deoplete-plugins/deoplete-dictionary

deoplete source for dictionary

deoplete-dictionary

deoplete source for dictionary

Description

This source collects "keyword_patterns" keywords from 'dictionary'.

Note: It uses buffer-local 'dictionary' set up.

Note: It also supports directory.

rank: 100

Configuration examples

" Sample configuration for dictionary source with multiple
" dictionary files.
setlocal dictionary+=/usr/share/dict/words
setlocal dictionary+=/usr/share/dict/american-english
" Remove this if you'd like to use fuzzy search
call deoplete#custom#source(
\ 'dictionary', 'matchers', ['matcher_head'])
" If dictionary is already sorted, no need to sort it again.
call deoplete#custom#source(
\ 'dictionary', 'sorters', [])
" Do not complete too short words
call deoplete#custom#source(
\ 'dictionary', 'min_pattern_length', 4)

deoplete-dictionary's Issues

completion with upper case

With ignorecase on, when I type Elep I correctly see elephant as an option. However, when I complete with it, elephant is used, i.e. my case is removed. Expected behaviour is completing as Elephant.

Am I missing an option, or is this something deoplete doesn't support? For now I've just quickly added it as part of my hack for #1, and it works as expected - basically, return [{'word': start + w[len(start):]'} ....

performance improvements?

I tried this out and completion for a 100k dictionary is surprisingly slow - say half a second. I quickly hacked something where I did the filtering in gather_candidates (something like return [{'word': w} for w in candidates if w.startswith(context['complete_str'])) and it's nearly instant and seems to work fine. This surprises me, as I thought matcher_head would be doing the same thing.

Does anyone

1. see why the current implementation is so much slower?
2. object to having an implementation similar to the above as a PR?