Software for Economists

Author: Alan G. Isaac
Last modified: 2018-06-25
Most recent (GitHub) version:
Caveat lector: do not assume the information below is accurate. In particular: the advice below comes without any warrantee whatsoever, and please confirm licensing information with the software copyright holders.

The following software list is idiosyncratic. I have tried to make it generally useful for economists. I favor cross-platform solutions, since I view the current near monopoly of the end-user operating system as a momentary historical aberration. Suggested changes and additions are welcome. A resource complementary to this page is RFE's Software page.

Two requests: Thank you.

Text Editors

If you write code, use a decent text editor! All other things equal, multi-platform solutions are always preferable. Nowadays having an editor that nicely handles unicode (UTF-8) is pretty important. (See Unicode hints.)

I recommend Vim. It is highly configurable, available on essentially every platform and quite wonderfully designed. On top of all this, it is charityware: you pay nothing for this excellent software, but users are encouraged to make a donation for needy children in Uganda. However, many good choice exist.

Some cross-platform editors:

Mode-based editing rocks! Try the free text editor Vim. (Or possibly NeoVim.) Just be sure to do the tutorial (or possibly this online tutorial).

If you feel unready to start mode-based editing, you can still try a modeless configuration of Vim called Cream.

See Vim user hints.

Python users may want to consider the interesting and powerful Python Tools for Visual Studio in Visual Studio Code. (There is additionally a Vim editing mode.)

The elephant in the room is GNU emacs: big, powerful, and free. (Vincent Goulet offers a modified emacs that is easy to install across platforms and is particularly useful for LaTeX users.) Some prefer the closely related XEmacs (but development on that has slowed). Many users will find the somewhat less powerful but much smaller mg3a provides sufficient emacs style functionality.

There are many cross-platform alternatives. Here are just a few of the interesting ones.
Lightweight but powerful; the default text editor for the GNOME desktop. (License: GPL)
Lightweight; based on Scintilla. (License: GPL2+)
Small and fast. (License: GPL)
Fast and powerful, with a particularly powerful macro language (S-Lang).
The Hessling Editor (THE)
Gets rave reviews; uses the very powerful REXX scripting language; has some unusual and excellent features (e.g., the all command, which Vim approximates with folddoopen).
Based on Scintilla (a cross-platform editing component); provides scripting with the Lua programming language.
Komodo Edit
The same editor used in ActiveState's commercial Komodo IDE!
Lapis editor,
Allows simultaneous selection of discontinuous segments of text.
Early entrant into collaborative editing.

Windows users:
The NotePad editor is always available as an accessory program, but it remains too limited for serious use. If you like the basic interface, try the fast and powerful Scintilla-based NotePad++. (License: GPL) For pure speed on a Windows platform there is the powerful VEdit. If you can afford it, there is also Lugaru's Epsilon, a powerful and fast yet lightweight editior with an extensive emacs-style command set.

Aside: you can still get the traditional vi editor, which has been free and open source since 2002.

Very specialized: the Java based Record Editor may be just what you need if you are working with fixed-field-length data files.


Monospaced Fonts

Once you have chosen your editor, find a decent monospaced font. (Crucial for programming; always examine this character sequence: Il1aoe0O.) My favorite is Andale Mono, a TrueType font that I find exceptionally readable even at small font sizes. There appears to be more than one legal yet free distribution, due to the original licensing terms, but you can also purchase the font. My second choice is the Unicode DejaVu font family. (Derivative from Bitstream's Vera Fonts; available as a Latin-Cyrillic-Greek subset; License) Many people like the free and open source Inconsolata.

Before listing more monospaced fonts, I must note the wonderful Scientific and Technical Information Exchange (STIX) font project. These are "a comprehensive set of fonts that serve the scientific and engineering community". and are available under a royalty free (and fairly liberal) license. (TeX users should check these out!) The fonts were released in May 2010 and updated in 2016.
Here are MiKTeX installation instructions from Ulrike Fischer:
  1. Unpack the zip-file somewhere outside miktex. You will get a folder stix-tds
  2. Add the folder stix-tds as a new root in miktex - settings, tab "roots"
  3. Activate the file either in a document with \pdfmapfile{} or globally by following the instructions in the documentation
  4. Test the font with this document:
    %\pdfmapfile{} %if the map file is not activated globally
    Some text, and a math formula \(a+b=\sqrt{c}\).

A few more freely available mono-space fonts ... Google’s Noto Sans Mono is getting very favorable reviews. Most systems include some version of CourierNew, which is mostly adequate, or LucidaConsole, which is pretty good. (Unfortunately, neither adequately distinguishes 0 from O.) Another good choice from Adobe is Source Code Pro,which in regular weight looks quite a bit like Andale Mono, but which somewhat overweights its boldface font. (OpenType, Latin-1 only; License: Open Font License 1.1) Another popular font is Envy Code R. (True Type, Latin-1 and Latin-1 Supplement; License: gratis but not redistributable) For more choices, filter on mono-spaced fonts at Google Fonts or Font Squirrel. User ratings of “programming fonts” can be found on Slant. The GNU FreeFont mono-space fonts are available as OpenType or TrueType. Some people like Anonymous (OFL). There is also a more narrowly focused but very useful APL font (SImPL.TTF). Those needing a small font-size might consider the ProFont bitmap font, which looks great at 9pt. (License: MIT)

Good commercial mono-space fonts:

Other Useful Fonts

Gentium Plus is a high-quality FOSS unicode font set with some math-symbol support.

The Arev fonts are designed to be useful for presentations containing math.

Spell Checkers

The most natural approach to spell checking is modular, but many editors nevertheless include their own spell checker. If yours does not, there are many free options. Perhaps the most obvious cross-platform option is Aspell. (If you will use the Windows release, see Aspell hints.) Aspell derives from the more famous ispell. (There seem to be separate implementations of ispell for Unix and OS/2 and ispell for Win95/NT. You can also try jspell in a DOS environment.) If you need a spell checker for general Windows applications, try All-Purpose Spell Checker or Spell Checker for Edit Boxes. (They are free.)

Vim 7+ includes a spell checker. You can also integrate Vim with aspell, which is my preference.

Dictionaries and Thesauri

These days, your favorite dictionaries and thesauri are probably available free online. My favorite crowd-sourced dictionary is Wiktionary, which is remarkably good, even for etymology. My favorite commercial dictionary is The American Heritage® Dictionary of the English Language. The DICT Development Group offers a free online dictionary that provides a fast, convenient interface to the Collaborative International Dictionary of English. In addition, you can submit a query for "yourword" directly as*&Database=*. Windows users can download for free the Freelang translation dictionaries.

Aiksaurus is a powerful, free, cross-platform thesaurus. Windows users can download WordWeb for free.

Stream Editors

The only stream editor I have experience with is sed, which is very fast, powerful, and cross platform. (License: GPL) For an introductory tutorial, try SED: A Non-interactive Text Editor. If you decide to use sed a lot, you may wish to try sedsed, and Python-based sed debugger.

Producing Scientific Documents

Basic Considerations

Scientific documents will generally include typeset mathematics. Proper typesetting aids communication, and ISO31/XI therefore provides typesetting standards. (See Claudio Beccari's discussion, but also the thread on StackExchange.)

There are three basic approaches to producing mathematical documents.
  1. WYSIWYG: What you see is what you get. Generally this means you will use a word processor that provides support for mathematics. Your printed document will look very similar to the document view in your working environment. (Which often means it is not typeset very well.)
  2. WYSIWYM: What you see is what you mean. Generally this means that your working environment will be a user-friendly graphical interface to an underlying typesetting language. Your printed document will look somewhat similar to the document view in your working environment, but substantial formatting will be added to the document when it is typeset.
  3. WYSIYM: What you see is your markup. Generally this means that your working environment will be your favorite text editor, and you will create a plain text document that incorporates markup commands to induce formatting and the typesetting of special symbols. Your printed document will look quite different from the document view in your working environment. Substantial formatting will be added to the document when it is typeset, and your markup commands will not be visible in the typeset document.


A good cross platform solution is Open Office, which includes the oomath formula editor. Open Office is free and open source software.

Another cross platform solution is AbiWord, which brilliantly uses MathML as its internal representation of mathematics. AbiMath can import MathML equations. AbiWord also support itex for equation input, which is quite wonderful.

For Windows users, the most common approach is Microsoft Word with embedded OfficeMath equations. Unfortunately for old documents, new versions of Word no longer support Equation Editor 3 or earlier. Unfortunately, many journals do not allow embedded equations in submitted articles. This is a rather clumsy environment if your writing includes much mathematics, but it is widely used.


Some working environments allow you to see your mathematics in a form closely related to how it will be printed, while reserving polished document formatting to a separate "typesetting" process. Anyone who has printed a web page is familiar with the difference between the screen display and the typeset printed version (or "print preview") of the document: just imagine directly editing such a screen display before typesetting your document. These are sometimes called WYSIWYM (what you see is what you mean) environments.

The most popular WYSIWYM environments are interfaces to an underlying LaTeX document. In such an environment, you will see for example the glyph β instead of the LaTeX code \beta.

The only cross-platform implementation I know of is LyX. Binaries are available for Linux, Mac, and Windows users. LyX can interface with the Maxima computer algebra system. (Just add the path to Maxima in Tools -> Preferences -> Paths.)

An interesting alternative is TeXmacs.

Windows users appear to have a number of commercial choices, the most popular of which seems to be Scientific Notebook and Scientific Word. (Scientific Notebook is not marketed as a LaTeX front end, but as of 2004 it still saves in LaTeX format and with a tiny amount of editing can be used that way.) See Scientific Word user hints.

Mathematica notebooks can also be used as an WYSIWYM interface for mathematical documents, with some support for both LaTeX and MathML export.

In the WYSIWYM world, MathML is an obvious alternative to the LaTeX as an underlying document format. (In principle, this should not matter much to a user, who should be able to work in the GUI environment and not worry much about the underlying format.) The Amaya browser/authoring tool enables the WYSIWYM production of mathematical documents, uses MathML as the underlying format. Unfortunately, I find Amaya slow and clumsy for the production of mathematics, and development has stopped.


The most powerful way to produce mathematics is to learn a markup language for type-setting mathematics. In this case, what you see is your markup (WYSIYM). This also provides a big advantage for the process of writing: you get to use your favorite text editor and thereby get access to powerful editing facilities. The most popular choice is LaTeX, which is almost the universal WYSIYM choice among mathematicians, physicists, and economists. It is also my personal choice. An alternative is Lout, which is much less widely used but has some strong advocates. A full Lout installation is remarkably small: the download is less than 2Mb. The math in a Lout source document looks a lot like math in the EQN language. (Speaking of which, some people insist that EQN and GROFF remain the right way to typeset mathematics.)

If you compile your LaTeX documents with LuaLaTeX, you can use the unicode-math package. This allows you to enter math as UTF-8 characters, in addition to using the usual LaTeX codes. This produces much more readable documents, but only a few fonts have substantial support for the Unicode math characters. (The unicode-math documentation includes a list of fonts, including TeX Gyre Pagelle Math and Asana Math.)

Meta Formats

There are a number of efforts to provide a meta-format from which both LaTeX and XHTML/MathML documents can be generated. The most notable and currently developed is probably GELLMU, although tbook is also interesting. I like the much simpler restructured text (reST), which allows inline and display math in LaTeX format. (See my reST hints.) Citation handling is currently (2005) weak, but bibstuff provides some basic capabilities. Math support is improving. It may also be worth trying reST plus Mozilla's itex for web math. Other metaformat options include: Almost Free Text (AFT), MarkDown, TextTile, and setext.


reST to HTML: just use the rst2html writer. If you have included math in the document, use the rst2mathml writer.

reST to PDF

The usual way of creating PDF from an reST document is the following: use the rst2latex writer (or the rst2latexmath writer) to convert the document to LaTeX format, and then use pdftex to produce the PDF.

rst2pdf produces PDF directly using the ReportLab library. Good math handling via Matplotlib's mathtext!

You can also produce HTML with rst2html and then convert it. Using Adobe Acrobat can work moderately well, either by printing from a browser or by directly importing the HTML document. Alternatively, you can use the xhtml2pdf converter (License: GPL or commercial) or the Prince XML converter.

LaTeX and Related Software

Background: LaTeX is a document formatting package based on TeX. TeX is typesetting software, arguably the best in the world (especially for mathematics). In this section I will concentrate on LaTeX and related resources.

To get started with LaTeX you will need two things: a text editor or IDE, and a LaTeX distribution (such as MikTeX, OzTeX, teTeX, or fptex). Amazingly enough, you can get everything you need for free. You can also get great help from other users, e.g. on comp.text.tex or the MikTeX users list.

AU Students:
For mathematical documents and for long documents (like dissertations!) you should use the LaTeX document preparation system. There is an AU Thesis Class, which automatically handles all the formatting of your dissertation or thesis. Instructions for using this with Scientific Word are included. Also see MacKichan Software's basic instructions and detailed instructions. You can also use this custom document class with LyX. That said, my recommendation is to us Vim with my MiKTeX menus, or with Vim-LaTeX.

If you are used to word processors, you may initially find the process of creating a LaTeX document a bit clumsy: you do not see the document formatting as you write. However for long documents this is actually an advantage, which you will grow to appreciate. You will create your LaTeX input file with your text editor. In addition to your text, you will add certain LaTeX formatting commands to this input file. You will then use your LaTeX distribution to process this input file, which will produce a device independent (.dvi) file that you can view and print. Details follow.

One caveat:
When you are working on an academic article, LaTeX is an excellent environment. But when submission time comes, you may pick a journal that does not accept LaTeX. What to do? i. Article format generally does not matter until acceptance: hardcopy submissions are still usually expected, and otherwise PDF is generally acceptable. So there is really no problem until an article is accepted at such a journal. ii. Most journals that express a preference either ask for standard LaTeX or ask for common word-processor formats, but even in the latter case LaTeX is usually acceptable. iii. If you have written an article in LaTeX and it has been accepted by a journal that for some reason will not accept LaTeX, you can easily convert your article to a different format.

TeX Tools

MiKTeX tools have been released for UN*X systems.

TeX Distributions

There are many commercial and freeware LaTeX distributions for all platforms. Each comes with its own installation instructions, which you must be sure to follow step-by-step. For Windows 2000 and XP, I have had wonderful experiences with the freeware MikTeX, which has managed to make installation a virtually painless 15 minute process if you read the readme and the prompts.

MikTeX is currently the most widely used LaTeX distribution on the AU campus. We are usually behind a version or two.

TeX Live provides a comprehensive TeX system for most systems. BaKoMa TeX is an inexpensive distribution for Win95 that got some good press, and it includes PostScript versions of the Computer modern fonts. (A very nice feature.) You can purchase a TeX system (with a lot of other useful stuff) from NTG as 4allTeX.
Mac OSX users: use the i-Installer! You can also look at Gerben Wierda's teTeX port to OS X comes with an i-Installer installation program and gets great reviews. As a front end, consider Richard Koch and Dirk Olmes's TeXShop.

Learning TeX

Novices can get acquainted with LaTeX by looking at a tutorial. Once you are using some version of LaTeX, you are likely to find LaTeX: A Document Preparation System by L. Lamport to be indispensable. There are many other good books on TeX and LaTeX. The free lshort2e.pdf has most of what a novice will need. (Check for the most recent version.) More detail can be found in Making TeX Work and many other excellent books. If you need to write your own class (and you probably do not!) try clsguide.pdf.

Users with questions should check the LaTeX FAQ, and the LaTeX homepage. If you are sure you questions are not answered there, ask on the comp.text.tex newsgroup. CTAN and the CTAN-Web, also offer a wealth of information about TeX and LaTeX. (Be sure to visit Graham Williams's catalogue of information.) You may also find some useful TeX Information at UCC. If you get truly ambitious, Victor Eijkhout's TeX by Topic is now available online. (Most users will never need such a detailed understanding of TeX, however.)


Adding drawings to LaTeX documents

LaTeX provides a basic picture environment that is pretty powerful. Still, you will often need more. You can easily include PDF, JPG, or PNG images in your documents. See Graphics for LaTeX for more details.

LaTeX Format Conversion

See the more complete list at

From LaTeX

Extensive conversion capabilities may be provided by pandoc. (Language: Haskell; License: GPL) See this blog for a LaTeX-to-Word example. One interesting possibility is plasTeX, which parses LaTeX for output in various formats.

If you use LyX, there is a LyX (via LaTeX) to OpenOffice format converter. (See the same URL for a Mac front end.)

LaTeX to PDF

You will often want to produce PDF files from your .tex files. (The PDF files can be read by anyone with access to a PDF viewer.) I have had excellent results using dvipdfm, which is included in most LaTeX distributions. Many people report excellent results using pdflatex, which is part of most LaTeX distributions. Also see Markus Kuhn's tips on how to use (La)TeX to effectively produce highest-quality PDF files.


Note that a successful translation into HTML or MathML can subsequently be read by any modern word processor. (Of course it will not look as good as in LaTeX!)

You can also go LaTeX to HTML via PDF:
LaTeX to PostScript
LaTeX to Word or Word Perfect
LaTeX to MathML

To LaTeX


Most existing presentation applications have serious limitations. On the other hand, for short one-off bullet-point presentations, with a few images, these applications are very easy to use. (E.g., you can just paste your images into a variety of templates.)
Two very interesting notebook-based approaches to presentation building are Mathematica and Jupyter. Mathematica is a powerful commercial symbolic algebra package which supports interactive presentations using the Wolfram Language. Mathematica truly pioneered notebook interfaces, which have since become very popular. Jupyter is a FOSS notebook web application that can serve as a front-end to dozens of languages, including Python (via the IPython notebook). The Jupyter Book project provides tools to convert a collection of Jupyter notebooks into publication quality documents.
For speed and ease of editing, it is nice to create presentations with a text editor. This means implies audio and video content will be included by explicitly linking to the underlying files. Some obvious candidates are

A fuller discussion of options can be found in Wiedman on Screen Presentation Tools. Here are some options using reStructuredText.

- rst2pdf
  (Good math handling via Matplotlib's mathtext!)
- rst2s5
- Bruce
  (You can embed a Python interpreter in your presentation page!)
- rst2odp
- rst2beamer
- rst2slidy (cannot find it, but here's Slidy: )

Roberto Alesina mentions that if you choose rst2pdf then fancy transitions are hard, handouts have to be a separate document, there is no way to easily center something vertically, and you need to create different versions for different screen aspect ratios. (On the other hand, everything end up in a single file, your PDF can play almost anywhere, and you know exactly how it will look.)

If you know LaTeX, the beamer is great, and you can compile beamer presentation directly with pdftex. (I found getting started with the beamer document class using MiKTeX's pdflatex to be completely painless.) Some people prefer powerdot (which supercedes Prosper). With powerdot you will need to go .tex to .dvi to .ps to .pdf. (Many integrated TeX environments can automatically implement that sequence for you.) Some people like the commercial Utopia presentations bundle.

