'*pattern*': [a, b] does not work when a and b are equal

[rbonvall]

I'm creating text objects for LaTeX code. In LaTeX, inline equations can be typed \(like this\) or $like this$. I tried to create text objects for both with the following configuration:

\  'paren-math': {
\     '*pattern*': ['\\(', '\\)'],
\     'select-a': 'a\',
\     'select-i': 'i\',
\   },
\  'dollar-math': {
\     '*pattern*': ['[$]', '[$]'],
\     'select-a': 'a$',
\     'select-i': 'i$',
\   },

but only the first one worked. I also tried '*pattern*: '[$][^$]*[$]', also with no positive results.

Am I missing something? Is there another simple way to create text objects where the delimiters are the same (like the ones vim has for quoted text) without defining the selection functions?

[kana]

Thank you for the report. It is a known issue. See also #2.

\  'dollar-math': {
\     '*pattern*': ['[$]', '[$]'],
\     'select-a': 'a$',
\     'select-i': 'i$',
\   },

but only the first one worked. I also tried '*pattern*: '[$][^$]*[$]', also with no positive results.

select-a and select-i imply that a pair of regular expressions is given to *pattern*, while select implies that a single regular expression is given to *pattern*.
So, if you want to avoid using functions, try the following definitions instead:

\   'dollar-math-a': {
\     '*pattern*': '\$[^$]*\$',
\     'select': 'a$',
\   },
\   'dollar-math-i': {
\     '*pattern*': '\$\zs[^$]*\ze\$',
\     'select': 'i$',
\   },

But it's better to support such usage. The function to select a region for '*pattern*': [begin, end] is textobj#user#select_pair. Do you have any idea to improve it?

[rbonvall]

Hi @kana,
thanks for providing a workaround. I'll take a look at the source code to see if I can fix the select_part funcion.

[kana]

vim-textobj-user internally uses searchpairpos() to find a region surrounded by a pair of patterns. It is a special kind of search function, because it takes care of nested structures. And it always fails if given patterns are equivalent. I looked into its implementation, and I guess that it is an intentional behavior to avoid getting stuck.

So that it seems to be necessary to implement a searchpairpos()-like function which also works fine with equivalent patterns.

[kana]

Perhaps it would be better to change the whole algorithm used in textobj#user#select_pair() instead of sticking to searchpairpos(). Anyway, I'll look into the problem more.