Racket `omni_start_map` should be something like `\k\+` #3870

benknoble · 2021-08-10T18:28:12Z

Racket, being a lisp, has far more than just alpha-numeric identifiers, which is what the default pattern matches. This means when I type file-name, invoking completion won't give file-name-from-path but completions like namespace—it has used only the name part. Using \k\+ seemed to work in cases like this, but I don't have an exhaustive test.

More broadly, and the reason this is an issue and not a PR, is I think the default should be \k\+—then it should be effectively scoped to the languages iskeyword setting, no?

I also don't understand why $ is included in the omni_start_map pattern: does that only allow completions when there is nothing else on the line? Perhaps word boundaries \<\> would be more appropriate?

Looking for feedback here.

PS Why not use the lisp pattern? Well, I don't think even it covers all possible lisp or racket identifiers, and \k\+ is so much easier to read.

The text was updated successfully, but these errors were encountered:

w0rp · 2021-11-11T11:13:31Z

I don't know what any of this is, but if you can create a pull request that fixes this without breaking any current functionality in languages like TypeScript, go for it.

benknoble · 2021-11-12T13:50:09Z

I am referring to s:omni_start_map in autoload/ale/completion.vim (apologies for not being clear). The declaration is around line 139, with sole use at line 311.

It seems like the intent was to allow filetypes to specify patterns to find the start of a word for completion, with a default of

\v[a-zA-Z$_][a-zA-Z$_0-9]*$

Unfortunately, this doesn't work for some languages. It explicitly disallows - in words that should start completion, for example; every lisp I've ever used has - in names commonly (some examples are in the original post). This means that trying to complete from a word that already has - in it breaks.

My suggestion is to use \k\+ (possibly with word boundaries)—then the default should work for all filetypes by leveraging their custom iskeyword settings (which define in vim what a word is for a filetype). If a filetype has this correctly set (and, e.g., racket, scheme, and common lisp should), then it should "just work."

If that's no good, then I'd like to at least add 'racket': '\k\+' or similar—the issue is that racket is actually a lot of languages, so there are lots of possible filetypes… and other lisps suffer, too, so it would be nice to add them… etc. So then you have this mechanism that controls completion (and breaks it, if not set right) but that is really hard to customize. (By really hard, I mean that the only way is to (a) edit your local copy of ALE and be careful not to lose that when updating or (b) submit a PR and wait for it to merge. That seems untenable in the long-term.)

The default `omni_start_map` is to restrictive for Lisps and Schemes like Racket, which permit hyphens (among other special characters). As recorded in dense-analysis#3870, trying to complete `file-name-from-path` when typing `file-name<C-x><C-o>` would give completions like `namespace` because the hyphen is ignored to find the start of the word for completion. Now the racket filetype searches for the start using the keyword class `\k`, which is more precise.

The default `omni_start_map` is too restrictive for Lisps and Schemes like Racket, which permit hyphens (among other special characters). As recorded in dense-analysis#3870, trying to complete `file-name-from-path` when typing `file-name<C-x><C-o>` would give completions like `namespace` because the hyphen is ignored to find the start of the word for completion. Now the default searches for the start using the keyword class `\k`, which is more precise and configurable for each filetype without modifying the source.

The default `omni_start_map` is too restrictive for Lisps and Schemes like Racket, which permit hyphens (among other special characters). As recorded in dense-analysis#3870, trying to complete `file-name-from-path` when typing `file-name<C-x><C-o>` would give completions like `namespace` because the hyphen is ignored to find the start of the word for completion. Now the racket filetype searches for the start using the keyword class `\k`, which is more precise.

benknoble · 2022-05-05T20:26:06Z

I've opened 2 different pull requests:

One only fixes Racket, which I feel is unsustainable.
The other modifies the default to be more precise per-filetype, but I don't know if this will break anything else.

benknoble · 2022-06-22T01:31:57Z

Reviving this, both PR #4186 and #4187 pass tests. It's not clear to me that passing the tests means "without breaking any current functionality in languages like TypeScript," but I thought I would mention it.

The default `omni_start_map` is too restrictive for Lisps and Schemes like Racket, which permit hyphens (among other special characters). As recorded in #3870, trying to complete `file-name-from-path` when typing `file-name<C-x><C-o>` would give completions like `namespace` because the hyphen is ignored to find the start of the word for completion. Now the racket filetype searches for the start using the keyword class `\k`, which is more precise.

The default `omni_start_map` is too restrictive for Lisps and Schemes like Racket, which permit hyphens (among other special characters). As recorded in dense-analysis#3870, trying to complete `file-name-from-path` when typing `file-name<C-x><C-o>` would give completions like `namespace` because the hyphen is ignored to find the start of the word for completion. Now the racket filetype searches for the start using the keyword class `\k`, which is more precise.