There are a few open source presentation programs that are not LaTeX based. They are more like using PowerPoint. OpenOffice includes a presentation program. GNOME Office includes Agnubis, but it is still work in progress.

Another simple but powerful possibility. The ReportLab Toolkit includes PythonPoint for presentation creation. It is amazingly powerful, but a bit complex. To make things much easier, you can very simply generate PythonPoint documents from restructured text (reST). Once you have the software installed, this is about as easy as it can get!

Presentations via RestructuredText

You can produce very nice presentations using reStructuredText with one of it presentation oriented writers.

S5 [2] can simply create beautiful browser based presentations. Instead of editing HTML you can use reST and convert the text file to a slide show with rst2s5. Also see S5 Reloaded and AJAX-S.

Falling short of a presentation package but still potentially useful for report writing is PyReport, which compiles a Python script and its (textual and graphical) output into a PDF file.

Last but not least: for incredible animated presentations, the technically inclined may consider Slithy.

Dynamic Annotation

Often it is helpful to be able to annotate a page during presentation. Windows users can use Zoomit. (Proprietary but free.)


One place where a good GUI is nice is for the visual formatting of tables. Here a nice GUI can provide a real advantage over markup. This need not mean abadoning LaTeX.

LyX supports table editing. (You can also copy table data from spreadsheets to LyX.)

(Commercial options include Scientific Notebook and possibly LaTable.) When the structure and content of the tables is the dominant consideration, however, LaTeX tables can often be easily and automatically generated. For example, many econometric packages will produce LaTeX formatted tables of results.

If your chosen econometric package does not support LaTeX table generation, you still have a variety of options.

Meta Formats for LaTeX

Other TeX Resources

There are extensive citation management resources.

Compiling TeX

Agent-Based Simulation

gridworld is a Python module providing some of toolkit functionality of NetLogo. (E.g., spatial and stationary agents; simple GUI construction with button, sliders, plots; etc.) Examples of the use of gridworld can be found in a supporting paper, along with a set of template models for exploring agent-based modeling.

Currently the best platform for introductory exploration of agent-based models is NetLogo, which includes an extensive Models Library. While NetLogo's statistical and mathematical support is limited, you can link it to R or link it to Mathematica. Additionally, NetLogo supports Java extensions and can be run by Java programs.

If you are new to NetLogo, I have written a basic overview.

Here are some useful NetLogo books:

  • Best overview and general introduction:
    Wilensky, Uri and William Rand (2015). An introduction to agent-based modeling: Modeling natural, social and engineered complex systems with NetLogo. Cambridge: MIT Press.
  • Best applied introduction:
    Steven F. Railsback; Volker Grimm (2011). Agent-Based and Individual-Based Modeling: A Practical Introduction. Cambridge: Princeton University Press. ISBN 978-0-691-13674-5.
  • Best short introduction:
    Gilbert, Nigel and Klaus G. Troitzsch (2005). Simulation for the Social Scientist, Second Edition. London: McGraw Hill.
  • Most focused on Economics:
    Hamill, Lynne and Nigel Gilbert (2016), Agent-Based Modelling in Economics, Wiley.
  • David O'Sullivan; George L.W. Perry (2013). Spatial Simulation: Exploring Pattern and Process. Wiley-Blackwell. ISBN 978-1-119-97079-8.
  • Britt Anderson (2014). Computational Neuroscience and Cognitive Modeling. London: Sage. ISBN 978-1-4462-4930-7.
  • José M. Vidal (2010). Fundamentals of Multiagent Systems Using NetLogo.
  • After NetLogo, perhaps the most popular multi-agent simulation toolkit is Repast. (License: New BSD; Language: Java or C++)

    Swarm is a famous and flexible platform for multi-agent simulation projects, originally written at the Santa Fe Institute in the mid-1990s. (Language: Objective-C or Java; License: GPL) Although the platform is extremely capable, use of Swarm seems to be declining, and there appears to be little ongoing development. The Java Agent-based Simulation (JAS) library attempts to reimplement Swarm in pure Java.

    Evolving Objects is "is a templates-based, ANSI-C++ compliant evolutionary computation library". (License: LGPL)

    Scientific Programming Languages

    Much scientific programming is done in very high level languages. Two factors to consider in making a language choice are what software in your area is already written in the language, and how much original coding you plan to do.


    A key reason why Python rocks for research is the NumPy array multi-dimensional library (License: BSD). NumPy is a very high quality library, and it has become my favorite approach to matrix (and multi-dimesional array) programming. This Python package is written as a C extension, so it is very fast. This means you can have full access to the ease of use and to the object-oriented power of the Python programming language while doing serious scientific programming.

    Additionally, an extensive scientific computing environment is provided by SciPy. (License: BSD; Dependency: NumPy.) For publication quality graphics, add the excellent graphics package Matplotlib. (License: BSD; Dependency: NumPy.)

    Most Python users will be best served by a Python distribution that packages together everything they need as a single install. For example: NumPy, SciPy, and Matplotlib (as well as much more) are included in the following distributions.

    Many scientific programming applications can be found on the SciPy Topical Software Wiki. ScientificPython offers some useful scientific and visualization modules. The are also some additional probability and statistics utilities available. Pycoin provides a Python interface to some of the COIN operations research packages. Additional functionality can be obtained from the NAGpy interface to the NAG Fortran library or the PY-ML interface to Mathematica. Python also has modules for interacting with gnuplot and R. You may also want to see the Python math tools at the Vaults of Parnassus. If you need to build GUIs as part of your Python project, consider the Boa Constructor. Additional Python stuff you might find useful for scientific computing: PAIDA (a Python implementation of the AIDA abstract interface for data analysis), and the resources listed at StatPy, and possibly various AI resources. Modular toolkit for Data Processing (MDP) is a Python data processing library implementing Principal Component Analysis (PCA), Independent Component Analysis (ICA), Slow Feature Analysis (SFA), and Growing Neural Gas (GNG). Finally, it is worth poking around CheeseShop.

    Being able to write code in one matrix programming language and run it in another environment can be useful. Examples include Treiber's pym Python program for converting MATLAB to Python, Cameron Rookley's gtoml Perl program for converting Gauss code to Matlab, and Christopher Stawarz's i2py script for coverting IDL to Python. If you want to create a translator, consider PLY (a Python implementation of lex and yacc).


  • Julia is a promising “high-level, high-performance dynamic programming language for technical computing”. (License: MIT) This is a very interesting and powerful language, which is already popular among economists. Follow the platform specific installation instructions. It is also possible to add Julia to a Jupyter notebook.
  • Java for Scientific Programming

    Java does not seem to have any eqivalent of Python’s SciPy package for scientific programming. Pieces of the functionality can be obtained as follows. (But stick with SciPy if you can.)

    Other Free Matrix Programming Languages

    Support for Matrix Programming in Other Languages

    I mention only a couple important items from the endless possibilities.

    VectorZ provides fast double-precision vector and matrix math and supports n-dimensional arrays.
    Parallel Colt provides support for dense and sparse linear algebra. This is a dependency of Incanter, which provides statistical data processing and visualization functionality.
    ND4J is a library for scientific computing on the JVM. (License: Apache 2.0) A very promising project, but (as of 2015) it is far from mature.
    FLAME is a high-performance dense linear algebra library.

    Commercial Options

    Popular commercial options include Mathematica, MATLAB, GAUSS, and IDL. For econometrics, the biggest collections of extant code are in MATLAB and GAUSS, although Python and Julia code have perhaps caught up. Of these two, Gauss has the advantage of real array broadcasting, while MATLAB relies on the relatively clumsy repmat and bsxfun for broadcasting. Recent versions of Matlab supports classes and GAUSS has structs, but neither language is truly object oriented, and both often copy array data when it is unnecessary (instead of referencing it). If you are going to write a lot of your own code, avoding needless copies can be critical. You can get this very nicely with the NumPy Python library.

    Interactive Data Language (IDL) is popular among astronomers and others doing intensive image processing. You can run most IDL programs using the GNU Data Language (GDL). You can convert your IDL programs to Python with i2py.

    If you are forced to work in Matlab you can still have access to Python via pymex, which "embeds a Python interpreter into Matlab".

    GAUSS and Related Resources

    See GAUSS user hints.

    Among matrix-oriented programming languages, GAUSS was once the first choice among U.S. econometricians. (Python and Julia now appear to be taking over.) It is known for speed and ease of use. Its collection of built-in statistical procedures is large but not as extensive as specialty languages such as S-Plus/Gnu-S, although this can be offset by its speed and by available free and commercial packages (such as GAUSSX).

    Getting Started with GAUSS

    Jens Peter Steffen offers a detailed comparison of GAUSS with MatLab and Ox.

    General Purpose Programming Languages

    This is an extremely short selection of possibly useful free programming languages.

    If your emphasis is computational, you will want to visit NIST's Guide to Available Mathematical Software and the Netlib repository. Also, see Jack and Hatem's list of free linear algebra software.


    I use Python whenever possible, which is almost always. (The main exception is when execution speed is at an absolute premium in computationally intensive projects.) Traditionally thought of as a scripting language, Python has evolved into an extremely powerful very high level general purpose programming language. It is still useful (indeed, a first choice) for scripting. Use it to find out why. In addition, Python has excellent support for scientific computing. The online documentation is excellent. If you lack programming experience, try Python for the Absolute Beginner and Byte of Python before trying the Python tutorial. Also, if you prefer video tutorials, Microsoft has assembled a multipart video introduction.

    One other note: although it is an interpreted language, you can compile Python. Interesting graphics and numerical libraries have been developed for Python.


    L3 is a large subset of Python along with a GUI interface that provides a "worksheet" like environment. (License: BSD.) Easy and powerful.


    For C/C++ see Bjarne Stroustrup's list. However, I will list a few free C and C++ compilers that are reputed to be quite good. If you are new to C, see Standard C by Plauger and Brodie.
    Btw, if you program in C/C++, you should probably be using the Cscope code browser; there is a native Win32 port of Cscope, which works with Vim's Cscope interface. There is online access to Numerical Recipes in C, but I have had occasional trouble getting through.

    Hackable Games in C: NetHack, derived from the venerable Hack game. Read about in in IEEE Spectrum. See the GitHub repository if you want to hack it.

    Other Free Languages

    Scripting Languages

    For scripting, the most popular choices currently seem to be Python, Perl, and Ruby. There have also been some interesting applications of the embeddable S-Lang interpreter and programmers library. Lua is especially small and portable, and is used in the up and coming LuaTeX extension of pdfTeX. (See my LuaTeX hints.)
    It is also worth mention that Jython implements Python as Java, allowing the ease of Python scripting to be combined with the power of Java. (As a very cool plus, Vim supports scripting with Perl, Python, and Ruby.)

    More Scripting:
    Active Perl is available for many platforms and is getting excellent reviews. Python is growing in popularity, as is Ruby (which finally has a Windows installer).

    Languages for Scientific Computing

    Haskell is worth considering. Graham Hutton complete Youtube course on Haskell. The massiv package supports multidimensional array operations.

    We have alread discussed C above. Certainly C++ and Fortran have to be mentioned. The GNU Compiler Collection (GCC) includes front ends and libraries for these (as well as Objective-C, Java, and Ada).

    Getting started with Java:
    Download the Java Development Kit (JDK), which includes the Java Runtime Environment (JRE). On Windows, you install this by running the downloaded executable. Then download Eclipse, which is an IDE that is popular for Java development. You “install” Eclipse simply by decompressing it. (E.g., on Windows, just unzip it. It does not make any registry changes.) Eclipse comes with tutorials. For a simple overview of installation, consider this online video.

    For very simple introductions to Java (suitable for those without programming skills), try David Eck’s free book and/or Java Programming for Kids, Parents, and Grandparents.

    The F programming language is a carefully selected subset of the most useful Fortran 95. You can compile F with the g95 compiler: g95 -std=F my_program.f95, available for free from G95.

    FreePascal includes very substantial Delphi compatibility. (License: LGPL and GPL)

    Ada, which is a preferred languge for the security conscious.

    Eiffel supports desing by contract. There is now open source licensing of Eiffel Studio. (License: GPL)

    MzScheme is a free and open source implementation for Linux, Mac, and Windows. (As a very cool plus, as of version 7 Vim supports MzScheme scripting.)

    Scheme, and R. Kent Dybvig's The Scheme Programming Language, Second Edition is now online.

    Sather-K is Eiffel based but adds innovations (and is free).

    Visual Eiffel Lite is free. Documentation is in Eiffel: Object Oriented Programming

    Other Useful Languages

    PostScript is discussed under Drawing Languages.

    Other Useful Statistical Software

    See the free stat software pages by John C. Pezzullo and Gene Shackman . (The latter offers some package comparisons.) Also see the QM&RBC Software page, Spatial Statistical Resources, the Econometrics Laboratory Software Archive (ELSA) at Berkeley, and many others.

    Build Systems


    Resources in this area are extensive. I will only mention a few idiosyncratically selected items. (For linear and non-linear programming, see also the separate list of solvers.) Do not forget to think about scaling issues. (Here is one discussion of scaling: go to the Aptech web site, click on Resource Library, then click on White Papers, and download Optimization with the Quasi-Newton Method.)

    Python Specific

    Visualizing Data

    Visual analysis of data remains very useful, and some innovations in this area are recent. Classic texts include: The Visual Display of Quantitative Information by Edward Tufte and Visualizing Data by William S. Cleveland. You can get the data and S scripts for Cleveland's book. Cleveland (with Becker and Shyu) developed Trellis Graphics™ for data visualization, which has been implemented for a number of platforms including S-Plus and R. (The name “Trellis” is trade-marked, so these are often referred to as lattice plots.)
    Other resources: Daniel Carr's page.

    For interactive data visualizations, you must consider Vega and Vega-Lite. These visualization tools implement a grammar of graphics and a grammar of interactivity, mostly using declarative specifications in JSON. Vega-Lite provides a convenient, very high level access to this grammar. Many languages now support this, including JavaScript, Python (Altair), Julia (VegaLite.jl), and Haskell (hvega). Voyager 2 is an interactive data exploration tool built on top of Vega. Some languages support this, including JavaScript, Python (Altair), Julia (DataVoyager.jl).

    Modrian "is a statistical data-visualization system written in JAVA." Good for categorical data and for maps. Plots are interactive. There is a supporting paper by Martin Theus.

    Symbolic Mathematics/Computer Algebra Systems

    Find a more list of Computer Algebra Systems (CASs) on Wikipedia. This section discusses only a few particularly interesting applications. For most users, commercial CASs will be the right choice, but some interesting FOSS solutions exist.

    Commercial CASs


    Web Resources

    Math on the Web

    There are a number of way to provide access to nicely formatted math on the Web. The most widely used is to provide PDF documents that contain nicely formatted math. The best way to produce good looking math in PDF documents is with TeX based software. Acceptable results can sometimes be achieved by converting a word processor document with equation editor objects embedded in it.

    The other major alternative is to somehow produce math in a web page. The stand for this is MathML, but as late as 2021, FireFox and Opera remain the only major browsers with good support. (Be sure to download the STIX fonts.) Producing MathML by hand is impractical, so there has been a search for feasible ways to rely on MathML. One of the most interesting approaches is to use reStructuredText with the rst2html5 writer. This approach uses LaTeX syntax to input mathematics. Possible output formats include MathML and MathJax. This is probably the simplest powerful way to put math on the web.
    Another interesting approache is ASCIIMathML, which simply includes LaTeX-like math in a normal HTML file. (A clever javascript renders the math.) Mozilla based browsers such as FireFox have excellent native MathML support, so they work well with ASCIIMathML. (But you need to install the math fonts. See the Wikipedia MathML article for a summary of support by other browsers.

    One other alternative is to convert LaTeX documents directly to XHTML+MathML documents. This can be done with LaTeXML, TeX4ht, or with the closely related TeX4moz. TtM looks like a similar effort.


    If you want to test your browser for standards compliance, start with the Acid 3 tests. You might also want to look at handling of MathMl, SVG graphics, or even a very simple XHTML. (Be sure to look at the tests where the document is served as XML.)

    I like three browsers: FireFox, Opera, and Chrome. Of these, I give FireFox the nod for being both open source and mature, but Chrome is also open source and is considered faster. Microsoft Edge is slowly becoming usable; it is certainly much more standards compliant than the old Internet Explorer. Try them all on the Acid 3 tests. An interesting "new kid on the block" is uzbl, which emphasizes speedy browsing but is highly extensible.

    See the Opera hints below. See the FireFox hints below.


    The ePub3 standard introduced MathML support. Quite a few ePub readers support MathJax.


    Please follow basic email etiquette [1|2], by quoting selectively and only as necessary to provide context for your reply. When appropriate use only ASCII text for your email. Finally, if you must use Outlook, do your interlocutors the courtesy of turning off TNEF attachments!

    Get a decent email client. Since there are so many good options, choose an email client that has a multi-platform presence and, most importantly, lets you use your own editor.

    Using your own editor will repay you handsomely, saving time and frustration. Here are examples of things any good editor will allow you to do simply. Preprocess your email to compensate for bad practices by the sender, delete or reformat blocks of text (e.g., a line, a sentence, or a paragraph) without reaching for a mouse, adjust capitalization (e.g., change caps to lower case or vice versa, or convert to first letter caps), and quickly insert text from any other file (i.e., read it in, so you do not need to cut and paste). One could propose that every application in which you need to edit text should include such facilities. E.g., there are many email clients with built in editors that have many such facilities. But then each time you change applications you will have to learn a new editing environment. Instead, whenver possible, you should use applications that provide support for an external text editor. By sticking with your preferred editor you do not have to learn a new editing environment and scripting language each time you switch applications. That of course is the point of modularity.

    See Wikipedia's list of email clients for a sense of what's out there. I like the free, cross-platform PINE, which is keyboard oriented, highly configurable, and has a good Windows version. But (despite occasional instability on Windows) I generally use Thunderbird. Another email client that is now cross-platform is Claws Mail. Big plus: unlike most Windows email clients, Thunderbird, Alpine (formerly pc-pine), and Sylpheed let you use your favorite editor to edit your email. Thunderbird supports external editors by means of a plug-in.

    See the Thunderbird hints and Alpine hints below.

    Another cross platform solution is now The Mutt Email Client—long popular among Unix users, and available to Windows users. (You can also get mutt along with fetchmail, and sSMTP in a unix-mail package for Windows, but you'll have to install Perl and Cygwin.) An email client that is not free but gets great reviews is The Bat, and but it seems to not yet support external editors. Another popular cross-platform email client is Mulberry, and it is available on many platforms. (License: Apache 2.0) One last note on email: if you use Vim as your editor, you might consider using vimail, a simple mail reader for Vim.

    If you are in an environment that forces you to use Outlook, you can at least write your email with a decent editor.

    POPFile sorts your email and filters spam. It is free and cross platform.

