OK, it is actually somewhat different. Under normal circumstances my example would be a syntax error, but there are two exclusions that I know of:

The specific code where I observed the parse error has \begin{...} and \end{...} occur within an environment definition, and aren't interpreted by latex:

\newenvironment{scilastnote}{%
\setcounter{lastnote}{\value{enumiv}}%
\addtocounter{lastnote}{+1}%
\begin{list}%
{\arabic{lastnote}.}
{\setlength{\leftmargin}{.22in}}
{\setlength{\labelsep}{.5em}}}
{\end{list}}

The unmatched { seem to be closed after the end of document. This seems to be easier to address.

Thinking a bit further, there are other related problems with the current parsing approach, consider these examples:

\large[a]{b} will be parsed as TexCmd('large', [OArg('a'), RArg('b')]), even though it's a \large followed by text [a]b.
\textbf a will be parsed as TexCmd('textbf') followed by a, even though a is an argument to \textbf here.
\textbf[a]{b} is a boldface [ followed by a]b.
{\large text} is parsed as a single RArg without a command.

Figuring out correctly what is the signature of every command is nearly impossible and requires a working latex state machine: one would need to know everything that is defined within imported packages, and also by the user within the current file. Because the command (re)definitions may occur at any place within a document, for example here

\documentclass{article}
\begin{document}
\textbf{this is boldface and the command expects one argument}
\renewcommand{\textbf}[2]{#2}
\textbf{this text is hidden because the command expects two arguments}{second argument that I don't actually need to provide!}
\renewcommand{\textbf}[1]{#1}
\textbf{this text is visible and the command expects one argument}
\end{document}

I believe the scope of this project is to provide a programmatic way to deal with a relatively small subset of latex—this is already a very useful goal, but then it's important to keep the goal clear.

One improved approach to parsing that I can imagine would be to do the following:

Check in a user-controllable set of known commands for the number of optioinal and required argument of a given command, and parse what belongs within a single command based on that input, whether there is a correct number of {} or not.
Apply the current heuristic as a starting point, so \a[b][c]{d}{e} is a command with 2 optional and 2 required arguments, although this may be even remotely close to true, and the command may have both more and fewer arguments.
Introduce a separate expression for a token group { ... }, and differentiate it from RArg.

Not sure how to deal with the low-level commands like \newenvironment, etc.

parse error caused by extraneous figure brackets #59

Description

Metadata

Metadata

Assignees

Labels

Projects

Milestone

Relationships

Development

Issue actions