Some proposed improvements for t-amsl.tex.

[Brooks Moses <bmoses@stanford.edu>]

Giuseppe and ConTeXt list members -

Having figured out how to use \grabuntil (see my previous messages to this 
list), I've realized that it's really an inconvenient solution in some 
cases, because it's searching for a literal bit of text rather than an end 
to the appropriate grouping level.  (Apparently there isn't a way to do the 
latter in TeX, given that AMSmath's \collect@body also searches for literal 
text.)  This means that, if I write a "LaTeX compatibility function" that 
does things like define \end{myenvironment} to expand to 
\stopmyenvironment, my redefinition will not work on any environments using 
\grabuntil.

This inconvenience happens to show up in t-amsl.tex's "align" and "gather" 
implementations, which is one place where I really need that compatibility 
function.  However, it turns out that there's a simple way to rewrite the 
implementation so to avoid needing \grabuntil.

Mostly, this works on the fact that the definition of Plain TeX's \eqalign 
function can be split apart into a before and after part, as follows:

\def\starteqalign{\null\,\vcenter\bgroup
   \openup\jot {\mathsurround=0pt}%
   \ialign\bgroup
       \strut\hfil$\displaystyle{##}$&$\displaystyle{{}##}$\hfil
       \crcr}
\def\stopeqalign{\crcr\egroup\egroup\,}

(This, aside from the \bgroup's and \egroup's, is copied directly from 
Appendix B of The TeXbook.)  Thus, we could expand the instances of 
\eqalign in the \startalign and \startgather definitions, shuffle things 
around appropriately, and obtain the following:

\def\startalign{\startformula
    \let\\\cr
    \starteqalign}
\def\stopalign{\crcr\stopeqalign\stopformula}

\def\startgather{\startformula
    \def\\{\cr&}%
    \starteqalign&}
\def\stopgather{\crcr\stopeqalign\stopformula}

This exactly duplicates the existing functionality in the t-amsl.tex 
package.  However, that functionality could be improved to be a better 
match to what's provided in amsmath.tex; in particular, align should accept 
multiple columns of equations, and gather should center its lines.  Thus, I 
propose that the following redefinitions be included in t-amsl:

\def\startalign{\startformula
   \let\\\cr
   \null\vcenter\bgroup
   \openup\jot {\mathsurround=0pt}%
   \everycr={}\tabskip=4pt plus1fil \halign to \displaywidth\bgroup
       \strut\hfil$\displaystyle{##}$%
       \tabskip=0pt &$\displaystyle{{}##}$\hfil%
       \tabskip=4pt plus1fil &&\hfil$\displaystyle{##}$%
       \tabskip=0pt &$\displaystyle{{}##}$\hfil%
       \tabskip=4pt plus1fil \crcr}
\def\stopalign{\crcr\crcr\egroup\egroup\stopformula}

\def\startgather{\startformula
   \let\\\cr
   \null\,\vcenter\bgroup
   \openup\jot {\mathsurround=0pt}%
   \ialign\bgroup
       \strut\hfil$\displaystyle{##}$\hfil
       \crcr}
\def\stopgather{\crcr\crcr\egroup\egroup\,\stopformula}

Also, a similar method can be used to replicate the split environment 
(which, unlike the above, should go inside an explicit \startformula and 
\stopformula pair).  I propose the following definition be added as well:

\def\startsplit{%
   \let\\\cr
   \null\,\vcenter\bgroup
   \openup\jot {\mathsurround=0pt}%
   \ialign\bgroup
       \strut\hfil$\displaystyle{##}$&$\displaystyle{{}##}$\hfil
       \crcr}
\def\stopsplit{\crcr\egroup\egroup\,}

This still doesn't address the fact that the lines in the align and gather 
environments are not given individual equation numbers as they should be, 
but I think that is a project for another day.

- Brooks


P.S. A question: is it intentional that the array environment in t-amsl 
produces notably different spacing than LaTeX's array environment?