    Document Formats

    What are some good document formats, and how can one transform between them?

    For writing math, nothing beats LaTeX. However, you can use reStructuredText with writers that support math to produce both LaTeX and XHTML MathML.


    Writing HTML is easy: learn the basics. Conversion to other formats is possible. Most word processors will read HTML and automatically convert to their own format. You can also convert HTML to PostScript (if you have access to Perl).

    After learning a few HTML basics, you can learn much more from the materials at the World Wide Web Consortium W3C). Most of the time you should be able to use XHTML Basic and Cascading Style Sheets (CSS). Here is some CSS Background, and the wonderful chapter 1 of Cascading Style Sheets 2.0 Programmer's Reference by Eric A. Meyer. For more information and links, go to the HyperText Markup Language (HTML) Homepage. Some authors may like to see the ISO standard as well.

    All modern Word Processors can export their documents as HTML: unfortunately, they often produce pretty bad HTML. You can get much better results with some of the HTML editors, such as the web development suite offered free by Evrsoft


    XHTML is a wonderful HTML related markup which provides for most needs when used with CSS. Unfortunately some browsers do not support it, unless you serve it as HTML. (That is the trick used in the present document.) Some authors discourage this practice, but it should work fine, as long as you validate your documents.

    XHTML: Introduction to XHTML, with eXamples" helpful | XHTML validation. With XML/XHTML (handled by an XML processor) you will be able to define your own entities in the doctype declaration. E.g., you could define <!ENTITY alpha "&#xxxx;" > (where xxxx = the numeric reference) to get around remembering numeric references. The XHTML+MathML document type looks particularly useful, and it is already used to provide math support in reStructuredText.

    Mathematical Markup Language (MathML) is now a wonderful reality. FireFox and other Mozilla based browsers offer native support. So does E-Lite: ICESoft's lightweight Java web browser offers full, native MathML implementation. Internet Explorer browsers currently support it only via plug-ins. Public discussion on features for math in HTML contexts takes place on To subscribe send a message to with "subscribe" as the content of the subject. There is a public archive. Also, consider WebEQ for Web math.

    XML: Perhaps the most interesting current XML development for non-programmers is OpenDocument Format (ODF). The is a promising open document exchange standard. The OpenDocument Format is an open, fully-documented, royalty-free standard for document exchange. It is supported by OpenOffice and the closely related LibreOffice. There is also a plug-in converter for MS Word (TM).
    By the way, docutils can write directly to ODF: see Dave Kuhlman's work (currently in the docutils sandbox).

    Transport Neutral Encoding Format (TNEF)

    What is that winmail.dat attachment you just got in your email? It is a proprietary Microsoft format that bundles together multiple attachments in a way that most email clients cannot unbundle. How obnoxious! So, what can you do about it?

    One possibility: ask your interlocutor for the simple courtesy of turning off TNEF attachments! (Instructions are the bottom of the article.)

    If you are unfortunate enough to receive a TNEF attachment from someone who cannot figure out how to configure their email client to send attachments that everyone can read, you may have to unpack the TNEF attachment yourself. The only cross platform solutions I know of are JTNEF and pytnef (License: LGPL). At this point, pytnef relies on the tnef command-line utility that is I think only available for unices (although it looks like it could be compiled for Windows).

    Email Scripts

    If you need to produce email in a scripting environment, Python offers all the basics in the smtplib module. You can even add MIME attachments with reasonable ease.

    News Readers

    Cross platform: Pan (GPL) or slrn or Gnus.

    Free for Windows: Xnews and Forté Free Agent.


    First let me offer an apparently outrageous suggestion: individual users can often substitute good habits for AV software, and even if you install AV software these good habits are crucial defenses. That said, whatever software you choose, you may wish to test it on the EICAR test file.

    i. use plain text to send and receive information whenever possible. All word processors and spread sheets can save as .txt. There is seldom a real need for using anything except plain text for email.
    ii. When plain text is not possible (because special formatting is required) you can almost always use HTML. All modern word processors and spread sheets can save as HTML. This is almost risk free, but to be really safe you can turn off Java and JavaScript in your browser preferences.
    iii. When page formatting is crucial, which it almost never is, use Adobe's PDF format. You may need a PDF "printer driver" to do this (such as Adobe Acrobat, Jaws, or some free alternatives such as PDFCreator).
    i. Many viruses are received as part of email attachments. Other malware (known as worms) will email themselves from an infected computer to many others. So the best practice would be to not open any email attachments. Unforunately, sometimes we need to open attachments. ii. If you must open an attachment, view the file extension. The default in Windows is to turn off your ability to do this, which means you can be tricked into thinking a dangerous Visual Basic script filename.jpg.vbs is really a safe .jpg file. You can fix by following this windows hint. For Windows users, especially dangerous extensions are .exe, .com, .pif, .vbs. Also dangerous are .doc and .xls. (Comment: dangerous file types can be included as ``attachments'' to .pdf files as well, which means you should not open such attachments with Adobe Acrobat either.)
    iii. When possible, avoid the use of susceptible proprietary file formats (such as Word or Excel). Related to this, avoid the use of Outlook unless you know how to turn off all its dangerous features. And if you simply must open a document in a MS format (e.g., because that's what a work colleague sent you), learn how to turn off the availability of macros. You should also be cautious with any other format that supports macros, which basically means you should avoid exchanging information in word processor formats (see i. and ii. above).
    iii. Never run a program file unless you know it came from an impeccable source. (For Windows users, this includes file extenstions .exe, .com, .pif, .bat, and .vbs.)
    iv. Windows users should disable autoplay, since it allows arbitary executable to run automatically on your computer.

    Additional Comment: Many new viruses are not detected by anti-virus software. Safe practices are necessary. Furthermore, if the recipient wishes to read what you send in their word processor, all modern word processors also import plain text (very safe), HTML (pretty safe), and PDF (usually safe) very simply.

    Multimedia Resources

    Since this is not a core interest of mine, I list only a couple resources.

    Attention Vista users: your operating system came without the Windows Media Player plugin. Fortunately, mozillaZine explains how to install the WMP plugin.


    Modeling Languages and Solvers

    Modeling Languages

    The modeling languages most widely used by economists are commercial: GAMS (with online documenation) and GEMPACK (with online documentation). Economists use these primarily for CGE modeling. (See these helpful CGE exercises with solutions.)

    Other alternatives are listed at A few are listed below.


    Naturally, many commercial modeling languages come with a variety of solvers. Here are some other solvers.

    Integrated Development Environments

    For Python, PyScripter has an excellent reputation. But it has limited support for threading. (Windows only; license unknown.)

    For Java, consider the Eclipse IDE. (Eclipse is a "general purpose" IDE.) Python users can also use Eclipse via the Pydev plugin.

    For C and C++, consider the Agide IDE. You can use Vim as your editor! On Linux/KDE, you can alternatively use KDevelop, which also support Vim as an editor.

    VimMate is a light GTK-based IDE that embeds the Vim editor.

    For Python:
    - the general purpose Leo IDE includes a unique and powerful outlining facility (Language: Python; License: MIT)
    - the relatively lightweight Eric IDE gets very good reviews. (Language: Python; License: GPLv3)
    - Jedi helps turn VIM into a Python IDE by adding autocomplete.

    For Android, Google has released Android Studio in 2014.


    ECMAScript is a standaradization of JavaScript, a dynamic, interpreted programming language whose origins are in client-side scripting in the browser. Most browsers support ECMAScript, and client-side scripting has been used to create economic applications. [#paine.stark-2004-cheer] These features made it a widely used language. Growth in language features, including support for functional programming, have made it a rather interesting language as well.

    If you are new to JavaScript, I have written a basic introduction.


    Java is a strongly-typed object-oriented language. If you want to program in Java, consider the Eclipse IDE. The JSwat debugger front end may also prove useful. If your application is mathematical you may be interested in the Java Mathematical Expression Parser (JEP). Other Java related resources can be found at

    Python is an obvious choice for server side scripting. You might consider Xitami as a cross platform server solution, or simply to test your scripts.

    Cross Platform GUI

    The obvious language independent FOSS choices currently (2005) seem to be TK and wxWindows, with TK being simpler, more mature, and a bit slower and wxWindows being faster and more powerful but less stable. Of course Java provides a language dependent solution.


    Python comes with the Tkinter module, which provides a Python interface to the TK graphical user interface (GUI). Tkinter documentation is spread around. For a very brief introduction, you can try Alan Gauld's tutorial. If you use Tkinter, consider using Python megawidgits (Pmw). (You can also easily add graphing extensions with BLT Toolkit. You may want to also get Mark Rivers's enhancements.

    tk_happy allows very fast, menu guided tkinter GUI creation.


    Until recently curses has been unavailable for Windows. Wcurses is a curses module for Python that works under Windows. There may also be some profit in considering curses applications on INTERIX.


    wxWidgets offers cross-platform API (C++ library) for writing GUI applications. wxPython is a Python extension module that wraps wxWidgets. Dabo is a cross-platform application development framework, written in Python atop the wxPython GUI toolkit. (Dabo capabilities resemble Visual Basic or Visual FoxPro, but it's cross platform.)


    Graphics Formats

    There is some useful background on the PyMol wiki.

    Encapsulated PostScript (EPS)

    Encapsulated PostScript (EPS) remains a very important standard for graphics inclusion in journals and books. It therefore pays to learn a drawing application or language that allows you to generate encapsulated PostScript (EPS). You can convert EPS to PNG or SVG. (See below.)

    Portable Document Format

    In many applications, EPS) is being replaced by Adobe's Portable Document Format (PDF). PDF) and page indepdence and support for new features, but loses the powerful PostScript drawing language.

    Portable Network Graphics (PNG)

    Currently the best choice for web graphics. (GIF and JPEG suffer patent problems, although these patents are finally expiring.) Before placing graphics on the Web, visit the Portable Network Graphics (PNG) homesite.

    MNG and APNG
    The animated version of PNG is called MNG, for Multiple-image Network Graphics. It supports looping (including nested loops), clipping, deltas, and other features, plus everything PNG supports-- including alpha transparency. Unfortunately implementations have been lagging. However FireFox supports the animated PNG (APNG) format, which is less complete but quite capable.

    Drawing Languages and Metalanguages

    With all the point and click drawing applications currently available, why bother with a drawing language? The answers are generally precision, speed, and fun, often in that order. This section concentrates on 2-D drawing, with emphasis on the PostScript drawing language, while mentioning a few other interesting possibilities.

    The so-called Grammar of Graphics has garnered attention because of its power and expressiveness. PlotNine is an implementation for Python. ggplot2 is an implementation for R. It is supported by two popular books: R Graphics Cookbook and ggplot2: Elegant Graphics for Data Analysis.


    PostScript is a device independent page description language. If you want to precisely render a vector graphic, you may well find it convenient to draw it yourself in PostScript. I offer an introduction to PostScript Drawing for the Sciences. (Still under revision; comments welcome!) See some PostScript hints below.

    Many people prefer to use a meta-language to produce their PostScript graphics. Most of these can also produce PDF. Some produce other formats as well.

    GhostScript and GhostView let you view and print PostScript files.
    Related: PLisp takes the basic structure of Common Lisp and uses it as an alternative representation of a PostScript program. While the basic graphic operators of PostScript are retained, Lisp-like function definition, control flow, and variable binding become available. Macros, defined constants, and function libraries have also been provided. The PLisp compiler generates PostScript code which is guaranteed not to contain errors in stack usage.

    Scalable Vector Graphics

    This popular format for web graphics can be useful when you need vector graphics and/or animation. It was supposed to be possible to embed SVG in PDF, and for a while Adboe worked actively on an SVG viewer (which is still available). This seems to have stalled, and indeed further development of the SVG standard unfortunately seems to have stalled. For extensive SVG manipulation and generation tools, see Batik (language: Java). An increasing number of graphics languages offer an SVG backend.

    Scalable Vector Graphics is a royalty free vector graphics standard for the Web, and it is a powerful animation format. (If you have simple, static graphics to produce, you might write them in PostScript and then translate using pstoedit.)
    One of many nice things about SVG is that it can be embedded into PDF, animations and all! (See this very cool example. Here is a perl script. Here is a Java converter. Here is a Python script. (If you can find Leonard Rosenthol's Acrobat plug-in, let me know. It seems to have disappeared.)
    Another nice thing is that you can move image generation to the client side.

    Browser support for SVG is improving but advanced features may be missing. Check your target browser. (You may find some useful information on Wikipedia.)

    Curren Kelleher offers an extensive tutorial on data visualization using SVG (via D3). The tutorial uses VizHub (MIT license).

    3D Drawing

    3D drawing: For simplicity and power, it will be hard to beat Sketch. (Langauge: C++ (binaries available); License GPL) Particularly helpful to LaTeX users is its support of tikz. (See Kjell Magne Fauske's nice introduction to Sketch, which includes full examples.)

    Try the Persistence of Vision Ray Tracer (POV-Ray). If you need high performance real-time 3D, consider Irrlicht. If you are an OpenGL user, superficie may be useful, although development appears quiescent. (License: GPL)

    3D Plotting

    3D plotting: For nice, accurate surface plots, there is probably no simpler solution than gnuplot. The SymPy plotting module is very interesting. QwtPlot3D appears very powerful, and PyQwt3D provides bindings for Python users who wish to produce scientific 3D plots. (Now supports NumPy arrays.). Other possibilities include Visual Python and Python Computer Graphics Kit. Also, OpenDX looks like a very powerful open source implementation of the IBM Data Explorer (DX), and there is also a Python binding. VisIt looks to be an extremely powerful way to produce 3D animations.

    The Visualization Toolkit (VTK) "software system for 3D computer graphics, image processing, and visualization" consisting of "a C++ class library, and several interpreted interface layers including Tcl/Tk, Java, and Python." (License: BSD-like, but containing some patented technology.)
    Hint: Find yourself a good interface to VTK. VTK provides wrappers for Python, Tcl, and Java, but the wrappers resemble VTK more than the supported language. Python users may like MayaVi, but this still requires VTK knowledge. For true ease of use, consider Enthought's mlab Python framework for access and manipulation of MayaVi2 objects. (Also note that PyVTK can write VTK files from NumPy data. TVTK provides a powerful and "Pythonic" interface to the Visualization Toolkit. With some feature resemblance to MayaVi, the possibly more powerful Paraview is written in Tcl/Tk rather than Python.

    More 3D plotting solutions:

    HippoDraw produces interactive graphics for data analysis. Hippodraw includes Python bindings and a Python extension module.

    Rendering Engines


    Simon Yuill offers a nice animation tutorial (with a Python emphasis).

    Pms.Blt provides a Python interface to the BLT Graph plotting library.

    Blender supports powerful 3D drawing and animation, which can be assisted by Python scripts.

    HarddiskOgg can record analog audio "directly to Ogg Vorbis/MP3/Wave with on-the-fly normalization". (License: GPL)

    HTML % includes a VIDEO element, supported by FireFox 3.5+ and Chrome 3.0+. (See Video For Everybody for usage hints.) This enables native OGG/Theora video playback in Firefox 3.5! For conversion of video formats, you might try HandBrake. (License: GPL)

    Plotting in Python

    See the SciPy TopicalSoftware Wiki for a more complete list of options.

    2D Graphics

    3D Graphics

    Network Graphics


    Graphics Format Conversion

    Convert to EPS

    A cross-platform standard for graphics is Encapsulated PostScript (EPS). EPS is sometimes the expected graphics format by publishers when you submit drawings for a book or journal article, although PDF is becoming more universal. A common file conversion problem remains taking a file produced in a different format and converting it to EPS.

    Convert to PNG

    PNG is the current standard for web images. If you have produced an image for the web in a different format, see the LibPNG list of converters. There are many options listed there, but if your image is PostScript or PDF, consider GhostScript.

    Convert to PDF

    GhostScript can convert many file formats, including PS to PDF. You can combine GhostScript with the Redirection Port Monitor (RedMon) to produce PDF files (see helpful directions). FreeDist is a Visual Basic program can automate this process for Windows users. FreePDF seems to offer similar functionality to FreeDist. PDF995 is another free (gratis) but not open source alternative.

    If you need to convert entire HTML documents to PDF, the easiest way to is to open it in a browser that supports conversion and simply save as PDF. (Chrome supports this as a Print destination.) There are also stand-alone converters. Try HTMLDOC. Perl users can try html2ps and then use GhostScript. If your document is well formed XML, CSSToXSLFO offers very interesting possibility. It converts an XML document, together with a CSS2 style sheet, into an XSL-FO document. The document "can then be converted into PDF, PostScript, etc. with an XSL-FO-processor", which is included. (License: Public Domain)

    Abiword allows you to convert/print from the command line. It reads ODF and RTF (which can be produced by Word).
    unoconv is a Python script that uses the uno/pyuno access to the OpenOffice API. You my need to start an instance of oowriter: problems were reported with the --listener option as late as 2009.

    Other Graphics Format Conversions

    Drawing Applications

    There is an almost limitless variety of high quality commercial drawing applications with pricing roughly linked to functionality. Rather than attempt to list or compare those, I will focus on a few cross-platform alternatives.

    Metalanguages for PostScript

    Python users have a number of fine alternatives. For serious 2-D drawing, I like PyX very much. PyX also has some support for 3D graphics. (See my PyX hints below.)

    While PyX is great for data-based graphs, I often find myself using Matplotlib for these. (Matplotlib is a wonderful 2-D plotting package which also supports powerful interactive graphics.) PyX and Matplotlib support the generation of EPS and PDF plots with LaTeX text. (Additionally, Matplotlib provides an impressive mathtext capability independent of LaTeX.)

    Matplotlib and Pyx currently take different approaches to math text. PyX uses LaTeX to typeset the math, so you must have a working LaTeX installation. Pyx generates PostScript, PDF, and SVG graphics with this approach. Matplotlib uses its own parser and layout engine to typeset math text with Type 1 computer modern fonts. You do not get the full power of TeX typesetting, but you can produce outputs in a variety of formats (including EPS, PNG, and PGF). Here are some details.

    Matplotlib supports spark graphs (sparklines) via the sparkplot module. (Sparklines can also be produced with PIL using the Sparkline Generator Web Application.)

    Other possibilities:
    Asymptote is a powerful and popular language for drawing 2D and 3D graphs. The syntax resembles Metapost. Support LaTeX typesetting of labels. Asymptote allows easy generation of animations (animated GIFs and MPEGs)! (License: GPL 3.0)
    Kiva "is a multi-platform Display PDF 2D drawing engine that supports multiple output backends, including wxWidgets on Windows, GTK, and Carbon, a variety of raster image formats, PDF, and Postscript." (Requires ReportLab for PDF output.)
    PyChart is excellent for quick, simple, flexible charts.
    DISLIN is free for non-commercial use and is available as Python extensions.
    VUESZ is a PyQT based scientific plotting package.

