[mini.ai] imitate word textobject another way

[aidancz]

Contributing guidelines

• I have read CODE_OF_CONDUCT.md

Module(s)

mini.ai

Question

There is an example that imitates the word textobject with optional trailing whitespace:

require("mini.ai").setup({
	custom_textobjects = {
		w = { "()()%f[%w]%w+()[ \t]*()" },
	},
})

This works well, but I want to achieve the opposite behavior: imitate the word textobject with optional leading whitespace. However, this does not work:

require("mini.ai").setup({
	custom_textobjects = {
		w = { "()[ \t]*()%f[%w]%w+()()" },
	},
})

This is very close to the behavior I want:

require("mini.ai").setup({
	custom_textobjects = {
		w = { "()%f[ \t][ \t]*()%f[%w]%w+()()" },
	},
})

but it does not work correctly at the first word, i.e.

for text:
**the quick brown fox jumps over the lazy dog**

now:
**the( quick)( brown)( fox)( jumps)( over)( the)( lazy)( dog)**

i want:
**(the)( quick)( brown)( fox)( jumps)( over)( the)( lazy)( dog)**

Is this possible?

[echasnovski]

Yes, it is possible by using composed pattern. It is a bit convoluted, but it is essentially a way to say "match one of patterns at this place, whichever has the smallest width". Here is the example for this particular case:

require('mini.ai').setup({
  custom_textobjects = {
    w = { { '^()%w+()', '()%f[ \t][ \t]*()%f[%w]%w+()()' } },
  },
})

However, my honest suggestion would be to try to avoid this. Although it makes it work with next / prev search methods, it makes it only work with Latin characters (as this is what %w is used for) and not respect 'iskeyword' option.

[aidancz]

Thanks for the reply
However, this cannot select the first the because there is an asterisk before it, it is not at the start of the line.

I previously thought about using a composed pattern, like this:

require("mini.ai").setup({
	custom_textobjects = {
		w = { { "()%f[ \t][ \t]*()%f[%w]%w+()()", "%f[%w]%w+" } },
	},
})

or

require("mini.ai").setup({
	custom_textobjects = {
		w = { { "%f[%w]%w+", "()%f[ \t][ \t]*()%f[%w]%w+()()" } },
	},
})

The result is that their order doesn’t matter; as you said, the pattern with the smallest width wins.

[echasnovski]

Ah, sorry, I misunderstood the meaning of **. Didn't realize that they were part of the text.

I am afraid, in this case it is either add patterns on a case-by-case basis or pivot from using Lua patterns in favor of custom function. My suggestion would be the first one, but the following one makes it work for the original case (but there are still cases for which it won't work and even through an error):

require('mini.ai').setup({
  custom_textobjects = {
    w = {
      {
        '^()%w+()',
        '^[^ \t%w]+()()%w+()()',
        '()%f[ \t][ \t]*()%f[%w]%w+()()',
      },
    },
  },
})

[aidancz]

Oh! Wait, I think I just thought of a method 😄

require("mini.ai").setup({
	custom_textobjects = {
		w = { { "%f[%w \t]%w+", "()%f[ \t][ \t]*()%f[%w]%w+()()" } },
	},
})

finally works

[aidancz]

Thank you!

[echasnovski]

Yep, seems like it indeed works, at least for more cases than other suggestions! Very nice :)