Math formatted columns in tabular environment handled incorrectly

[shardulc]

This issue is about a very specific use case but I think it is common enough to make an issue. The screenshot below demonstrates the problem.

That is, when specifying columns to be in math mode using the tabular environment of the array package, you have to write >{\(}c<{\)} but the syntax highlighter does not recognize the \) that ends math mode. Perhaps this occurs because the { starts a new scope before \) but I do not know how the highlighter works. Can someone please take a look at it?

[Aerijo]

You're right, the issue is caused by the curly brackets containing the end mathmode delimiter. As curly brackets are scoped to their own rules, their contents are effectivlely hidden from any outside rules. For example, look at the following (broken) latex:

\( this is math {\)} this is still math \) this is not math

Pasting that into atom will display the same issue as yours.

Unfortunately, it is difficult to avoid this behaviour. A possible general solution would be to stop scoping individual appearances of{ and }, but that could be annoying for other reasons. In your case, the begin{tabular} grammar could be extended to look for an opening curly brace immediately following the begin statement. If it finds one, it will scope contents as verbatim until the next (balanced) closing curly bracket. This would not fix the issue I provided, but it would fix yours.

@yudai-nkt I don't mind the second option. It's not the best solution, and could lead to invalid highlighting for other reasons, but I think mathmode tables are common enough to justify it.

[Aerijo]

Alternatively, mathmode could be optionally ended by an (unbalanced) }. As far as I can tell, it is never valid to end mathmode in a different group to the one it begins in. For example, the following is valid (using \text fromamsmath; $ can be replaced by \( and \)):

text $ math $
text $ math { still math } $
text $ math { still math \text{text $ deeper math $ }} $

but the following will break

text $ math {$}
text $ math {$ deeper math $} $
text {$} math {$}

Therefore, I believe that ending mathmode with } would be fine, as if the user has not already balanced it, they either are doing something like for the table, or are going to get an error anyway. As we cannot reliably predict which is the case, I would say letting the user run into the error is preferable to ruining syntax highlighting for the rest of the document.

However, this approach hinges on being able to reliably balance { and }. Otherwise, highlighting would break on some of the valid examples. As it is, I believe the current scoping will hide matching {} pairs from the math pattern, but would need to test to confirm.

And whatever fix is implemented, this is the kind of thing that needs ... shudders in apprehension ... specs.

[Aerijo]

End result would look something like this:

This was done by just adding |(?=\\}) to the end pattern of the mathmodes.

(Interesting to note: if the end math command occurs by itself, it will only be scoped as a generic control symbol)

[shardulc]

@Aerijo Isn't there any way to simply ignore scopes when trying to start and end math mode, just highlighting everything between \( and \) regardless of what comes in between? While typesetting math, it is anyway possible that some notation requires an unmatched {/}, so this is bound to come up again.

[Aerijo]

It's true that math can need unbalanced {/}, but these would need to be escaped (like \{), and that is already handled properly. I would not like it ignore everything within math mode, as I like seeing \frac highlighted differently to the product frac.

Plus, it would still look wrong in your example, as the }c<{ would appear as math (which I don't think it is).

[Aerijo]

@yudai-nkt What do you think about this? I've yet to find a problem with it, and it's quite easy to implement. It serves as a sort of 'failsafe', such that even if the construction is valid (like in this issue), the error does not propagate across the file.