    Eukleides is an amazing Euclidean drawing language that can save your drawings as pstricks or in EPS format. It is designed to be very natural for geometric thinking. (License: GPL)

    Graphviz is an open source package with special abilities in drawing directed (via dot) and undirected (via neato) graphs. The focus is on Graph Visualization. pydot provides a Python interface to dot.
    yED is a "powerful graph editor that is written entirely in the Java programming language. It can be used to quickly and effectively generate drawings and to apply automatic layouts to a range of different diagrams and networks."

    Artdraw may prove promising: it is a PostScript-like syntax that adds alpha blending.

    If you are considering PostScript graphics on the way to PDF documents, is worth noting that ReportLab allows direct creation of PDF graphics and reports.

    Data and Function Driven Graphics

    Before exploring other possibilities, try gnuplot. It meets most non-interactive data driven or function driven plotting needs. It is free and open source, multiplatform, very fast, and very powerful. (There is a Mac version, and here are some instructions for installing under OSX.) You can download official release. New users may profit from Langlais's "short course" or the IBM tutorial.
    Features: In addition to plotting 2-D and 3-D graphs of arbitrary functions, gnuplot can also quickly plot large data sets. It has a powerful yet intuitive scripting language. GNUPLOT can provide output in many different formats, including the LaTeX picture environment and EPS. The LaTeX and epslatex output "terminals" allow for math symbols etc. (After LaTeXing the resulting .tex file, the .dvi file can be converted to PostScript using dvips -o. The Ghostview previewer can then be used to convert the PostScript file, fonts and all, to EPS.) All told this is quite a powerful plotting tool with extremely flexible output. New users should read Introduction to GnuPlot, but skip the institution specific material on starting gnuplot and head straight for the "Simple Example". See the gnuplot hints below.

    Other Options
    One other possibility deserves mention: Google Chart. Chart construction becomes a simple matter of URL construction. IBM offers nice examples of use. Chart construction can be simplified by scripting (e.g., in Python or in Ruby).

    Vector Graphics

    UML, Flow Charts, Etc.

    Unified Modeling Language (UML) may not be truly unified—it is clumsy at best for functional programming—but it still offers a remarkably useful collection of chart types for modeling and simulation. Here is a tiny selection of available resources, plus a few (somwhat) related flowchart resources. For more options, see Wikipedia's list.

    Papyrus seems the most complete option. However, it is a complete modeling environment, and it is accordingly heavy weight. The remaining options are primarily for drawing key UML diagrams.

    Here are a couple older flowchart tools.

    Image Manipulation

    Scripted Image Manipulation with Python


    Edward Tufte offers an online discussion of sparklines. creates sparkplots using Python and Matplotlib.

    Here is a cool sparkline application: Sparklines in data: URIs. Here is Joe Gregorio's a code sample implementing the idea in Python.
    import Image, ImageDraw
    import StringIO
    import urllib
    def plot_sparkline(results):
       """Returns a sparkline image as a data: URI.
           The source data is a list of values between
           0 and 100. Values greater than 95
           are displayed in red, otherwise they are displayed
           in green"""
       im ="RGB", (len(results)*2, 15), 'white')
       draw = ImageDraw.Draw(im)
       for (r, i) in zip(results, range(0, len(results)*2, 2)):
           color = (r > 50) and "red" or "gray"
           draw.line((i, im.size[1]-r/10-4, i, (im.size[1]-r/10)), fill=color)
       del draw
       f = StringIO.StringIO(), "PNG")
       return 'data:image/png,' + urllib.quote(f.getvalue())
    if __name__ == "__main__":
        import random
        html = """
                <p>Does my sparkline 
                    <img src="%s"> 
                fit in a nice paragraph of text?
        print html % plot_sparkline([random.randint(0, 100) for i in range(30)])

    Graphics for LaTeX

    If you produce documents in LaTeX, you would naturally like the fonts and equations in your graphics to match those in your text. There are a variety of solutions.

    The simplest answer is probably to use gnuplot. See for a good discussion.
    See the TikZ hints.
    If your figures are simple (or if you are patient), just use the very powerful TikZ drawing language directly in your LaTeX document. TikZ does not depend on PostScript, you can use it directly with pdflatex. (Conveniently, it is also compatible with the dvips/ps2pdf route.) Sean Farley offers an useful tutorial.
    There are many TikZ libraries available, including pfgplots for drawing plots and tkz-graph for drawing graphs. These can greatly simplify your drawing tasks.
    gnuplot since version 4.4 has a TikZ “terminal”. Very nice!
    TikzEdt (License: GPL3) and KtikZ can nicely help you create tikz figures. (Mykel Kochenderfer has compiled a KtikZ Windows binary.) Related functionality is provided by tikz2pdf. (License: BSD)
    If your figures are complex but drawable by hand, you can create them with the TpX drawing application, which can output tikz code. (You can set the default output format to tikz before you start your drawing, or you can change the format in the picture properties.)

    For great results even in complex drawings, try the PyX graphics package for Python. This will require an initial investment of experimentation and learning, but the results (as EPS or PDF) are fabulous.

    Some solutions require the use of PostScript.

    Often you will want to produce your graphics as EPS files, which can be included in a LaTeX document. However there are several alternatives.

    Sometimes you may want to include Encapsulated PostScript (EPS) files in LaTeX documents. This will probably take you two steps: i. tell LaTeX that you will be including a graphic, and ii. convert your .dvi file, in which you may not be able to view the graphic, into a PostScript or PDF file, in which you can view the graphic. (The second step assumes your .dvi viewer does not include a PostScript filter, but some do.) The best way to accomplish the first step is to use the \includegraphics command of the graphicx package, probably with the dvips option. So in the preamble of your LaTeX document you will have
    and then in the body of the document you will have something like

    The best way to accomplish the second step is with dvips for PostScript output or dvipdfm for PDF output.
    Comment: if you use a .dvi viewer that can display your included PostScript files (e.g., TrueTeX or yap with GhostScript), you will probably find it saves time not to view them every time you view the document. Just use the dvips option for the graphicx package so that the PostScript just gets passed through. Then remove that option when you want to see everything.
    ] Comment: According to Robin Fairbairns: The packages graphics, graphicx and epsfig were all released in their present form at the same time (summer 1994). epsfig is capable of generating some confusion because a package of the same name existed before that time (it was written as a joint replacement for the long-established, and somewhat buggy, epsf and psfig packages); the old epsfig doesn't have the characteristics we have come to rely on. The graphics package forms the basic kernel of the whole bundle: it converts graphics primitive commands into \special commands[*] for the device drivers that the bundle supports. The graphicx package adds a user interface, and makes possible the merging of commands (like scale and rotate: which in graphics need to be composed, and hence produce a longer sequence of commands both at the TeX level and potentially also at the driver level). To first order, the epsfig package simply provides a different user interface, which it translates into graphicx-like commands. There _are_ trade-offs. graphics' user interface is very "traditional tex", with little sophistication. graphicx and epsfig use a key-value syntax which is noticeably more difficult (and hence cpu-intensive) for the packages to decode. graphicx is middle of the road as far as decoding complexity goes, but has useful extra power in its graphical expressiveness by comparison with graphics. (Also, if you never learned psfig, epsfig's psfig-like syntax won't excite you much.)
    [*] or, in the case of pdftex, into pdftex commands for the job.
    If you want your EPS files to be truly portable, you need to embed the fonts. If you use a drawing application, make sure it is configured to embed the fonts. If you convert your .dvi files to PS, you can tell dvips to embed all fonts with the -Pdownload35 option
    dvips -t letter -Pdownload35 file.dvi -o
    If you are producing PDF files, you can use pdffonts to check for font embedding. If using ps2pdf you can use the option -dEmbedAllFonts=true. See this document for more discussion.
    Users of PostScript and pdflatex face an annoyance: pdflatex does not directly support the inclusion of EPS files. However, according to Kai Stirker, one can include EPS files in pdflatex files as follows. Note that you will need to have installed epstopdf, which you get when you install GhostScript. Then just \includegraphics{filename.eps} as usual to include your EPS graphic, and pdflatex will conver them to pdf format on the fly. (Be sure to return the extension! Otherwise you will get conversion the first time, and use of the old PDF after that.)

    You can use flpsed to add editable annotations to a PostScript document. (License: GPL; Platform: X11.)

    Other interesting solutions:

    Other Useful Software and Sites


    Laptop (and cell phone) theft is a constant problem. The Prey Project offers a multiplatform, FOSS theft recovery tool. It uses Google geo location and a web camera if available, and is available for the Android operating system. (A related commercial product is Undercover by Orbicule, which has additional functionality and is very user friendly.) Here are some other possible steps to take. You can run autossh as a service that creates a link between your home router and laptop. Then whenever your laptop has internet, you can ssh into it from home. You can also install a keylogger and schedule (e.g., with launchd or cron) to say sftp the log every day and then restart the log file. (Personally I would only use open source keyloggers, such as PyKeylogger, but these tend to be less focused on hiding themselves.) Probably better than any of this is vigilant use of a laptop cable lock.


    Quantum GIS is an extensible geographic information system. (License: GPL)

    Open Access

    Planning an open access journal (business models, tips, etc.). Software to create and manage online open access repositories for articles: Eprints (from Southampton University, perl); DSpace (from MIT, java); CDSWare (from CERN, python); FEDORA (from Cornell and U. of Virginia, java). Also: Copernicus is a company specializing in the publication of open access journals.

    Fun, Useful, and Open Hardware

    RaspberryPi now has an official manual, which includes introductions to programming languages (including Python and Java).


    Windows keyboards are a painful design for touch typists. Consider Happy Hacker Lite II, which unfortunately is a bit pricey. Short of that, at least remap the prominent caps-lock key to something useful: [1 | 2 ]

    Touch Typing

    Touch typing will speed your writing and programming immensely. I know of two free cross-platform solutions. Tux Typing is targeted to kids but is generally useful. Vim users can use the TTCoach script.

    Personal Information Managers (PIMs)

    Cross-platform PIMs Chandler, which is very promising. (License: Apache. Language: Python.) Also look at Aethera. In the meantime, the Palm Desktop is free, available on Windows and Mac, and quite functional (although it does not use the iCalendar standard).

    Citadel is a promising groupware application.


    There are lots of good calendars out there. But if you are "always connected" I have not been able to beat Google Calendar. (I have not tried to use it offline.) Note that you can add Google Calendar to your FireFox sidebar. Comment: if you ever see a corrupted screen when viewing Google Calendar, try clearing your browser's cache and cookies.

    Generally it seems a good idea that your calendar support the iCalendar standard. (This means you can easily move it between applications and across platforms.) It also means you can easily write scripts that interact with your calendar. For example, you can use the iCalendar. (License: LGPL; Language: Python.) An alternative is the Python-PDI library for personal data interchange, which also handles the vCard format. (License: GPL; Language: Python.) Aside: GCALDaemon offers synchronization between Google Calendar and iCalendar compatible calendar applications. (Language: Java; license: APACHE version 2.0) (Gina Trapani offers some installation hints.)

    I like Sunbird a lot (it is the Mozilla offering). Going under the name of Lightening is a calendar extension for Thunderbird that shares the Sunbird code base. For an interesting and "light" calendar, you can try Remind. (Linux Journal has a review.)

    Address Books

    The address book situation is a scandal. Applications have not settled on a standard for data sharing and do not even adequately document their own formats. The best documented formats, VCARD and LDIF, are not great. For such a widespread need, this is a rather astonishing situation.

    Thunderbird comes with a pretty good address book tool. Unfortunately, it saves address books in Mork (.mab) format, and if you back these up it cannot import them! Fortunately, there is a Mork parser library (Language: Python) Thunderbird 3 was supposed to replace the horrible Mork format with MozStorage, thereby adopting the storage format now used in FireFox. But the current situation is still bad. However, there is an address book extension that improves things.

    Most PIMs include address book functionality. Most email clients do as well. You can also get such (database) functionality in stand-alone applications. On Linux you can use Gaby or GTK-Agenda. There is also the Contacts application.

    Project Management

    See Wikipedia for a comparison of project management software. Here is another comparison.

    Here is one that is not yet listed there: ProjeLead (License: GPL) Probably Trac is best known in the open source community.

    Personal Financial Managers (PFMs)

    Open Financial Exchange

    Open Financial Exchange (OFX) "is a unified specification for the electronic exchange of financial data". The format is an open standard. It is possible to convert CSV to OFX.

    Office Suites and Related Software

    Advice on office suites: do not use them for research. Word processors may slow down your writing; use a good text editor. Spreadsheets are a risky way to handle data [ 1 | 2 | 3 | 4 | 5 | 6 ]: use a simple matrix programming language instead. The virtues of existing presentation packages do not outweigh their weaknesses, especially if you need to present mathematical formulae.

    Nevertheless, office suites can be convenient. Here are some basic considerations. FOSS office suites are currently a high-quality multi-platform solution. I recommend either LibreOffice (License: LGPL) or OpenOffice (License: Apache). Of course most of the people you know will be using a commercial office suite, but you will be able to save your files in formats they can read, and as long as they are not using advanced features (e.g., templates, complex macros, or complex spreadsheet formulae) you will be able to open their files as well. Most users do not use such features. FOSS suites meet most personal and business needs.

    Since office suites are often expensive, here are some cheap or free commercial alternatives.
    Google Drive (good intermediate functionality; easy collaboration; limited advanced functionality)
    Word Perfect Office (home of the famous Word Perfect word processor)
    Soft.Maker is a cross-platform commercial suite.
    Hancom Office

    Document Exchange

    We exchange documents for three primary reasons: for viewing, for interaction, and for modification. Word processor formats are a popular choice, and the versioning capabilities of modern word processors can be very useful in a collaborative setting. Plain text can often satisfy viewing and modification needs, and there are great version control tools for text formats. When more structure is needed, HTML will usually suffice. When formatting must be tightly controlled or access needs to be controlled, PDF is often a good choice.

    Word Processors

    AbiWord is a cross-platform, full-featured word processor with a Word®-like interface. If you want a quick viewer for MS .doc files (e.g., for your email), this loads much faster than most alternatives. It also has a wonderful equation editor, which accepts LaTeX input in a dialogue box.

    If you want even lighter weight than the very nice AbiWord, on Linux you can try Ted. Some Linux users like frames-oriented KWord.

    Nowadays it is perfectly feasible to use a decent HTML editor as your word processor. The Amaya browser/authoring tool may be worth considering, although I have found it a bit slow and clunky. (Somewhat offsetting this is its native MathML support.) Another possibility is the web development suite offered free by Evrsoft.

    There have been some interesting efforts to provide fairly direct use of PostScript for document markup. A popular one was Graham Freeman's QuickScript, but I can no longer find it on the web. Another alternative, Tinydict, is still around.


    It is now perfectly possible to rely on internet applications for word processing. I have had good experiences with Google Docs for basic word processing tasks and online collaboration. Zoho Writer looks even more capable, and has good math support as well. (Including the ability to produce math by inputting LaTeX.)

    Bibliography Management

    You will need to decide what format to keep your bibliography in, and what software to use to manage it.

    Citation Management:
    If you use a commercial word processor you will probably use the commercial EndNote to manage your reference and format your citations. However there are some interesting alternatives. In particular, LaTeX users have the traditional (and very powerful) BibTeX to manage the citations in their documents. Even users of commercial word processors can take advantage of BibTeX by using Bibtex4Word. New users should see the BibTeX user hints. Now for the design of citation reference and citation style, they have the even more powerful biblatex (which greatly simplifies writing new styles). (Users switching from natbib will want to know about the natbib-compatibility-mode, which allows use of natbib's commands (e.g., \citep, \citet) instead of the commands usually provided by biblatex (e.g., \parencite, \textcite).) Python users can consider the bib4txt component of bibstuff, which should particularly interest reStructuredText users who need bibliographic support. (I recommend you get the most recent BibStuff.) (BibStuff License: MIT; Language: Python) Chris Putnam's Bibutils "interconverts between various bibliography formats using a common XML intermediate." (Bibutils License: GPL; Language: C; binaries are available for many platforms.)

    Users of Open Office or MS Word can insert references directly in their documents with Bibus.

    CiteULike is an interesting "in the cloud" solution for managing references.

    Citation Management

    The current choice for citation management appears to be Biber+Biblatex. (Other possibilities are discussed below.) The traditional choice for citation management is BibTeX. (See the BibTeX user hints.)

    Here is how to transition from the traditional Bibtex with natbib to the more modern Biber with biblatex. (If you have not been using natbib, you can leave out that option.) After changing your LaTeX document as indicated below and compiling it with pdflatex as usual, run biber on the document’s basename (instead of using Bibtex).

    %\usepackage{natbib} %REMOVE natbib package
    %ADD biblatex package (with Biber backend)
    %ADD .bib files (one by one, in preamble)
    Lorem ipsum dolor \dots
    \printbibliography %ADD printbibliography command
    %REMOVE old bibliography commands

    In the background, the choice of the biber back end for biblatex will lead to to the creation of a .bcf file, which biber operates on. But you should not need to know about this unless choose to pass to biber the full filename (basename + .bcf) instead of just the basename. In your document, you will need to keep the natbib option to biblatex as long as you use natbib style citation commands. Or you can replace all your citet and citep commands with textcite and parencite commands.


    Traditional spreadsheets:
    It has become very popular to use spreadsheets in education. This has some downsides. Traditional preadsheets are OK for very simple, one-off applications using small amounts of data. For serious work, they are too opaque, because they are too hard to document. (If you have ever tried to figure out another person's complex spreadsheet, you know what I mean.) Particulary well publicized was the Excel error whose correction overturned a much cited Reinhart and Rogoff study. But studies indicate that almost all large spreadsheet have serious errors.1,2 Basic data handling glitches have been documented even in the most popular spreadsheet applications. The lack of transparent scripting facilities means that it is very difficult to understand someone else's spreadsheet or even one’s own spreadsheet developed months or years ago. If all you want to educate someone to do is one-time, never-to-be-revisited, quick-and-dirty computations, then teach them a traditional spreadsheet. Otherwise teach them a format that allows serious documentation and transparency of manipulations. At the very least, eschew traditional spreadsheets in favor of newer alternatives, such as scientific spreadsheets and engineering spreadsheets.

    Comment: some problems with spreadsheets can in principle be overcome by using them directly just for data storage and then preforming manipulations via a scripting language that exploits the spreadsheet's computational and graphics facilities---on a Windows platform, perhaps VBA or better yet Python with win32com; on a Unix platform, perhaps ABV or Python or a plug-in that supports scripting. Of course this is not how most people use spreadsheets. There are also engineering spreadsheets that allow explicit scripting and treat data in a more object oriented fashion. These are tools that can be properly used for serious work.

    Comment: Python can read and write tranditional spreadsheets. For simple spreadsheets see the CSV module. For more advanced needs, consider xlrd and xlwt.

