ycm completes html error

[zeayes]

In html, YCM treated dot as part of identifier.

vim -version

VIM - Vi IMproved 7.4 (2013 Aug 10, compiled Mar 10 2016 14:43:34)
Garbage after option argument: "-version"
More info with: "vim -h"

vim html keyword config:

autocmd FileType xml,css,scss,sass,html,javascript set iskeyword& | set iskeyword+=-

YCM config:

Plugin 'Valloric/YouCompleteMe'
" set completeopt=longest,menu,preview
set completeopt=longest,menu
let g:ycm_error_symbol='>>'
let g:ycm_warning_symbol='>*'
let g:ycm_key_list_select_completion=['<c-n>']
let g:ycm_key_list_previous_completion=['<c-p>']
let g:ycm_confirm_extra_conf=0
let g:ycm_complete_in_strings=1
let g:ycm_complete_in_comments=1
let g:ycm_use_ultisnips_completer = 1
let g:ycm_seed_identifiers_with_syntax=1
let g:ycm_min_num_of_chars_for_completion=2
let g:ycm_min_num_identifier_candidate_chars=0
let g:ycm_collect_identifiers_from_tags_files=1
let g:ycm_collect_identifiers_from_comments_and_strings=1
let g:ycm_autoclose_preview_window_after_insertion = 1
let g:ycm_autoclose_preview_window_after_completion = 1
" [ 'same-buffer', 'horizontal-split', 'vertical-split', 'new-tab' ]
let g:ycm_goto_buffer_command='vertical-split'
let g:ycm_global_ycm_extra_conf = '~/.ycm_extra_conf.py'
let g:ycm_rust_src_path = '/usr/local/rustc-1.6.0/src'
let g:ycm_filetype_blacklist = {'tagbar' : 1, 'nerdtree' : 1}

[vheon]

ycmd uses this regex https://github.com/Valloric/ycmd/blob/master/ycmd/identifier_utils.py#L60-L63 for html and of course is not perfect; but to me the problem here is that what you're writing is not plain html but a templated html which have a different syntax compared to html. I don't know what would be the best fix for this:

• add a regex for a templated html and handle those situations properly
• make the html regex immune to those situation

😕 guys thoughts?

[micbou]

The issue is not specific to a templated html. For example:

<div>test.test</div>

test.test will be extracted as an identifier.

We could exclude the dot character but according to HTML specs, dot is a valid character in tag names and attribute names/values even if I am not aware of any tag and attribute names containing a dot and dots in attribute values are rarely (never?) used.

[vheon]

The issue is not specific to a templated html. For example:

ops, sorry

[zeayes]

Could YCM support vim's iskeyword as identifier?

[micbou]

Could YCM support vim's iskeyword as identifier?

Probably but YCM currently uses regexes to extract identifiers which are much more powerful than Vim iskeyword so I think it is better to improve the regex.

[Valloric]

Like @micbou said, supporting iskeyword is not the answer (and it would be too complex), improving the regex to understand common template chars ({{ and }}) is.

[zeayes]

thanks!

[forresthopkinsa]

Why is YCM using regex to parse HTML? I thought we were past this haha