Re: tex-validate-region reports false errors on some LaTeX commands

Re: tex-validate-region reports false errors on some LaTeX commands

[martin rudalics]

 > The easy fix is to simply ignore open parentheses (also open square
 > brackets) that are preceded by a backslash.  If things are fixed this
 > way, then mismatched uses of \(, \), \[, and \] will not be diagnosed,
 > but that is better than reporting zillions of false errors.

We could give the backslash before paren style characters some
syntax-table property (punctuation?).  We'd match "\(" with ")" then.
Would that harm much?

Re: tex-validate-region reports false errors on some LaTeX commands

[Joe Wells]

martin rudalics <rudalics@gmx.at> writes:

>> The easy fix is to simply ignore open parentheses (also open square
>> brackets) that are preceded by a backslash.  If things are fixed this
>> way, then mismatched uses of \(, \), \[, and \] will not be diagnosed,
>> but that is better than reporting zillions of false errors.
>
> We could give the backslash before paren style characters some
> syntax-table property (punctuation?).  We'd match "\(" with ")" then.
> Would that harm much?

The correct solution would require that \(, \), \[, and \] are treated
the same as \begin{math}, \end{math}, \begin{displaymath}, and
\end{displaymath}.  (The latter are defined in terms of the former in
latex.ltx.)  Emacs already parses \begin{XYZ} and \end{XYZ} for use in
forward-sexp.

Anyway, the code in tex-validate-region seems to operate in two
passes.  The first pass verifies that the syntax parsing underlying
forward-sexp is happy.  The second pass seems to be oriented at making
ordinary parentheses/brackets/braces match.  I'm not entirely sure why
two passes should be needed.  Can anyone explain?

Note that strange things happen with syntax parsing in LaTeX.
Consider this text:

  (\begin{quote})
  \end{quote}

If you put point to the left of the “(”, then forward-sexp moves to
the right of the “)”.  If you put point before the \begin{quote}, then
forward-sexp moves to the right of the \end{quote}.  This all happens
despite the fact that the structures are not balanced properly.  Can
anyone explain this bizarreness?

-- 
Joe

Re: tex-validate-region reports false errors on some LaTeX commands

[Glenn Morris]

Joe Wells wrote:

> The correct solution would require that \(, \), \[, and \] are treated
> the same as \begin{math}, \end{math}, \begin{displaymath}, and
> \end{displaymath}.

I've installed a hack fix along these lines in the trunk. I don't want
to put it in the release branch, since I don't think the
tex-validate-* functions are very important, and the sexp motion
commands seem to be used quite a bit in tex-mode.

>   (\begin{quote})
>   \end{quote}
>
> If you put point to the left of the “(”, then forward-sexp moves to
> the right of the “)”.  If you put point before the \begin{quote}, then
> forward-sexp moves to the right of the \end{quote}.  This all happens
> despite the fact that the structures are not balanced properly.

The latex-forward-sexp function is not rigorous.

Re: tex-validate-region reports false errors on some LaTeX commands

[Joe Wells]

Glenn Morris <rgm@gnu.org> writes:

> Joe Wells wrote:
>
>> The correct solution would require that \(, \), \[, and \] are
>> treated the same as \begin{math}, \end{math}, \begin{displaymath},
>> and \end{displaymath}.
>
> I've installed a hack fix along these lines in the trunk. I don't
> want to put it in the release branch, since I don't think the
> tex-validate-* functions are very important, and the sexp motion
> commands seem to be used quite a bit in tex-mode.

Okay.  I don't understand the distinction between trunk and release
branch here, but I assume this must be sensible.

>>   (\begin{quote})
>>   \end{quote}
>>
>> If you put point to the left of the “(”, then forward-sexp moves to
>> the right of the “)”.  If you put point before the \begin{quote},
>> then forward-sexp moves to the right of the \end{quote}.  This all
>> happens despite the fact that the structures are not balanced
>> properly.
>
> The latex-forward-sexp function is not rigorous.

Indeed, it is a fairly gross kludge.

It would be nice if font-locking could somehow set up syntax-table
text properties to make this work correctly.  However, I think the
syntax-table text property feature is not sophisticated enough for
handling LaTeX.  You can only assign syntax-table text properties of
the form (SYNTAX-CODE . MATCHING-CHAR).  It is probably necessary for
LaTeX to be able to do something of the form (SYNTAX-CODE
. MATCHING-STRING).  For example, “\begin{xyz}” would need a
syntax-table text property that says it is matched by “\end{xyz}”.
Right now you can only specify single matching characters.

-- 
Joe