    Scientific spreadsheets, often called engineering spreadsheets, are much better working environments than traditional spreadsheets. A scientific spreadsheet treats data as objects and allows programmatic manipulation in a spreadsheet-like environment. If you use this properly, it can overcome many of the problems with spreadsheets while still allowing easy graphics creation. Serious spreadsheet use will involve a lot of scripting. You should therefore choose a scientific spreadsheet that supports an easy to use yet powerful scripting language. (I recommend Python.) Here are a few possibilities.

    If you are interested in spreadsheet structures, you may be interested in Fenfire generally and ZigZag more specifically.


    Perhaps the simple yet powerful NoSQL will meet your needs. More traditional approaches follow.

    SQLite "is a small C library that implements a self-contained, embeddable, zero-configuration SQL database engine". (License: Public Domain.)

    FireBird "is a relational database offering many ANSI SQL standard features that runs on Linux, Windows, and a variety of Unix platforms."

    OpenOffice includes a database. GNOME Office includes GNOME-DB, which is reputed to be very good. For enterprise needs, PostgreSQL has a good reputation: PostgreSQL is a free-and-open-source, SQL-compliant, object-relational database management system, reputed to be highly-scalable. Using Ruby with Rails and say MariaDB or PostgreSQL is getting a lot of buzz for developing data-base backed web applications.

    Simple but structured needs may even be met by Open-RJ.

    Middleware, including Object Request Brokers (ORBs)

    Project Managers

    faces is a user friendly project management tool. (License: GPL; Language: Python)

    Task Coach is a lightweight task manager. (License: GPL; Language: Python)

    Task Juggler is very popular. (License: GPL)

    OpenOffice includes a database. GNOME Office includes MrProject, "a project planning, scheduling and tracking tool.


    CYCAS is the only cross-platform CAD package I am aware of. The most popular CAD program remains the Windows® based AutoCAD.

    Data Formats

    Consider HDF5 for substantial databases. (Overview.) Binaries. Windows installation.

    Desktop Publishing

    Scribus is a featureful, cross-platform desktop publishing application. Supports PDF 1.4 and Python scripting. "The Scribus file format is XML based; open and completely documented." (License: GPL) Used by professionals to send multipage full color layouts sent directly to commercial print houses. Supports CMYK separation, PDF generation, and much of the toolsets you get from commercial packages like Quark or InDesign.


    HylaFAX offers an open source version for a variety of *nix platforms. (This OpenOffice setup discussion may be helpful.) I do not know of any FOSS fax software for Windows or Mac: let me know if you do!

    Instant Messaging

    Pidgin IM (formerly Gaim) is a free and open source instant messaging client for a variety of platforms. ScatterChat is a secure enhancement of gaim.

    Mac users can consider Fire, which supports most popular instant messaging services. (GPL)


    Character Composers

    A compose key allows inputting special characters as key combinations. Some text editors provide support for this (e.g., digraphs in Vim). Linux includes compose-key activation as a native ability. For Windows users, a few are provided via the Alt key and the numeric keypad. However, WinCompose provides better functionality. (License: WTFPL (essentially public domain).) (There is a Youtube tutorial for WinCompose.)


    Calculators are still useful. Here are a few of many.

    Reverse Polish Notation:
    The dc desktop calculator, for those comfortable in a shell (including cmd.exe), is cross-platform, wonderfully fast, and ever handy---highly recommended! Another cross platform possibility is kalc. For Linux you can try rpncalc. For Windows I like the freeware RPNCalc, which is also ported to WinCE. RPN is an interesting calculator for the PalmOS. And finally for PostScript enthusiasts, I cannot refrain from mentioning PSCalc.

    For a console calculator, try the cross-platform bc. (The bc manual is very good.) For a graphics calculator, try the cross-platform GraphCalc. Or just use gnuplot as a graphics calculator. GNU Emacs Calc has many features, including graphics using GNUPLOT. CoCalc is a freeware JAVA calculator.

    Rascal is enough more than a calculator that it may be considered a simple matrix programming language. This is also true of cxc, and Iguana seems to be evolving in the same direction.

    File Compression

    Remote Access

    Terminal Emulation
    File Transfer (FTP)

    Check TheFreeCountry for a longer list.

    Python users get a basic cross-platform solution automatically via the ftplib module. (For an even lighter approach, try this TFTP module.) The Python Cookbook uses this module to provide a threaded ftp client. As of August 2005, ftplib does not provide any secure services, however. For SSH2 services you can use paramiko, which relies on pycrypto. These are very easy installs! (Get paramiko from the releases tab on GitHub, and see the README file for instructions. Windows users can also look at Voidspace for prebuilt PyCrypto binaries.) I provide an example (usable) SFTP client using these modules, but you can even have a graphical interface with SFTP support via wxSFTP. (If your needs are more substantial, you can use pycurl as a Python interface to libcurl, which should be able to meet all your ftp needs.)

    If you need FTP/S, you can try TSLite, which offers a pure Python solution for FTP over TSL.

    For Windows: I find the free and open source FileZilla to be excellent for FTP and SFTP. (As of August 2005, however, I have had some problems with FTP/S in passive mode.) Other possibilities: WinSCP. OpenSSH for Windows. WS FTP had a free (for academics and students), easy to use, ftp program for Windows platforms, but I think the free version went away.
    For Mac: CyberDuck.
    For Un*x: NcFTP. (Also runs on Windows and Mac, for those willing to learn ftp commands.) OpenSSH.

    Optical Character Recognition

    GOCR, Kooka, OCRShop, ClaraOCR

    Report Generation

    Generating reports is a common need. These can be informal or presentational, and may or may not need to be printed.

    gawk, a powerful report generation language which takes only a moment to learn well enough to produce simple reports. (E.g., you can manage student grades with Gawk much more easily than with a spreadsheet.)

    ReportLab allows direct creation of PDF graphics and reports.

    iText can be a very powerful report generator if you know a bit of Java.

    Apache Formatting Objects Process (FOP) offers amazing flexibility. Java based. (License: Apache 2.0)

    Printing Reports

    Program controlled printing can be a challenge, especially under Windows. If you have the option of producing your reports as PDF or as appropriately formatted text, then you can usually just ask your operating system to use a default application to print them. Otherwise, you will have to work harder.

    Tim Golden offers a very useful discussion of Windows printing under Python.

    GNU a2ps allows conversion of text (and other) files to PostScript, as does enscript and ASCIIprint.

    wordaxe can be used for hyphenation. For example, you may want to generate your own PostScript or PDF. As a more specific example, the wordaxe hyphenation library offers ReportLab support. For an easy interface, try rst2pdf, which offers cascading stylesheets and supports TTF and Type1 font embedding.

    Version Control

    There are many version control systems (VCSs): see the Wikipedia list. If you just need revision control for a single-user, single-machine project, you can use RCS. For simple version control on multi-user or multi-machine projects, consider Subversion (SVN). Currently distributed VCSs are the rage, with Bazaar (License: GPL), Mercurial (License: GPL), and especially Git (License: GPL) having particularly high profiles. (Windows users of Git should consider GitHub for Windows.) Having a distributed VCS is not important for most small projects.


    See my Subversion hints.

    Despite the recent popularity of Git, Subversion is probably the most widely used version control system. There is an extensive user manual. (Be sure to read about the basic work cycle. Simplest bottom line: change to your working directory and enter svn update, then make changes to your files. Before you commit do another update, and finally commit the changes with svn commit -m "My message about the changes.".)

    The pysvn project provides a Python based graphical interface to SVN.

    SVN on windows: download and execute the Windows installer. Users who dislike the command line should consider TortoiseSVN as a user-friendly Subversion client.

    SVN with Vim: vcscommand provide keybindings for SVN control from the Vim editor.

    Free and low cost SVN repositories:
    assembla (includes some free options)

    Subversion users may find it helpful to generate graphical reports of the repository changes. A Python-based solution is svnplot. A Java-based solution is StatsSVN.

    Miscellaneous Office Needs


    Track stolen laptops and cell phones with Prey.

    E-Book Readers and Converters

    Calibre has quite amazing conversion capabilities and supports many e-book formats.


    Outlines can help writers and programmers.[ 1, 2 ] Many good editors offer substantial outlining capability. Examples include the VimOutliner and the emacs outline mode.

    There are many commercial and free outliners. The Java Outline Editor offers serious outline functionality.

    Mind Mappers

    FreePlane is an active fork of FreeMind, which is a highly capable cross-platform mind mapper. (Language: Jave; License: GPL) For a more document oriented approach, try Kdissert. (License: GPL) MindMup provides free online mind mapping, including online collaboration. (License: MIT. Supported by Google Drive.)

    Tree Sheets is an odd cross between a spreadsheet and a mind mapper.

    Making Posters

    The basic situation. The graphics problem.

    Text Preprocessors

    The hoary standard is m4. An interesting alternative is empy. Good luck.

    Operating Systems

    After the Windows and Mac operating systems, most people will think first of GNU Linux, an excellent, free, Open Source version of the Unix operating system. Another popular free and open source Unix relative is FreeBSD. Another popular version of the Unix operating system is Solaris, a FOSS version of which is available as illumos (formerly OpenSolaris).

    Plan 9 is very roughly a faster, more secure multi-processor Unix. There is an open source implementation of BeOS called Haiku, but in 2015 it seems that organization issues may lead to a forked project calles Senryu. (Haiku can install and run within Windows.) QNX's Neutrino Real Time Operating System is also a compact and stable operating systems. (Can install and run within Windows, and thus does not require a separate partition or boot loader!) An interesting effort to produce an open source (and binary compatible) substitute for Windows NT is ReactOS, which already has basic console functionality. A few people may also be interested in FreeDOS.

    If you pick the GNU/Linux Operating System you may want to visit Furthermore, if your interests are computational, you may want to consider the Quantian distribution.
    A recent review of Web OSs highlights another way to get basic OS functionality.


    Things have gotten better with the advent of PowerShell, but Windows users still suffer from a dearth of good shells. Here are a few options.

    Unix Tools

    Unix Tools for Windows and System Tools
    Unix Tools for MacIntosh

    Fink has an extensive collection of Unix tools for Mac users.

    View and Manipulate PDF

    PDF Maniuplation
    PDF Viewing



    Note that password protecting your Word files does not encrypt them. Password protection of PDF files is easily cracked, although Adobe requires that readers respect the protection. Some editors (e.g., Vim) allow easy, strong encryption of your text files.

    Stuff for Kids (and Educators)

    Banshee music player. (License: MIT)

    CDex CD ripper. (License: GPL) On Windows, CDex may provide an alternative to the poor iTunes error correction (i.e., the presence of pops and clicks).

    Tomboy note-taking application.

    Open Standards

    Data Exchange

    Please let us stop playing nationalist games with number formats before someone gets badly hurt as a result. For scientific data exchange, the decimal separator is a point (ASCII 46). It is not a comma (ASCII 44). Neither is it any other UTF code point. It is just that simple. Pretending otherwise is like pretending that science does not use the metric system.

    Possibly the simplest open standard for fixed periodicity time-series data is the open-databank standard. Most time series econometrics packages support some version of this.

    The most widely used data exchange format is comma-separated values (CSV). There actually is a standard, in the form of RFC 4180. Sadly, many files with the .csv extension do not conform to this standard, so caveat emptor.

    AstroAsciiData provides simple database functionality for ASCII tables.

    HDF5 "is a general purpose library and file format for storing scientific data". Some support can be found in the econometrics community, both in commercial and free packages. Python users have excellent support via PyTables. (To use PyTables with Python 2.4 in Windows platforms, you will need to get the Windows binaries.) Octave and R provide HDF5 support. Commercial applications supporting HDF5 include Mathematica and MATLAB version 6.5+, among many others.

    The netCDF libraries "support a machine-independent format for representing scientific data". The netCDF distribution includes interfaces for C, Fortran77, Fortran90, and C++, and interfaces are reportedly available for Ada, IDL, Java, MATLAB, perl, Python, R, Ruby, Tcl/Tk. The interface provided by ScientificPython is supposed to be excellent. PyNIO adds to this support of additional formats. Another possibly faster way to read the netCDF format is pupynere, a pure Python solution. (License: MIT)

    User Hints

    CSV Hints

    CSV stands for comma-separated values. Distinguish CSV and DSV. CSV is described by RFC 4180. Do not give a .csv extension to files that fail to match this description.

    Email Client Hints

    Be sure to set your client to produce plain text email. I have written up a few reason and clues.

    Mahogany Hints

    Make Vim your editor and call it as
    c:\programs\vim\vim73\gvim.exe -c "so $VIM\vimfiles\mail.vim" %s
    where mail.vim is where you put any mail processing commands (e.g., set ft=mail tw=75). You are welcome to use my mail commands.

    Under OS X Mahogany uses the Unix OpenSSL library, so you need to: 1. install openssl libraries (you can get them from fink) 2. point M to then: in the "helpers" page of the options dialog, add the correct paths to them, then it should work

    Thunderbird Hints

    External Editor
    Using your own editor when writing email is strongly recommended. Get the "External Editor" extension. For example, use Vim with Thunderbird as follows. Define the External Editor command to load Vim as: \pathto\gvim --nofork -c "setf mail", so you will get 'mail' syntax highlighting and ftplugins loaded. Or if you have defined say mail.vim, a file of useful macro, you can instead use e.g., \pathto\gvim.exe --nofork -c "so $VIM\vimfiles\mail.vim". After installing the extension, edit the composition toolbar (the one for replying to email, not the main application toolbar) and add a button for the external editor. You can then click that button and your current email reply will be opened in Vim.
    By the way, you may find my ai_mail.vim to contain some useful abbreviations and commands for email. Just drop it in your %VIM%\vimfiles folder, and change your External Editor command to, e.g., "C:\Program Files (x86)\Vim\vim82\gvim.exe" --nofork -c "so $VIM\vimfiles\ai_mail.vim".
    Thunderbird and Acrobat: Acrobat will not attach PDFs to Thunderbird emails, unless you you set Thunderbird as your default client in a very specific way. Open the Control Panel and double click Default Programs. Pick "Set program access and computer defaults/Custom (expand this)/Choose a default e-mail program, and then, click on Thunderbird. (As of Windows 7 and Acrobat 9.)
    Make sure your 'textwidth' when editing the email is no larger than Thunderbird's "Wrap plain text messages at" option (in your Composition options); otherwise when Thunderbird receives your text it may wrap it to a smaller size.
    Gmail allows IMAP access, Google provides detailed instructions for configuring Thunderbird, and Thunderbird automates the process of adding your Gmail account.
    Key Configurations:
    The keyconfig.xpi extension allows you to easily change your key bindings. Carefully follow the instructions. I like to associate the d key with goDoCommand('cmd_delete'); and the x key with goDoCommand('cmd_MsgCompactFolder');.

    Pine Hint

    If you prefer to POP your mail, you can get more control over this than Pine offers with radiks, a tiny mail transfer program (MTA) for win32 (Windows 9x/NT/2000) which reads mail from a pop3 server and transfers it (filtered, if you want) to your local hard drive.

    SciPy user hints

    Use the Lapack routines if you need speed. E.g., for a QR decomposition use scipy.linalg.flapack.dgeqrf instead of scipy.linalg.qr.

    For numarray <= version 1, there is a matrixmultiply (dot) bug: The multiplication is done correctly, but the post-multiplying matrix may be changed after the multiplication is performed. David Cooke suggests that quick fix is to define your own matrixmultiply (or dot). E.g.,
    def dot(a, b):
        return, b.view())
    The original b is then not touched, and the screwed-up matrix is discarded.
    Suppose you find you need to import a module with the same name as your script. First best is to change the name of your script, but you can also
    import sys
    import sets
    to remove the current directory from Python's search path. If you then need access to any modules in the current directory, use sys.path.insert(0,'')" first.

    As late as 2004, the Numeric and numarray documentation incorrectly stated that and, or, and not would operate element-by-element on arrays.

    Lists are sorted in place, so mylist.sort() return None, but mylist is now sorted.

    R Hints

    Rossiter wrote an accesible Introduction to the R Project.

    Windows users may want to read R Guide for Windows Users by Kevin Wang. Lund University offers some supplementary documentation; also Harrell's R utilities along with his guide to LaTeX tables and S may be useful. Ikaha offers some introductory discussion of generating graphics in R.

    Be careful with the definition of R’s PRNG for the geometric distribution: rgeom has support on nonnegative integers, since R has p(x) = p (1-p)^x. A perhaps more common parameterization is p(x) = p (1-p)^(x-1): which you can match by adding 1 to the values returned by rgeom.

    You can write LaTeX documents containing R code whose ouput is to be included in the document with Sweave.

    Windows users can access R from Python with the RDcom server/client. This allows you to call R from Python as follows:
    >>> from win32com.client import Dispatch
    >>> sc=Dispatch("StatConnectorSrv.StatConnector")
    >>> sc.Init("R")
    >>> print(sc.Evaluate("2+3"))
    Other approaches to R/Python interaction include the following. Walter Moreira's RPy module (requires pyWin32 under Windows) allows you to "embed" R in Python and transparently send data to it. (You can easily pass objects in both directions using OmegaHat's R/S Plus Python Interface.) pyRserve is a simpler solution, which nevertheless allows running R on a remote server.

    Bounding Box Hints

    Occasionally you will need to provide or adjust the bounding box information for a graphic. Here are some resources.

    AbiWord user hints

    AbiWord supports easy conversion to LaTeX, and thus easy generation of PDF versions of your documents. Here is a simple command line approach, given a document named test.abw to produce a document test.pdf.
    abiword --to=latex test.abw
    pdflatex test.latex

    AbiWord has a special feature: it supports some vi keybindings! To implement the vi keybindings, you need to make two changes your AbiWord.Profile file. (Windows users should search for this in the Documents and Settings folder.) First add another Scheme to your AbiWord.Profile file: Just add the following lines to your AbiWord.Profile file right below the already existing "_custom_" Scheme.
    By setting KeyBindingsCycle="1" (1 is on, 0 the default is off) you will be able to press F12 to switch keybinding mode: the mode will appear in the Status Bar just to the right of INS. The second step is to go to the beginning of AbiWord.Profile and find the Select tag; modify its scheme attribute to refer to the new Scheme you just typed. It should then look like this:
    This is very cool. You get into insert mode by pressing i and out by pressing <esc>. A number of basic cursor motion keys (including h j k l w b a A ^ $) and core commands (including O o D x) are implemented, along with / for a search dialog. See for a list of implemented commands, but a few do not currently work.

    Acrobat hints

    When you add comments to a document, the comments are stamped with a name, a date, and a time. This may well be more information than you want to provide. You can remove the name stamp from all your comments as follows. Show the comment list: pick Comments » Show Comment List from the menus. Select all comments: click on one, and then press ``ctrl-a``. Change the comment properties: right click on your selection, click ``Properties``, pick the ``General`` tab, and change the Author field (e.g., nothing). Unfortunately, you cannot remove the date and time stamps this way. However, only the "modified" stamp shows to the reader, when you change the author this way, it should update the modified time on all the comments.

