Small improvements of completion functions #1203
Conversation
Prevent `haskell-utils-repl-response-error-status` and `haskell-process-get-repl-completions` from trying to split nil-ish responses, because this yields an error. Rather return immediately `'no-error` in response parser and `nil` in completion function.
Prevent `haskell-completions-completion-at-point` from tries of getting completion candidates for nil prefixes. Likely this is right way for this function, moreover passing nil prefix to `haskell-completions--simple-completions` yields error.
geraldus
changed the title
|
Is there a way to trigger this |
|
@gracjan didn't understood your question well. I've faced the situation, when GHCi fails to start:
In this case REPL process is alive, and not marked as restarting and all requests to REPL returns |
Stop static completions to work inside comments and strings (except pragma comments). Let other completion function handle this cases, also keyword candidates are not so useful in comments and string.
|
Oops. I've found another one bug, force-pushed new changes. TLDR: first fixes for static completion function added condition case for general type of prefix with nil value associated as result. This was wrong, because instead of just |
|
+1 for these fixes: |
|
@purcell by the way, do you have some special setup for company mode and haskell? |
|
Not really -- I mostly rely on |
|
@purcell I have found useful to override (set (make-local-variable 'company-backends)
(append '((company-capf company-dabbrev-code) company-dabbrev)
company-backends))This combines completions from haskell mode with dabbrev-code completions outside strings and comments (useful for local identifiers defined in where and let blocks) and provides dabbrev completions otherwise. I want to put it in docs. |
Small improvements of completion functions
|
@geraldus Good tip, thanks. |
As a user of company mode I see errors in mini-buffer very often:
Small investigation lead me to proposed changes