Syntax doldrums

Question

I am a long time user and fan of LaTeX, but I hate the syntax! I wish it could be more like Python or Haskell. The syntax of those programming languages is much more readable than LaTeX which is ironic because you are much more likely to spend time reading raw LaTeX than programming code.

I would like it to be something like

We have
\equation
    A =
    \bmatrix
        a & b
        c & d
where ...

Instead of

We have
\begin{equation}
    A =
    \begin{bmatrix}
        a & b \\
        c & d
    \end{bmatrix}
\end{equation}
where ...

This is a simple example but you all know how onerous the curly brackets \begin's and \end's can get in complicated examples

Is there an alternative interpreter for LaTeX that has nicer syntax?

Answer 1

You need a very recent expl3 (January 8th, 2011) for the following code to work (I added \ior_str_map_inline:nn a few minutes ago).

\RequirePackage{expl3,l3str}
\ExplSyntaxOn
\bool_new:N \g_strange_tmp_bool
\tl_new:N \g_strange_last_line_tl
\tl_new:N \g_strange_line_tl
\int_new:N \g_strange_indentation_int
\int_new:N \g_strange_last_indentation_int
\seq_new:N \g_strange_indentation_seq
\seq_new:N \g_strange_environments_seq
\tl_new:N \l_strange_tmpa_tl
\use:n
  {
    \ExplSyntaxOff
    \int_set_eq:NN \tex_endlinechar:D \c_minus_one
    \ior_str_map_inline:nn { \c_job_name_tl }
      {
        \tl_gset_eq:NN \g_strange_last_line_tl \g_strange_line_tl
        \int_gset_eq:NN \g_strange_last_indentation_int \g_strange_indentation_int
        \bool_if:NTF \g_strange_tmp_bool
          {
            \tl_gset_rescan:Nnn \g_strange_line_tl { } {#1~}
            \int_gset:Nn \g_strange_indentation_int
              { \str_length:n {#1} - \exp_args:No \str_length:n { \use:n #1 } }
            \int_compare:nTF { \g_strange_indentation_int > \g_strange_last_indentation_int }
              {
                \seq_gpush:Nx \g_strange_indentation_seq
                  { \int_use:N \g_strange_last_indentation_int }
                \seq_gpush:Nx \g_strange_environments_seq
                  { \str_substr:Nnn \g_strange_last_line_tl {1} {-1} }
                \seq_get:NN \g_strange_environments_seq \l_strange_tmpa_tl
                \exp_args:No \begin { \l_strange_tmpa_tl }
              }
              {
                \tl_use:N \g_strange_last_line_tl
                \int_while_do:nn
                  { \g_strange_indentation_int < \g_strange_last_indentation_int }
                  {
                    \seq_gpop:NN \g_strange_indentation_seq \l_strange_tmpa_tl
                      \int_gset:Nn \g_strange_last_indentation_int \l_strange_tmpa_tl
                    \seq_gpop:NN \g_strange_environments_seq \l_strange_tmpa_tl
                      \exp_args:No \end { \l_strange_tmpa_tl }
                  }
              }
          }
          {
            \str_if_eq:xxT { #1 } { \token_to_str:N \endinput }
              { \bool_gset_true:N \g_strange_tmp_bool }
          }
      }
  }
\endinput

\documentclass{article}
\usepackage{amsmath}
\document
    We have
    \equation
        A =
        \bmatrix
            a & b \\
            c & d
    where ...
%

The trailing % at the same indentation level as \document is very important. I decided to keep \\ to separate lines of the matrix, because supporting that would require changing the definition of \bmatrix. Here, you simply need to include the code before the document starts, and then stop worrying about it.

Answer 2

Is there an alternative interpreter for LaTeX that has nicer syntax?

In general, TeX is very poor when it comes to "alternative interpreters" due to the fact that writing a program that can understand the grammar of TeX is a very, very hairy problem (and turing-complete too).

Rather than looking for an "alternate implementation" of TeX it's self, try looking for an "alternate grammar" that is close to your ideal input that can be compiled to TeX. Pandoc has been mentioned as a tool that can parse light-weight markup languages such as Mardown and reStructured Text and convert them to LaTeX. Pandoc is also written in Haskell and pretty easy to extend.

Or, you could roll your own. Perhaps something like CoffeeScript could be written for LaTeX. CoffeeScript is a language that mashes up the best parts of JavaScript, Python and Ruby syntax yet compiles to JavaScript to spare the overhead of writing a complete language implementation from scratch.

Answer 3

I'm not sure where this lies between a comment and an answer. The length pushed it towards the location of the latter.

Here is perhaps a promotion of LaTeX in lieu of your hate for its syntax that matches that of your example to a fair degree.

Whenever you define an environment in LaTeX, for example via

\newenvironment{myenv}
  {<begin myenv>}% \begin{myenv}
  {<end myenv>}% \end{myenv}

it (...LaTeX) defines two commands: \myenv and \endmyenv. To test this, you'll notice an error in your console when trying to compile the following minimal example:

\documentclass{article}
\newcommand{\mycmd}{test}% \mycmd
\newenvironment{mycmd}{hi}{there}% \begin{mycmd}...\end{mycmd}
\begin{document}
\mycmd
\end{document}

That error being

! LaTeX Error: Command \mycmd already defined.
               Or name \end... illegal, see p.192 of the manual.

See the LaTeX manual or LaTeX Companion for explanation.
Type  H   for immediate help.
 ...                                              

l.3 \newenvironment{mycmd}{hi}{there}

Even though you didn't define the an additional command by the name of \mycmd. So, using your code snippet, is it very possible to have LaTeX understand exactly what you're after if you use

We have
\equation
  A =
  \bmatrix
    a & b
    c & d
  \endbmatrix
\endequation
where ...

whatever your definition for equation and bmatrix may be (existing and redefined or not). In fact, the following minimal example compiles without problem:

\documentclass{article}
\usepackage{amsmath}%
\begin{document}
We have
\equation
  A =
  \bmatrix
    a & b \\
    c & d
  \endbmatrix
\endequation
where ...
\end{document}

As commented by egreg, there are instances where this my be "surprising" and yield output contrary to one'e expectation. However, the extent of this may not influence you without more detail.