    GAUSS user hints

    1. Output to the output window is stored in a buffer which has to reach a certain capacity before anything shows up. You can use the sentence print /flush;; to force GAUSS to flush the buffer. You can also use the "configure/preferences/cmd_window" menu and uncheck the box for "Buffer output".
    2. Aptech will supply students with GAUSS Light free of charge! (I've been using it extensively for teaching, as a result.) Compared with the full version, there are matrix limitations (ie.., 10,000 elements and/or 100 X 100 matrix limit, 1 Mb global symbol memory, 1 1/2 Mb total workspace) and no source level debugger. Here are the download instructions I received from Aptech. It seems the universal binary for the Mac is for OS X v.10.4 or later and only target the Intel platform!? (But I'm not certain...)
      Please note that files can only be downloaded via COMMAND LINE FTP, not
      a browser (such as Internet Explorer, Netscape, Firefox, etc.).  You
      must use a command window to download the files.  To open a command
      prompt window, click on Start, then Run, and type cmd.
      GAUSS Light 8.0 for Windows
      Below is a list of files that you may wish to download and
      what each is:
     [GAUSS Light 8.0 software]     [all documentation zipped together]
      Save the files that you download to either your root directory 
      (usually C:) or a directory that you set up that has no spaces
      in the title; there can be problems saving the files to a 
      directory such as C:\Documents and Settings\<username>.
      You *must* know the name of the files you wish to download.
      the necessary files via command line ftp, do the following
      at a terminal prompt:
      1. Type the following at the ftp prompt:
      2. Once you are in the site, type: 
         then when prompted enter your email address as the password.
      3. Type the following at the ftp prompt to get the files 
         (ftp> represents the ftp prompt you should see, do not type that): 
        ftp> get 
       (remote-file) /outgoing/ 
        ftp> get 
       (remote-file) /outgoing/ 
        ftp> bye
      Approximate file sizes
      8.8 MB
      8.5 MB
      Please note that files can only be downloaded via COMMAND LINE FTP, not
      a browser (such as Internet Explorer, Netscape, Firefox, etc.).  You
      must use a command window to download the files.  To open a command
      prompt window, click on Start, then Run, and type cmd.
      GAUSS Light 8.0 for Mac Universal
      Below is a list of files that you may wish to download and
      what each is:
      GAUSS_Light_8.0_Mac_OS_Xu.dmg  [GAUSS Light 8.0 software]           [all documentation zipped together]
      Save the files that you download to either your root directory 
      (usually C:) or a directory that you set up that has no spaces
      in the title; there can be problems saving the files to a 
      directory such as C:\Documents and Settings\<username>.
      You *must* know the name of the files you wish to download.
      the necessary files via command line ftp, do the following
      at a terminal prompt:
      1. Type the following at the ftp prompt:
      2. Once you are in the site, type: 
         then when prompted enter your email address as the password.
      3. Type the following at the ftp prompt to get the files 
         (ftp> represents the ftp prompt you should see, do not type that): 
        ftp> get 
       (remote-file) /outgoing/GAUSS_Light_8.0_Mac_OS_Xu.dmg 
       (local-file) GAUSS_Light_8.0_Mac_OS_Xu.dmg
        ftp> get 
       (remote-file) /outgoing/ 
        ftp> bye
      Approximate file sizes
      13.9 MB GAUSS_Light_8.0_Mac_OS_Xu.dmg
       8.5 MB
    3. Syntax highlighting is available for TextPad and (my crude effort) for Vim.
    4. As late as version 6, there is an error in quantile() that affects computation of the median. Fix this yourself, or use median() instead. To fix it, you can use the following code, suggested on the 'gaussians' list by Ron Schoenberg:
      proc quantilenew(x,e,a,b,c,d);
          local w, wt, f, r, i, z;
          if minc(e) < 0 or maxc(e) > 1;
              errorlog "Inadmissable quantile levels";
              if not trapchk(1);
          w = a + (rows(x) + b) * e;
          wt = floor(w);
          f = c + d*(w - wt);
          if wt == 0 or wt == rows(x);
               errorlog ftos((maxc(1/minc(e)|1/(1-maxc(e)))),"*.*lf",1,0) 
                         $+ " rows required for this level";
               if not trapchk(1);
          r = zeros(rows(e),cols(x));
          i = 1;
          do until i > cols(x);
              z = sortc(x[.,i],1);
              r[.,i] = z[wt] + f .* (z[wt+1] - z[wt]);
              i = i + 1;
    5. GAUSS now has procedures for integrating over an infinite support. Documentation can be found in the file, inthp.src.
      ** ===========================================================================
      **   Procedure     Format                       Purpose                 Line
      ** ===========================================================================
      **   inthp1   y = inthp1(&f,pds,ctl);   integration over -inf,+inf
      **   inthp2   y = inthp2(&f,pds,ctl,a); integration over a, +inf non-oscil
      **   inthp3   y = inthp3(&f,pds,ctl,a); integration over a, +inf oscill
      **   inthp4   y = inthp3(&f,pds,ctl,c); integration over a, b
      ** ===========================================================================
    6. To transform part of a multi-dimensional array to a matrix: first transpose the array with atranspose, so that the two dimensions that you want in your matrix are in the two last dimensions of the array. Then extract the matrix with getmatrix using the appropriate locator.
    7. Representing infinities: __INFn < x < __INFp for all real numbers x, and the usual "extended" real number properties apply. (Note the double underscore for these special constants.)
    8. If you need to trap an error so your program can continue after an error is detected, use the trap and scalerr commands. (See the example for 'scalerr' in the GAUSS Command Reference.)
    9. GAUSS provides more than one command for singular value decomposition. Ron Schoenberg comments: "If you are having problems with svd1, switch to lapsvdusv. Svd1 is an older but faster version based on Eispack source code. Lapsvdusv (as well as other lap* functions) are built on the more recently developed, more advanced Lapack source code. They are somewhat slower than the old functions however, and that is why the older functions have been kept in GAUSS."
    10. To raise a negative number to a non-integral power, use complex(b,0)^x, where b<0 is the base and x is the exponent.
    11. You can create a 4×4 matrix of null strings as let string a[4,4]="";. But what if the dimensions r, c are not known ahead of time? Then use a=""$+zeros(r,c); instead.
    12. Version 6 introduced a variety of new functions for working with Excel files, including a function for computing the terminating column: xlsGetSheetCount, xlsGetSheetSize, xlsGetSheetTypes, xlsMakeRange, xlsReadm, xlsReadsa, xlsWrite, xlsWritem, xlsWritesa. Note that the function xlsMakeRange computes the terminating column.
    13. Aptech provide an ODBC database connectivity kit free to registered users. It works with SQL statements. Contact for the current URL.
    14. Graphs: GAUSS allows the user to set the viewer application via the _pvwr global. The default on Unix/Linux is ghostview; on Windows it's vwr.exe. When you use tgauss to batch process your GAUSS program, you may want to to suppress the auto-exec viewer functionality: just set _pscreen=0. If your program draws several graphs, the default for GAUSS 3.6+ is that only the last one is accessible in the TKF file viewer. GAUSS draws each over the previous one. To get GAUSS to tile the graph windows (as in previous versions)
      call setvwrmode("many");
      at the top of your command file. To return to putting all graphs in a single window (for example, to produce an "animated" effect),
      call setvwrmode("one");
      Note that if you do not like the default colors for the lines you can change them: see the help for _pcolor. If you do not like the default colors for the graph characteristics (ticks, frame, labels, background, etc) you can change them: see the help for _pmcolor.
    15. You can save GAUSS plots in Encapsulated PostScript (EPS) format. Graph conversion is determined the the global variable _pcmdlin. To convert a graph into a ps file (flag -c ) stored in "c:\filename.eps" (flag -cf ), include the line
      _pcmdlin = "-c=1 -cf=c:\\filename.eps";
      in your program right before creating the graph. (Note the double backslash.) Alternatively, in the window displaying a GAUSS graph, pick File, Convert, PostScript. Then search your disk for cvt*.eps to find your new .eps file. (The filename and other parameters are configurable in pqgrun.cfg, which you should definitely take a look at. For example, you need to set cvt_color = RGB if you want color graphs.)
      As of GAUSS 7, I noticed that _pcmdlin is not documented. This suggests that you need to set the .tkf file names by setting the _ptek global, and then use the tkf2eps function to convert the .tkf files to Encapsulated Postscript. The syntax is: ret = tkf2ps("mytekfile.tkf", "myepsfile.eps"); (see the GAUSS Command Reference).
    16. GAUSS handles complex numbers. For example, you can produce the complex number 2-3i with the code i = sqrt(-1);x=2-3*i;. Note that GAUSS does not set the value of i by default; and you can pick any variable you want for this purpose.
    17. In the for command in GAUSS the loop indicator is not a floating point. Up to v.3.2.44, you need a hack if you want to pass the loop indicator to a procedure (e.g., to convert it to a character value). For example, the following code works.
        for i (1,10,1);
          j= i;
          print $ftocv(i+0,8,0) $ftocv(j,8,0);
      but it won't if you just pass i (or, oddly, 0+i). (In order to get a fast loop, the loop indicator is given special status in the compiled code and is not simply a float. The work-around generates some extra machine code that solves the problem. See Aptech's bug page for details).
    18. As of version 3.6, the use of for loops is not safe in recursive procedures. This is supposed to be fixed in version 3.7 or later.
    19. If your computations should not produce complex numbers, turn complex numbers off by calling sysstate(8,0). This will cause a fatal error if log, ln, or sqrt, are passed negative arguments. If you expect real results from a procedure that can produce complex values (e.g., fft), force the matrix to be real like this: if abs(imag(x)) ≤ 1e-12 ; x = real(x); endif; For real symmetric matrices, be sure to compute eigenvalues with eigh instead of eig: this guarantees real eigenvalues and is much faster than eig. Furthermore, if you use eig() for a real symmetric matrix, rounding error in the calculations will nearly always produce complex eigenvalues with imaginary parts that are very small.
    20. There is something of a problem with the inverse ChiSquare function cdfchii through GAUSS 4.0 and possibly later. Test with the following code snippet: cdfchii(0.9998,13); As one might expect, the same problem arises for the inverse gamma function. Test with the code snippet 2*gammaii(13/2,0.9998);
    21. GAUSS can generate .ps or .eps files. If you want to add these to a Web document, use ghostscript to convert to a portable anymap .ppm file, then use ppmtogif to convert to gif.
    22. If you are using the Windows version of GAUSS and you are using OPTMUM, set __optmum = 1. To keep the iterative information from going to an output file, set output off before calling Optmum and then turn it back on after the Optmum returns. For better screen output when __optmum = 2, set the font to "terminal" (there's a font menu in the Gauss window).
    23. To determine if a variable holds numeric or character value is to do a string comparison (it works with character but not with numeric). For example:
      let x1 = 4; let x2 = a;
      s1 = "" $+ x1; s2 = "" $+ x2;
      if (s1 $== x1); "ok"; endif; if (s2 $== x2); "ok"; endif;
      Only the second equivalence (s2 $== x2) is true.
    24. GAUSS 3.5 installation: Aptech advises you to put the license file into the \Gauss35\Flexlm. However, as soon as Gauss35 starts up, it begins looking for the license file in the \Gauss35 directory. Make sure you do not have an old license file there; it can cause a lot of trouble.
    25. To get nice looking output from, e.g., Maxlik, open a DOS window before calling the program. E.g.,
      call DOSWinOpen("Your Title Here", error(0));
      {x,f,g,cov,ret} = maxlik("tobit",0,&lpr,x0)); @or whatever@
      (Unfortunately the DOSWinOpen will cause GAUSS for Windows 3.5 (up to Kernel rev. 3.5.17, GUI rev. 3.5.14 to shut down.)
    26. Up through GAUSS 3.5, there is a bug in the polyeval() procedure that prevents it from calculating matrix polynomials. The bug is easily fixed. Just open the file poly.src and change the line
      y = y * x + diagrv(zeros(rc,rc),c[i]);
      y = y * x + diagrv(zeros(rx,rx),c[i]); @changed rc to rx@
    27. Up through GAUSS 3.6.20, if a space is placed immediately after previously #define'd flag in #ifdef statement, a code between #ifdef-#endif is not compiled. For example, the following code
      #define A
      print "---";
      #ifdef A 
      /*space ^ */
      /*here  | */
         print "A";
      print "===";
      as its output.
    28. In order to assign values to a sequence of names, use varput() and ftocv(). For example,
      for i(1,5,1);
    29. It is possible to call GAUSS from a batch file, using tgauss.exe. Since tgauss doesn't use a user interface, and can be much faster. Just create a dos batch file containing lines like c:\gauss\tgauss.exe -b c:\mypath\myfile.gau where you need to adjust the paths appropriately. The -b flag is used to call GAUSS in batch mode. Save the batch file (e.g., as filename.bat). As with any batch file, you can double click it to run it or run it from the command line.
    30. Suppose you want to print a large matrix to a text file. You may be able to just reset 'outwidth' to provide the needed line length, but outwidth has a maximum value. If your matrix is too big for this, convert the matrix into a string array using ftostrC and then write it to a file using fputst. Ron Schoenberg provided the following example (given the matrix x):
          sa = ftostrC(x,"%8.4lf");
          fp = fopen("matrix.asc","w");
          call fputst(fp,sa);
          fp = close(fp);

    Aspell user hints.

    Installation hints for Windows users: Go to and get the setup program, available as a complete installer. Run the setup program. Next get and install the a dictionary (e.g., the English dictionary). Now you are read to run aspell. E.g., at a command window type c:\programs\aspell\bin\aspell check c:\temp.tex (assuming that's where you put it, of course). You may want to change your path so that aspell is in it.
    Running Aspell from Vim: To run aspell from Vim, just add a mapping to your _vimrc or .vimrc file. Here are two examples. The first assigns spell checking to the familiar F7 key.
    map <f7> <esc>:w<cr>:!aspell check %<cr>:e! %<cr> 
    (Use the complete path if you didn't put aspell in your path). Note that if you take this approach the changes are permanent! (I.e., you cannot "undo" them.) Leave out the last command if you want to be able to change your mind. Here is a slightly different version.
    map <Leader>s <esc>:w<cr>:!aspell --dont-backup -c %<cr>:e! %<cr><cr> 
    To understand these, see the Aspell and Vim helpfiles! In either case, make sure aspell is in your path or use the fully qualified file name.

    Windows command shell hints.

    Redirection: see Microsoft's documentation. I often need to redirect all script output to a text file, both stdout and stderr. Example:
    c:\Python27\python.exe -c "import scipy;scipy.test()" >c:\file.txt 2>&1
    In PowerShell, use the out-file commandlet.

    PyX user hints.

    Installing PyX from SVN gets you the most recent changes and is trivially easy. (Preliminary: install SVN.) At the command line, create a build directory with svn co pyx and then change to your new pyx directory and enter install.
    PyX will produce EPS files with negative bounding box parameters. This is perfectly legal but confuses some applications. In particular, LaTeX users of dvipdfm may see a clipped figure. Fixing this is simple: ask PyX to write your figure with a specified paperformat. E.g.,
    One lovely use of PyX is to include an existing EPS file and add things to it (e.g., LaTeX text, or JPEG graphics). When you write out the result with writeEPSfile method of your canvas, be sure to set bboxenlarge=False to ensure a "tight" bounding box.

    For 2-D data plots, Titus Winters has written a nice PyX tutorial.

    Be sure to see the examples and gallery on the PyX website. I provide a heavily commented elaboration of the shaded integral example.

    For a simplified interface to PyX, you can add PyXgraph or, if you would like an interactive command line interface, PyXPlot. For geometry, geoPyX is limited but useful.)

    MiKTeX user hints.

    MiKTeX comes with dvipdfm configured as
    D "mgs.exe -q -sPAPERSIZE=a0 -sDEVICE=pdfwrite -dCompatibilityLevel=1.2 -dUseFlateCompression=true -dNOPAUSE -sOutputFile=\"%o\" \"%i\" -c quit"
    Some applications (e.g., PyX) produce negative bounding boxes. (See Pyx hints for a PyX solution.) While perfectly legal, this creates a problem for dvipdfm. During conversion to PDF, the EPS files are clipped at the papersize boundary rather than the figure boundary. I don't know how to force pdfwrite to behave nicely, and epstopdf does not work because it intentionally alters the bounding box for the PDF file. You can try it though by changing the MiKTeX configuration to:
    D "epstopdf --outfile=\"%o\" \"%i\" "
    If you are generating PDF files directly, you can also feed PDF images directly to dvipdfm using specials:
    	\special{pdf:image (mydir/myfile.pdf)}

    TextPad user hints.

    Adding a program as a Tool:

    1. From TextPad's Configure menu, select Preferences.
    2. Click on Tools, then click on the Add button, then select Program...
    3. The "Select a File" window pops up. Select the executable file (.exe) that you want to add as a tool.
    4. Click on Apply. Double click on "Tools" and you should see your executable file listed as a tool. Click on it, and you will see some dialog boxes. Make sure they are correctly filled, being especially careful with the "Parameters" box. (You will find it useful to look at TextPad's help on Tool Parameter Macros.)

    Adding TeX as a Tool (thanks to Alan Cibils):

    To LaTeX (the disk copy of!) your current document, select your TeX from the "Tools" pull-down menu.
    IMPORTANT: When you LaTeX a document in TextPad, a Command Results document is opened. When LaTeX finishes executing, it displays the message: Process completed with exit code 1. YOU MUST CLOSE THIS WINDOW AFTER EVERY RUN. Failure to do this will cause subsequent LaTeX or BibTeX runs to hang.

    To avoid the afore mentioned problem, I recommend running MikTeX as a DOS command, which can also be done as a TextPad tool. If you decide on this option, go to the TextPad menus Configure, Preferences, Tools. Click on add and select DOS Command. A window will appear in which you type the full path to the command you wish to execute. For example, for MiKTeX type:
    c:\texmf\miktex\bin\texify.exe $BaseName (If you use the default MiKTeX installation, you do not need a fully qualified path name.)
    Click OK, then Apply, then go to the tool you just added (it should be listed under Tools when you expand it) and make sure to UNcheck the "capture output" box.
    If you do all of this, then when you select the tool, a DOS window will pop open and show you the execution results. This is useful for when you have compilation errors, you can hit enter, "X" or whatever, whereas the other way you can't.
    Older versions of TextPad have a small bug: the DOS window closes upon the end of execution of your tool, whether the close window box is clicked or not.

    Adding BibTeX as a Tool:

    Note that the texify command offered in MikTeX automatically runs BibTex for you.

    To LaTeX your current document, select Bibtex32 from the "Tools" pull-down menu.
    IMPORTANT: When you LaTeX a document in TextPad, a Command Results document is opened. When LaTeX finishes executing, it displays the message: Process completed with exit code 1. YOU MUST CLOSE THIS WINDOW AFTER EVERY RUN. Failure to do this will cause subsequent LaTeX or BibTeX runs to hang.