Consider a file like ``` #lang racket (require racket/gui) ``` Type `Go(eventspace-`. Pressing <C-x><C-o> to trigger omnicomplete should suggest ``` eventspace-handler-thread eventspace-shutdown? eventspace-event-evt ``` It does not (instead producing "top-level" completions, as if `(eventspace-` wasn't even there). Debugging, `ale#completion#OmniFunc(1, '')` correctly returns `1`, but when given `(0, 'eventspace-')` it returns either the empty list or generic completion results as described above. I'm not entirely sure of the mechanism, but it seems that b:ale_completion_info.prefix is the key, and that this is set by `ale#completion#GetPrefix`. Calling `ale#completion#GetPrefix('racket', line('.'), col('.'))` with the cursor on a space _after_ `eventspace-` returned `''`! Now, it returns `eventspace-` and the completions work correctly again. Ref dense-analysis#4293, dense-analysis#4186, dense-analysis#3870

Consider a file like ``` #lang racket (require racket/gui) ``` Type `Go(eventspace-`. Pressing <C-x><C-o> to trigger omnicomplete should suggest ``` eventspace-handler-thread eventspace-shutdown? eventspace-event-evt ``` It does not (instead producing "top-level" completions, as if `(eventspace-` wasn't even there). Debugging, place the cursor on a space _after_. Now `ale#completion#OmniFunc(1, '')` correctly returns `1`, but when given `(0, 'eventspace-')` it returns either the empty list or generic completion results as described above. I'm not entirely sure of the mechanism, but it seems that `b:ale_completion_info.prefix` is the key, and that this is set by `ale#completion#GetPrefix`. Calling `ale#completion#GetPrefix('racket', line('.'), col('.'))` returned `''`! Now, it returns `eventspace-` and the completions work correctly again. Ref dense-analysis#4293, dense-analysis#4186, dense-analysis#3870

Consider a file like ``` #lang racket (require racket/gui) ``` Type `Go(eventspace-`. Pressing <C-x><C-o> to trigger omnicomplete should suggest ``` eventspace-handler-thread eventspace-shutdown? eventspace-event-evt ``` It does not (instead producing "top-level" completions, as if `(eventspace-` wasn't even there). Debugging, place the cursor on a space _after_. Now `ale#completion#OmniFunc(1, '')` correctly returns `1`, but when given `(0, 'eventspace-')` it returns either the empty list or generic completion results as described above. I'm not entirely sure of the mechanism, but it seems that `b:ale_completion_info.prefix` is the key, and that this is set by `ale#completion#GetPrefix`. Calling `ale#completion#GetPrefix('racket', line('.'), col('.'))` returned `''`! Now, it returns `eventspace-` and the completions work correctly again. Ref #4293, #4186, #3870

Consider a file like ``` #lang racket (require racket/gui) ``` Type `Go(eventspace-`. Pressing <C-x><C-o> to trigger omnicomplete should suggest ``` eventspace-handler-thread eventspace-shutdown? eventspace-event-evt ``` It does not (instead producing "top-level" completions, as if `(eventspace-` wasn't even there). Debugging, place the cursor on a space _after_. Now `ale#completion#OmniFunc(1, '')` correctly returns `1`, but when given `(0, 'eventspace-')` it returns either the empty list or generic completion results as described above. I'm not entirely sure of the mechanism, but it seems that `b:ale_completion_info.prefix` is the key, and that this is set by `ale#completion#GetPrefix`. Calling `ale#completion#GetPrefix('racket', line('.'), col('.'))` returned `''`! Now, it returns `eventspace-` and the completions work correctly again. Ref dense-analysis#4293, dense-analysis#4186, dense-analysis#3870

benknoble added the enhancement label Aug 10, 2021

This was referenced May 5, 2022

racket: start completion in a full keyword #4186

Merged

omni-compl: start completion in a full keyword #4187

Closed

hsanson closed this as completed Jun 23, 2022

benknoble mentioned this issue Nov 3, 2022

omni: racket should complete at end of keywords #4352

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Racket `omni_start_map` should be something like `\k\+` #3870

Racket `omni_start_map` should be something like `\k\+` #3870

benknoble commented Aug 10, 2021

w0rp commented Nov 11, 2021

benknoble commented Nov 12, 2021

benknoble commented May 5, 2022

benknoble commented Jun 22, 2022

Racket omni_start_map should be something like \k\+ #3870

Racket omni_start_map should be something like \k\+ #3870

Comments

benknoble commented Aug 10, 2021

w0rp commented Nov 11, 2021

benknoble commented Nov 12, 2021

benknoble commented May 5, 2022

benknoble commented Jun 22, 2022

Racket `omni_start_map` should be something like `\k\+` #3870

Racket `omni_start_map` should be something like `\k\+` #3870