    Add a .dvi viewer to your Tools, to view your LaTeXed document:

    To view your LaTeXed document, select your viewer from the Tools pull-down menu.

    Since you can choose to have MikTeX produce .pdf files instead of .dvi files, you may wish to add the Adobe .pdf viewer to your Tools, to view your LaTeXed document:

    To view your LaTeXed document, select your viewer from the Tools pull-down menu.

    Reassign Keyboard Short-Cuts

    Reassign Ctrl-e to WordRightEnd as follows: choose Customize from the Configure menu, select the Keyboard tab, choose Cursor from the Categories list, choose WordRightEnd from the Commands list, type Ctrl+e in the Press New Shortcut box, click Assign, and click OK. You will then want to assign Ctrl+Shift+E to WordRightEndExtend.

    User Hints: Mathematica

    Get all secion header cells with Cells[CellStyle → {"Section"}]. The result is a list of cell objects, on which you can set any options by assigning to its CurrentValue:

     CurrentValue[cell, ShowGroupOpener] = True,
     {cell, Cells[CellStyle → {"Section"}]}
    Using version control with Mma notebooks is a bit messy but can be made fairly workable. In a Mathematica notebook that you want to put under version control, go to Format » Option Inspector... in the menu. In the top-left dropdown, set the option scope Selected Notebook. Find FileOutlineCache under Notebook Options » Private Notebook Options, and set this option to False. (This may make opening the notebooks a bit slower.) Similarly, under Editing Options, set TrackCellChangeTimes to False
    Command approach:
     PrivateNotebookOptions → {"FileOutlineCache" → False},
      TrackCellChangeTimes → False
    Save your notebook and it will save with much less clutter and many fewer changes. To do even better, as a preliminary to committing your changes, delete all your output cells and then save your notebook. (You can use the menus or FrontEndTokenExecute["DeleteGeneratedCells"] for this.)
    You may find SciDraw helpful for better looking Mathematica drawings. In order to make your Mathematica plots more stylistically compatible with a LaTeX document, consider using:

    Scientific Notebook and Scientific Word user hints

    On campus: To fix font problems on the display, make sure the Windows Multilingual Support is installed. If that is not an option, go to Tools, User Setup, Font Mapping and check the Use Symbol Font mapping box.

    Be sure to learn the keyboard short-cuts: see especially the help for Key Prefixes. For example, you can logical negation as ctrl-s,N and the partial derivative sign (∂) as ctrl-s,d. (See the Help search.) Also, I strongly recommend changing the default settings under Tools, UserSetup, Math: set it so that two spaces switches to math and a space at the end of math switches to text.

    Be sure to visit Yngve Svendsen's Scientific WorkPlace Unofficial Homepage (although I have had some connection difficulties).

    Graphics: When using the JPGs or TIFFs, edit the file tcispech.ini and change the setting for MaxGraphicsMemoryUsage. Try doubling or tripling the default setting of 4000. When using EPS graphics, save your document using the Portable LaTeX filter and change the options for the graphicx package to use dvips. The graphics will only appear when printing, but they should appear.

    To use Scientific Word on the EagleNet, you will first need to install some (non-executable) files on your computer (which ought to have say 30Meg free disk space and 8Meg RAM); do this by entering "enetupgr sword" (without the quotes) at the DOS prompt. (Note: be sure to keep your macros in a separate file, which you \input in your preamble.) Be sure to see the Scientific Workplace Unofficial Homepage, which include instructions on how to use MiKTeX with SWP.

    You can get all the functionality of Scientific Word except for the typesetting and associated TeX functions in Scientific Notebook for Windows 9x and NT. Many useful functions are in the free Demo version, although this version cannot print, save, or copy from read/write documents. While you are on the AU campus, you are in luck: Scientific Notebook is available as an EagleNet Application.
    Users must first open EagleNet Applications. (If you don't have EagleNet on your Start menu, just pick Start, Run and the type `nal' in the dialog box.) You should see a Math/Statistics/Research folder on the left, in which you will find the Scientific Notebook icon. Double click the icon to run the program.
    The first time you run the program, several system files will be copied down to your hard drive before the program opens. Each time after that initial setup the program will run as soon as the icon is double-clicked. The program is accessible to the major academic user groups and all student users. Since it is a 5-user license, you may occasionally find you have to wait until someone logs off to use Scientific Notebook. I do not expect this to happen, but if it happens to you let me know and I will apply for more copies.

    Subversion User hints

    Subversion is easy to use and excellent for individuals and small groups. Here is a ten minute tutorial. Windows users who prefer to avoid the command line can just use TortoiseSVN.

    Remember your work cycle must be a very rigid UMUC sequence: 1. before you start work, *update* 2. *modify* your working copy (do your work, and test it) 3. *update* again (and resolve any conflicts) 4. as soon as possible, *commit* your changes
    To work across platforms, you must not use file or folder names that are illegal on any of the platforms. Windows users cannot use the following (which are reserved device names): CON, PRN, AUX, NUL, COM1, COM2, COM3, COM4, COM5, COM6, COM7, COM8, COM9, LPT1, LPT2, LPT3, LPT4, LPT5, LPT6, LPT7, LPT8, and LPT9. They also cannot use these names followed by an extension, for example, NUL.txt or com1.png.
    Suppose a contributor creates com1.png on a Mac and commits it to the Subversion repository. When Windows users try to update, the update fails horribly. (How horribly? Unfixably. They will have to get a new checkout after the problem is fixed.) The contributor of the file should ``svn rename`` it to a name legal for Windows, and commit that change. Then the problem for Windows users will go away.
    Bottom line: never ever use such file or folder names!

    How to resolve Subversion Conflicts

    Suppose you want to add a directory mydir under version control, but it contains a bunch of files you do not wish to track. You can add the directory but not the files as follows: svn add --non-recursive mydir.

    One disappointing thing: it would be nice to add a directory and all its files except those matching the svn:ignore property. Subversion does not currently support this. Here is one work around.

    You can set an editor (e.g., for commit messages and property editing) in the Subversion config file. On Windows you should be able (after installing Subversion) to find it here: \Users\YourUserName\AppData\Roaming\Subversion. Search for editor-cmd and follow the instructions. (On Windows, I am currently using editor-cmd = C:/programs/Vim/vim74/gvim.exe -f. If you use a path with spaces in it, you will have to quote it.)

    Once you ensure that Subversion has access to a text editor, you can edit the list of patterns it will ignore in any directory. E.g., svn propedit svn:ignore . (note the final period) will edit the patterns for the current directory.

    svn diff -r HEAD myfile shows differences between myfile and the repository. svn status --show-updates shows you which files in the repository have changed.

    It is useful (and probably best practice) for any plain text files to svn propset svn:eol-style=native *filename*. This is not the default, but each user can make it a personal default by editing the [autoprops] of their Subversion config file. From Subversion 1.8 onwards, it can be set as a default at the repository level with Repository Dictated Configuration.

    It is a good idea to lock binary files, to help ensure only one person changes them at a time: svn propset:needs-lock yes *filename*.

    Docutils User hints

    If you are using rst2latex, it will include the following in your header: \usepackage[T1]{fontenc} . Unfortunately, on its own, this will probably lead to the use of bitmap fonts and thus "fuzzy" screen output. According to Guenter Milde, there are several reasons for this choice. The real question is what to do about it. The recommended workaround is to select a T1-encoded "Type 1" (vector) font. For example you can use Latin Modern either by adding it to your LaTeX style sheet (\usepackage{lmodern}) or by using a Docutils command line argument (--stylesheet=lmodern). Another popular solution is the "Times/Helvetica/Courier" combination, which you can get by adding the following to your style sheet:
    \usepackage{mathptmx}            % Times for serif and math
    \usepackage[scaled=.90]{helvet}  % downscaled Helvetica for sans serif
    \usepackage{courier}             % Courier for teletype (mono-space)
    or with the following Docutils command line options: --documentoptions="scaled=.90" --stylesheet="mathptmx,helvet,courier".
    You can also just turn it off with the following Docutils command line option: --font-encoding=OT1 (or even --font-encoding="").

    JabRef User hints

    1. Go to and click Web Start, to start the JabRef reference manager.
    2. To enter a new article, click the green + and click Article. To enter a new working paper, click the green + and click TechReport.
    3. Multiple authors are separated by and (*not* by commas).
    4. Do not put quotes around titles.
    5. Use two hyphens for page number ranges (e.g., 99--101). For page numbers, put only the page ranges (e.g., no "pp.").
    6. For articles you will often have volume and issue number data, like vol. 2 no. 1 or 2(1). The volume is entered on the Required tab and the issue number is entered on the Optional tab.
    7. Months are entered like this: #jan#, #feb#, #mar#, etc. (Three letter abbreviations bracketed by hash marks.)
    8. The url field is under the General tab.
    9. If you are missing information for any Required fields, you can often get the missing information from RePEc ( if you do a Google search on the title.

    Matplotlib User Hints

    Matplotlib is included in many Python distributions, including Enthought Canopy. If you install it your self, be sure to read the installation instructions. Before installing Matplotlib, you will certainly need to install NumPy.

    To get the most from Matplotlib, you need to clearly understand the difference between a figure instance and an axes instance.
    A figure instance the basic container for an entire chart, like a figure in a journal article. A figure instance can contain multiple axes instances, each of which can be used for plotting or drawing.
    We can call the add_subplot method to add non-overlapping axes instances to our figure. (But if you call add_subplot again in a way that overlaps your previous subplot, this **replaces** your previous axes in the current figure.)

    import matplotlib.pyplot as plt
    fig1 = plt.figure()
    ax1 = fig1.add_subplot(1,1,1)
    #now draw on ax1
    fig2 = plt.figure()
    ax2a = fig2.add_subplot(2,1,1)
    #now draw on ax2a
    ax2b = fig2.add_subplot(2,1,2)
    #now draw on ax2b

    You can use the subplots convenience function to do figure and axes creation at one go.

    fig, ((ax1,ax2),(ax3,ax4)) = plt.subplots(2,2)

    Here is a simple example for creating a figure and displaying it onscreen. (For completeness, we illustrate the explicit choce of a GUI “back end”, although this is usually not necessary.) Note the hierarchical approach: the pyplot interface creates the figure object, the figure object creates the axes objects, and the axes objects create the other objects (including, implicitly, the axis objects).

    # preliminaries
    import matplotlib
    matplotlib.use('TkAgg')  #note choice of GUI backend!
    import matplotlib.pyplot as plt
    # create a new figure
    fig = plt.figure(1)
    ax = fig.add_subplot(1,1,1)
    ax.set_title('Easy Line Plot')
    # display your figure onscreen 

    Here is a simple example for creating a figure and saving it, without displaying it onscreen. (I.e., no GUI.)

    # preliminaries
    import numpy
    import matplotlib
    matplotlib.use('Agg')  #note choice of backend: no GUI!
    import matplotlib.pyplot as plt
    #create some artificial data for this example
    x, y = numpy.random.rand(2,50)  #the data (50 points)
    # create a new figure 
    fig = pyplot.figure()
    #add an axes instance to draw on
    ax = fig.add_subplot(111)
    #do your drawing
    ax.plot(x, y, marker='.', linestyle='', markerfacecolor='blue')
    ax.set_title('Scatter dots')
    # save figure to file
    fig.savefig('/temp/myfig.png', dpi=96)

    Matplotlib's pyplot facilities are convenience facilities for the user with modest needs. For some examples (e.g., embedding Matplotlib in other applications) it is important not to use these facilities. In this case, you can altogether avoid using pyplot, but you will need to handle figure and canvas creation yourself.

    # example by John Hunter.
    from matplotlib.backends.backend_svg import FigureCanvasSVG
    from matplotlib.figure import Figure
    fig = Figure()
    ax = fig.add_subplot(211)
    canvas = FigureCanvasSVG(fig)

    As another example, here is our earlier example redone to be more fully object oriented.

    # preliminaries
    from matplotlib.backends.backend_agg import FigureCanvasAgg as FigureCanvas
    from matplotlib.figure import Figure
    # create a new figure
    fig = Figure()
    ax = fig.add_subplot(111)
    ax.set_title('Easy OO Line Plot')
    canvas = FigureCanvas(fig)

    All Matplotlib plotting functions trigger autoscaling. This affect, e.g., setting the axes limits. To prevent your limits from being overridden, call xlim/ylim after any plotting functions. Alternatively, you can turn autoscaling off:

    Subplots are pretty close to each other: if you stack two and label both horizontal axes, the top plot's label overlaps the bottom plot's axis. Fix this by using Figure.subplots_adjust, like this:
    subplots_adjust(hspace=0.4)  # default: 0.2

    Fans of Edward Tufte may wish to look at etframes.

    Note that if you set_aspect on a axes instance you will still need to apply_aspect as well (or call draw).

    Matplotlib is capable of fully object oriented graphics programming. There is a very useful tutorial. If you want to embed Matplotlib in a GUI, you must use the object oriented API and not pyplot. The Matplotlib website includes examples of embedding. (For example,

    wxmpl is supposed to be the best way to integrate Matplotlib into wx.

    Changing Matplotlib backends: you have to call 'use' before importing matplotlib.pyplot; the backend choice must be set at the time of the matplotlib.pyplot import. For example,
    import matplotlib
    from matplotlib.pyplot import *
    If you need to control the placement of a Matplotlib figure, you need to work a bit because it is "back end" dependent. According to John Hunter, "the matplotlib Figure is embedded in a FigureCanvas which is typically a GUI widget embedded in a GUI Window. In the pylab interface, the canvas is managed by a FigureManager, which has a window attribute on most of the backends." He provides some examples.
    # GTK backend 
    from pylab import * 
    import gtk
    manager = get_current_fig_manager()
    # see gtk.Window class docs at
    manager = get_current_fig_manager() 
    # see gtk.Window class docs at
    Another user offered a WX example:
    # WX* backend (manager.window is a wxFrame)
    manager = get_current_fig_manager()
    #or: manager.window.Centre()
    Todd Miller offered a TkAgg example:
    # set the window position to X=200, Y=300
    To get the equivalent of gnuplot's set xzeroaxis and set yzeroaxis, use axhline() and axyline(). Prior to version 0.63, Matplotlib does not have an equivalent of gnuplot's set xzeroaxis. (This draws a horizontal line across the xrange at y=0.) John Hunter explains how to achieve the same effect, and it provides many insights into matplotlib. "What you want to do is plot a line where the y coord is in data units and equal to zero, and the x coord is in axes units and spans the xrange. The line stretches from left to right regardless of the xlimits. Each axes provides a data transform and an axes transform, so that you can specify lines/text/etc in either coord system; axes coords are 0,0 for lower left and 1,1 for upper right. But what you want to do is mix the data and axes transforms for x and y. The transforms module provides a helper function for this. The following code is freestyle (untested) but should work:
    from matplotlib.transforms import blend_xy_sep_transform
    ax = gca()
    trans = blend_xy_sep_transform( ax.transAxes, ax.transData)
    plot([0,1], [0,0], transform=trans)
    You can set an arbitrary color for any matplotlib plot. E.g., pyplot.plot(d, f(d), color = (1,0,0)).
    To get rid of the redundant ticks on the top and right edges:
    ax = pyplot.gca()
    Note that you should call show only once in a script. But sometimes you want to alternate seeing plots and doing computations. See for a discussion.
    In an Axes3D object, the .xaxis and .yaxis refer to the axis for the draw space, not to the the 3d projected axis. The latter are .w_xaxis, .w_yaxis, and .w_zaxis. (Thanks to Ben Root for pointing this out.)

    EViews User hints

    Eviews 6 appears to have some bugs and some annoyances:

    Kawa offers a very useful collection of supplemental programs and data in EViews format, including the NBER historical series and the PENN World Tables.

    EViews 5 can read and write many data formats.

    Up to version 4, the save command is not correctly documented in the EViews documentation. If you provide a filename as an argument, it is really equivalent to File.SaveAs. That is, it saves under the new name and changes the name of the current workfile. In fact, as of version 4, this is the only way to change the name of a workfile in RAM.

    In Eviews3, you could use scat(r) to get in a single graph multiple scatter plots with their regression lines. This is not possible in EViews4, where the replacement linefit command only uses the first two members of its group argument.

    missing observations: @obs(x[,s]) gives you the number of non-missing observations in x in the current sample or in sample "s" if the option is listed. One can access the number of non-missing observations for a group with @comobs.

    String substitution is very useful in EViews. For example, you might write set a string variable, say %param1, equal to a complicated function of parameters, which recurs in a system estimation. As a simple example:
    %param1 = "c(1) + c(2)"
    system1.add y = c(1)* x1 + x2/({%param1})
    Oddly, string substitution only works for procs (commands) and not for assignments (using = ). For example,
    scalar trythis = ({%param1})
    doesn't work. (In this case, EViews will look for an object named c(1)+c(2) without evaluating the expression.)

    Backwards compatibility: Eviews 3 and EViews 4 are not fully compatible. See the version 4 User's Guide page 532: "A Note on EViews Backward Compatibility" and page 580/581: "Converting from Version 3 Sspace" for some changes from version 3 to version 4. Briefly:

    1. "makeendog" is changed to "makegroup"
    2. "makestat" is changed to "makestats" (do not confuse with "makestate")
    3. with solving models the assign statement is replaced by scenarios, and the default scenario operates as if you had specified "assign @all _0" in version 3; hence in version 4 endogenous variables are not overwritten but new variables are created instead.
    4. legend(s) becomes legend columns(1)
    5. the r option for scat is not supported.

    ffmpeg User hints

    To reduce the image resolution, use a video filter. E.g., ffmpeg -i myinfinle.mp4 -vf scale=iw/2:ih/2 myoutfile.mp4 (Here iw and ih hold the input resolution width and height. See the filters documentation.)

    To extract part of an MP4 file, copy the audio (acodec) and video (vcodec) for a stime range. E.g., ffmpeg -ss 00:0:00 -to 00:10:00 -i myinfile.mp4 -acodec copy -vcodec copy myoutfile.mp4 extracts the first 10 minutes. (See the documentation of time durations.)

    You can pipe a series of frames (e.g., Matplotlib graphs) to FFMPEG to produce an animated graph.

    You can create PNG files named (frame000.png, frame001.png, etc) and do ffmpeg -b 8000 -f mpeg2video -r 30 -i frame%03d.png movie.mpeg. You can also produce an AVI like this: ffmpeg -r 10 -i frame%03d.png -f avi -vcodec msmpeg4v2 -y movie.avi. Works great! Note that AVI supports very low framerates; MPEG does not.

    Vim User hints

    Vim is free and open source software; get it here: If you are new to Vim, get some documentation. Be sure to do the Vim tutorial and to read the help for "navigation". If you want additional documentation, consider Vi Improved—Vim by Steve Oualline or Vim ge-packt by Reinhard Wobst.

    Windows installation hints:
    Installation creats a vim directory. Below that is a vimfiles directory. For Windows users who are new to Vim, i. copy my share_vimrc.vim file to your vimfiles directory, and then ii. look for the _vimrc file in your vim directory and add at the bottom of that _vimrc file the following command: source $VIM/vimfiles/share_vimrc

    Getting started:
    Look for a file called vimtutor or vimtutor.bat. Set aside 20 minutes for the tutorial, and then run the tutorial (e.g., by clicking the file icon). When you want to go deeper, look at the Vimcasts videos.

    The Vim cursor stays in the viewing window, in contrast with some Windows applications. If you want to cursor down and then return to where you started, set a mark before moving. (See :h marks.) But, if you want to be able to page down and then page back up to where you started, use ctrl-d and ctrl-u rather than Pg-Dn and Pg-Up.

    Digraphs allow simple and intuitive keyboard entry of unicode characters. Vim digraphs are based on RFC 1345. See ``:h digraph-table``.

    To find the non-ASCII characters in a file, use a regular expression. For example, to search for all characters outside of the ASCII range, search for [^\x00-\x7F]. (Alternatively, use decimal values: [^\d0-\d127].) However to search for individual non-ASCII characters, see the help for %o.

    Vim provides excellent syntax highlighting in various color schemes. There is an online gallery. You can get a list of all the installed colorschemes by entering colorscheme <ctrl+d> at the Vim command line. (Do not forget the space.) I particularly like koehler, a rather dark color scheme. (Download from GitHub.) You can set this by entering colorscheme koehler at the Vim commmand line. Another popular dark color scheme is solarized. (Download from GitHub.) To determine your current color scheme, enter echo g:colors_name at the command line. (If you did the tutorial, you know you can reach the command line by pressing Esc

    A buffer is a file open for editing. Use the :ls command to list your current buffers. This list is not interactive. However, while looking at this list, you can e.g. enter :b10 to switch to buffer 10.

    Vim uses tabs in a unique way. They are called tab pages, and they are essentially an arrangement of windows open on some of your buffers. (A buffer is a file that has been opened for editing; a window is an area to view a buffer, and a tab page is a collection of windows.

    Vim has excellent syntax highlighting. However, by default it does not use a lot of context. In many settings, I find it useful to set syn sych minlines=300. When necessary, you can syn sych fromstart, but this may cause a slowdown in very large files. (See the help for syn-sync.

    Vim cooperates with some scipting languages, including Python. See :help :python, :help python-examples, and especially :help python-vim. Start with
    :python import vim
    :py cb = vim.current.buffer
    Note that cb is now an iterable container of strings. You can assign to the i-th line by assigning to cb[i].
    Exiting a buffer: it is usually recommended to exit a buffer using bnext. However I sometimes prefer to delete a buffer as I leave it. Examining the help for ``bar``, we see that ``:w|bd`` can be used to write and then delete a buffer.
    With minimal effort, Vim cooperates wonderfully with LaTeX. Windows users should install MiKTEX, and then perhaps add my MiKTeX menus to the _gvimrc file.

    Vim 7+ includes spell checking: setl spell. Spell checking is affected by syntax highlighting, which identifies groups (e.g., comments). For spell-checking longer LaTeX documents, I find I need to syntax spell toplevel. This ensures spellchecking of text that is not in a syntax item. To make this permanent, you can include it in ~/.vim/after/syntax/tex.vim. You can also set the values of syn sync maxlines and syn sync minlines in the same file. (See the help for :syn-sync.) If the file is not too big, you can even always start from the top with syntax sync fromstart.

    Learn to yank to registers. For example, to yank (i.e., copy) the entire buffer you are working on into the clipboard, use :%y+. (The colon move to command mode, the percent sign picks all lines, the y is to yank, and the plus sign picks the clipboard register.)

    To remove the BOM from a file, set nobomb.

    There is a nice book devoted to learning Vim: Vi IMproved by Steve Oualline, also called the VIM book. You can view the Vim book online.

    Be sure to try Jeff Lanzarotta's bufexplorer.vim. (One warning: as of early 2001, the buffer delete command in bufexplorer uses bd!, so changed buffers will be deleted upon request without a warning. Change this to bd.)

    Thesaurus: You can adapt the Moby thesaurus for use with Vim.

    Ctags is no longer included with Vim (it has grown into a project of its own.)

    A useful plugin is Universal Text Linking, which executes URLs in plain text (e.g. call webbrowser, link MS Word documents, text files, embed vim commands).

    Vim supports folding! You may find it useful to begin your writing with an outliner.

    For user questions you can turn to the Vim mailing list. There is an archive.

    Note that for all systems the Vim distribution is split into an archive with runtime files (documentation, syntax files, etc.) and a binary and/or source archive. You should download at least two archives!

    You can find many useful hints at Softpanorama Western Orthodox Editors Page.

    Count the occurences of a pattern with :%s/pattern//ng. Count the number of words in a file buffer with g ctrl-g (in normal mode). Count the number of words in a visually selected block with g ctrl-g (in visual mode).

    To change line spacing see the help for linespace. To change the font in gvim, you can use the menu dialog Edit/SelectFont, or you can include your selection in your _gvimrc file. (I use set guifont=andale_mono:h9,courier_new:h10.) Text editors should only use monospace fonts.

    If you type very slowly, you need to slow down Vim's macro recognition. See the help for map-typing. I recommend resetting timeoutlen according to the example in the help for timeoutlen.

    '. takes you to your last edit

    PostScript User hints

    If you want to fill an entire page with background color, you do not need to know the dimension of the page. Just start your PostScript program by setting the color and then giving the command clippath fill. (Of course few printers can print to the exact edge of the page.)

    Transparency: PostScript uses an opaque imaging model. It does not support transparency. However, there is a workaround available if you have Adobe Distiller. You can add a "PDFMark" to your Postscript file to set the transparency for an object. When Adobe Distiller processes this PDFMark, it will set the transparency for the object. Lancaster offered an early discussion. (Start there rather than with the pdfmark Reference Manual.) Also, note that GhostScript supports transparency.

    LaTeX User hints

    Use pdflatex to compile your documents. Use tikz for your drawings.

    To get double spacing, use \usepackage{setspace} \doublespacing. This should leave your figures and tables alone.

    Do not use the eqnarray environment. Lars Madsen explains why. Instead use one of the beautiful amsmath environments.

    For nice theorem and proof formatting, consider combining the following three packages: amsthm, thmtools, and thmbox. (The last is just to redefine the proof environment.

    If you use linerange markers with the listings package, make sure your markers do not include hyphens. (These will be treated as indicating the linerange break.)

    See Voss's FAQ supplement.

    The pict2e package provides much needed additional functionality to the LaTeX picture environment. As of June 2004, however, there is a small bug: be sure to end qbezier and cbezier commands with an immediate comment character, to avoid breaks in line continuity.

    Using LaTeX to produce PDF files with graphics is pretty easy as long as you are producing clean graphics. But you need to pay attention to some details.

    If you are producing your PDF files with dvipdfm, then you will find it easiest to use EPS graphics. (EPS is also the primary standard for publishers.) Historically this has been my preference, but lately I've been working more with pdflatex.

    1. Problem: a few applications (mostly on Windows) do not produce EPS graphics. In this case, you can produce JPEG files and either i. learn how to include .jpg files, or ii. convert JPEG to EPS or sam2p (which is fast and easy), iii. switch to pdflatex
    2. If you are using pdflatex to produce your PDF files, you cannot include EPS graphics directly. You can either
      1. convert your EPS graphics to PDF, using eps2pdf
      2. produce your graphics in a supported format, such as PDF or JPEG
      3. set up your document to convert EPS to PDF on the fly
      Trevorrow suggests leaving off the extension (.eps) when you includegraphics and putting the following in your preamble:
      % we are running LaTeX, not pdflatex
      % we are running pdflatex, so convert .eps files to .pdf
    In any case, use the graphicx package and include your graphics. Example:
        \caption{Test Figure}
    Note the forward slash in the path name, as always with TeX. (You can omit the path if you i. keep your graphics files in the same folder as your .tex file, or ii. configure your LaTeX to look in the folder containing your graphics files.)

    On the one hand it is easy enough to zip up all your EPS files with the embedding LaTeX document. On the other hand, one of the attractions of the LaTeX picture environment was that your drawings were present in your .tex file. Here is a way to get almost the same thing with EPS files. Add the moreverb package. Put your EPS file in a verbatimwrite environment in your .tex file. You will have to provide a file name that it will be written to, but once it is written you can read it into a figure as usual. (An alternatie is to generate your figures with embedded gnuplot commands, using the egplot package.)

    If you need to wrap text around a figure, use the wrapfig package. Include the package (in your preamble) and then the syntax is almost like that for the figure environment.
    \caption{Example Caption}

    BibTeX User hints

    Gnuplot User hints

    Since gnuplot is a plotting package, it offers limited data manipulation facilities. These can be increased by the use of gawk, which gnuplot can talk to nicely. Alternatively you can manage your data in Python and use a Python interface to gnuplot, such as If you use Jupyter’s IPython, which is a good idea, you can try the included enhanced version: In either case, if you use gnuplot's fit command, you should consider how to get the results back into Python.

    There are palettes (themes) for gnuplot.

    Be sure to use the valuable documentation that gnuplot includes. Begin by typing "help introduction" at the gnuplot prompt, since this explains the syntax used in the help documents.
    Note: if you download gnuplot for Windows as a ``.zip`` file, the gnuplot help file (wgnuplot.chm) does not work properly. Windows blocks it. To unblock it, right click on the file, pick ``Properties``, and then press ``Unblock``. Please be sure you have an official release before doing this.

    The command to load a script is load. Provide your file name in quotes, using forward slashes.

    Remember that gnuplot distinguishes integers and floats by the presence of a decimal point. Make sure your real numbers are recognized as real by including a decimal point. Integer division produces an integer result. For example, $3/2=1$ but $3./2=1.5$.

    To easily plot a cobweb plot (i.e., Verhulst diagram with gnuplot, see this article.

    Setting the color of plots is not well documented. Do it by setting the linetype variable. To see the linetypes, type test at the gnuplot command line.

    To draw a 3-D scatterplot, try
    splot 'data.asc' with points palette

    You can draw a line segment by setting an arrow with the "nohead" parameter.

    Suppose you want a log scale on the y axis but you want to display the numbers as base raised to an exponent.
    set log y 2.0
    set format y '2^%L'
    The basic way to place single points is with the label command. Use the 'front' option to make sure that they are placed after lines are plotted. If you know the location, you can also place a point as a data plot of a single value. E.g.,
    plot '-' with points ps 3
    1 1
    centers a point at (1,1).

    If you would like to use gnuplot to create a drawing with no plots in it this is possible. Just be sure to set the yrange, then plot 0/0 after all your drawing commands.

    Up to version 3.8j, gnuplot arrowheads are not offset to account for their miter. This means they generally extend past the point where you expect them to stop. There is no satisfactory workaround, but you can try drawing a filled arrow with a very narrow pen and then redrawing just the tail (as a headless "arrow") with a thicker pen.

    Kawano offers a brief overview of how to generate numerical plots using gnuplot.

    FireFox hints

    I consider one extension to be indispensible: Session Manager. Another interesting extension is Taboo which allows you to replace the practice of keeping many tabs open with a single page of site links.

    Windows hints

    The default in Windows is to prevent you from seeing the extension to a filename. You can fix this as follows: open the Control Panel and open Folder Options, then pick the View tab and uncheck Hide File Extensions.

    Windows File Search: Basic file search in Windows is indexed. To search all files, use advanced search and check "Include non-indexed". It can also be useful to change what is indexed.
    To drop to a command shell from File Explorer, use Shift+RightClick to open a menu with that option.

    Windows 10 PowerToys includes a Keyboard Manager, for key remapping. This is useful for mapping CapsLock to Ctrl.

    Opera hints

    Opera hints:
    You can disable the Opera splash screen by adding/editing this line in your opera.ini file, [user prefs] section.
    You can pre-empt certain misconfigured servers that do not correctly handle RFC2616 add the following to the [Adv User Prefs] section of opera6.ini:
    HTTP Accept Charset=windows-1252, utf-8, utf-16, iso-8859-1, iso-8859-15;q=0.6, *;q=0.1
    The article on the Opera website about how to install the Adobe SVG Viewer states: "Please note: SVG may not work correctly, some users report that the images do not animate, even though they are displayed." You can overcome this with a bit of work. Install the Adobe SVG Viewer version 6 beta release of v6 of the Adobe SVG viewer, then copy three DLLs to the Opera Plugins directory. (The Adobe SVG Viewer v6 beta release passes almost all of the W3C SVG compliance suite.) Just follow the installation instructions. Once it is installed, copy the following files to the Opera\Program\Plugins directory:
    Restart Opera and test the plugin at the Adobe SVG test page. You should see 3 animated images.

    gretl hints

    To install gretl on an Intel Mac, make sure you do a 3-step install:
  • Make sure X11 is installed. (On Tiger you will need to install from your system disk. On Leopard, update your X11 from
  • install the GTK+ Framework,
  • install the **intel** disk image.
  • The last two are at

    LaTeX hints

    New LaTeX users on Windows may wish to read these hints

    Suppose you want to use a non-standard document class, say a thesis class to format your LaTeX thesis. You need LaTeX to be able to find the class (.cls) and class option (.clo) files. Here are some options for providing this information. (For more information, see the MiKTeX documentation.)

    To include graphics in your document, use the graphicx package. Keep your images grouped together, and point to them with the \graphicspath command. (Some knowledgeable users suggest setting TEXINPUTS instead, e.g., Note the odd syntax for \graphicspath: each path must be in braces and must end with a forward slash. If you want an images directory searched recursively (say because you put each chapters images a separate folder), end the path with a double slash. E.g., \graphicspath{{images-eps//}{images-pdf//}}.

    TikZ hints

    Most important: the TikZ/PGF manual is very helpful. Also, Crémer provides a useful introduction.

    The intersection coordinate systems are very useful. E.g., (A |- B) takes the first coordinate from A and the second coordinate from B. See the PyX manual.

    The arc command feels odd, especially to those with PostScript experience. Instead of inputting a circle center, you must give the start point for the arc. Suppose we want to draw an arc from from a circle centered at (1,1) with a radius of 2, say from 30 degrees to 60 degrees. You can do that like this: \draw (1,1) +(30:2) arc (30:60:2)

    Parametric plots are easy. E.g., \draw[domain=-20:0, smooth, variable=\t, samples=200] plot({cos(\t r)*exp(0.1*\t)}, {-sin(\t r)*exp(0.1*\t)});

    To place text along a line, use the sloped keyword: \draw (0,0) -- (5,1) node[above,very near end,sloped] {text}

    You can use tikzset to set the default arrowhead style for your document: \tikzset{>=latex}.

    Predefined line thicknesses are: ultra thin, very thin, thin (the default), semithick, thick, very thick, ultra thick. You can also set the line width explicitly: \draw[line width=0.5pt].

    A very light gray is black!10. A very dark gray is black!90.

    Use pgfplots for plots. The manual is good, and most tasks have very easy and obvious solutions. A few odd things are obscure. E.g., adding a zero-line to a plot.

    When supplying images to accompany articles, depending on the journal, the TikZ format may not be acceptable. Fortunately, it is easy to create standalone PDFs from TikZ figures. Use the standalone documentclass.

    % myfig.tex
    \documentclass[tikz]{standalone} %standalone figure
      \draw (0,0) -| (5,5) -| (10,10); %tightly cropped!

    LuaTeX hints

    See the LuaTeX manual.

    Recent TeX distributions include LuaTeX. The first thing most new users will wish to attemp with LuaTeX is compilation of legacy LaTeX documents. You can compile a LaTeX document named mydoc.tex with the command lualatex mydoc.

    Unicode hints

    To enter arbitrary Unicode characters on a Windows platform, see On Linux, consider the IBus input bus. If you use Vim on either platform, see help utf-8-typing (and help digraphs-default).

    Windows 10 hints

    To see your recent files, open the Run Dialog Box (keyboard shortcut Windows Key+R) and enter shell:recent. You can open (or delete) files here.

    To search for a phrase in a given type of file in FileExplorer, use the search bar on the upper right hand corner with the type: and content: prefixes. E.g., content:"my phrase" type:.tex. (First, make sure you are indexing the contents.)

    LyX hints

    If you are using Subversion for version control, make sure you set svn:eol-style to native on your .lyx files.

    To edit the LaTeX preamble, go to Document » Settings » LaTeX Preamble.

    Google Drive hints

    It is possible to provide a direct download link for items on your Google drive by making a small change in the URL. Here is an example for a Google spreadsheet. In a browser, open the file to edit and you'll see an address like, where the ID will be specific to your document. Just change the end of this to specify export and a format argument. For example,

    GnuWin32 hints

    GnuWin32 includes many useful GNU applications, including grep, sed, and awk. If you download everything (recommended), it will take a while. If you are just looking for a grep implementation, you can try grepwin. (License: GPL)

    Installation (with thanks to Arturo Servin) and the GetGnuWin32 developers.
    1. Download the getgnuwin32 project maintenance utility. (This will download and install the tools for you.)
    2. Run the executable file you downloaded. It will ask you for a destination to decompress the file. This will be a temporary directory: you will eventually delete it. I will assume you use c:\temp\getgnuwin32. (You can choose whatever you want.) To install, you can just follow the instructions. (If you want details you can check the readme file included with the package.)
    3. Comment: If you are behind a proxy, you may have to configure wget (using bin\wget.ini).
    4. Run download.bat. (This file is well documented. If you wish to change the defaults---e.g., the default mirror site for download--- edit the file before running it, and save your changes.) As it runs, download.bat will download the applications packages. This can take a while.
    5. When dowload.bat finishes running, run install.bat. (This will decompress the downloaded packages.)
    6. Copy the folder getgnuwin32\gnuwin32 to a new folder C:\programs\gnuwin32. (Or anywhere else you want.)
    7. Change to your newly filled C:\programs\gnuwin32 folder and run the file update-links.bat. This should complete the installation process (by updating any orphan links). You are now ready to use your new tools.

    There are more options. I do not use them and they can confuse some windows applications using the same name as those in the gnu package, so I will not explain. If you are interested, check the readme file.

    Adobe, PostScript, and Acrobat are registered trademarks of Adobe Systems, Incorporated. Apple and Macintosh are registered trademarks of Apple Computer, Incorporated. Windows and MS-DOS are trademarks of Microsoft Corporation.

    Copyright © 1998 & 2020 by Alan G. Isaac
    (mailto: a i s a a c @ a m e r i c a n . e d u)
    Some Rights Reserved: The content on this page is subject to the Creative Commons Attribution License version 2.0

    Top of This Page Economics Department Home Page