Fast word count function in Vim - vim

Question

I am trying to display a live word count in the vim statusline. I do this by setting my status line in my .vimrc and inserting a function into it. The idea of this function is to return the number of words in the current buffer. This number is then displayed on the status line. This should work nicely as the statusline is updated at just about every possible opportunity so the count will always remain 'live'.
The problem is that the function I have currently defined is slow and so vim is obviously sluggish when it is used for all but the smallest files; due to this function being executed so frequently.
In summary, does anyone have a clever trick for producing a function that is blazingly fast at calculating the number of words in the current buffer and returning the result?

Answer 1

I really like Michael Dunn's answer above but I found that when I was editing it was causing me to be unable to access the last column. So I have a minor change for the function:
function! WordCount()
let s:old_status = v:statusmsg
let position = getpos(".")
exe ":silent normal g\<c-g>"
let stat = v:statusmsg
let s:word_count = 0
if stat != '--No lines in buffer--'
let s:word_count = str2nr(split(v:statusmsg)[11])
let v:statusmsg = s:old_status
end
call setpos('.', position)
return s:word_count
endfunction
I've included it in my status line without any issues:
:set statusline=wc:%{WordCount()}

Answer 2

Here's a usable version of Rodrigo Queiro's idea. It doesn't change the status bar, and it restores the statusmsg variable.
function WordCount()
let s:old_status = v:statusmsg
exe "silent normal g\<c-g>"
let s:word_count = str2nr(split(v:statusmsg)[11])
let v:statusmsg = s:old_status
return s:word_count
endfunction
This seems to be fast enough to include directly in the status line, e.g.:
:set statusline=wc:%{WordCount()}

Answer 3

Keep a count for the current line and a separate count for the rest of the buffer. As you type (or delete) words on the current line, update only that count, but display the sum of the current line count and the rest of the buffer count.
When you change lines, add the current line count to the buffer count, count the words in the current line and a) set the current line count and b) subtract it from the buffer count.
It would also be wise to recount the buffer periodically (note that you don't have to count the whole buffer at once, since you know where editing is occurring).

Answer 4

This will recalculate the number of words whenever you stop typing for a while (specifically, updatetime ms).
let g:word_count="<unknown>"
fun! WordCount()
return g:word_count
endfun
fun! UpdateWordCount()
let s = system("wc -w ".expand("%p"))
let parts = split(s, ' ')
if len(parts) > 1
let g:word_count = parts[0]
endif
endfun
augroup WordCounter
au! CursorHold * call UpdateWordCount()
au! CursorHoldI * call UpdateWordCount()
augroup END
" how eager are you? (default is 4000 ms)
set updatetime=500
" modify as you please...
set statusline=%{WordCount()}\ words
Enjoy!

Answer 5

So I've written:
func CountWords()
exe "normal g\"
let words = substitute(v:statusmsg, "^.*Word [^ ]* of ", "", "")
let words = substitute(words, ";.*", "", "")
return words
endfunc
But it prints out info to the statusbar, so I don't think it will be suitable for your use-case. It's very fast, though!

Answer 6

Since vim version 7.4.1042
Since vim version 7.4.1042, one can simply alter the statusline as follows:
set statusline+=%{wordcount().words}\ words
set laststatus=2 " enables the statusline.
Word count in vim-airline
Word count is provided standard by vim-airline for a number of file types, being at the time of writing:
asciidoc, help, mail, markdown, org, rst, tex ,text
If word count is not shown in the vim-airline, more often this is due to an unrecognised file type. For example, at least for now, the compound file type markdown.pandoc is not being recognised by vim-airline for word count. This can easily be remedied by amending the .vimrc as follows:
let g:airline#extensions#wordcount#filetypes = '\vasciidoc|help|mail|markdown|markdown.pandoc|org|rst|tex|text'
set laststatus=2 " enables vim-airline.
The \v statement overrides the default g:airline#extensions#wordcount#filetypes variable. The last line ensures vim-airline is enabled.
In case of doubt, the &filetype of an opened file is returned upon issuing the following command:
:echo &filetype
Here is a meta-example:

Answer 7

I used a slightly different approach for this. Rather than make sure the word count function is especially fast, I only call it when the cursor stops moving. These commands will do it:
:au CursorHold * exe "normal g\<c-g>"
:au CursorHoldI * exe "normal g\<c-g>"
Perhaps not quite what the questioner wanted, but much simpler than some of the answers here, and good enough for my use-case (glance down to see word count after typing a sentence or two).
Setting updatetime to a smaller value also helps here:
set updatetime=300
There isn't a huge overhead polling for the word count because CursorHold and CursorHoldI only fire once when the cursor stops moving, not every updatetime ms.

Answer 8

Here is a refinement of Abslom Daak's answer that also works in visual mode.
function! WordCount()
let s:old_status = v:statusmsg
let position = getpos(".")
exe ":silent normal g\<c-g>"
let stat = v:statusmsg
let s:word_count = 0
if stat != '--No lines in buffer--'
if stat =~ "^Selected"
let s:word_count = str2nr(split(v:statusmsg)[5])
else
let s:word_count = str2nr(split(v:statusmsg)[11])
end
let v:statusmsg = s:old_status
end
call setpos('.', position)
return s:word_count
endfunction
Included in the status line as before. Here is a right-aligned status line:
set statusline=%=%{WordCount()}\ words\

Answer 9

I took the bulk of this from the vim help pages on writing functions.
function! WordCount()
let lnum = 1
let n = 0
while lnum <= line('$')
let n = n + len(split(getline(lnum)))
let lnum = lnum + 1
endwhile
return n
endfunction
Of course, like the others, you'll need to:
:set statusline=wc:%{WordCount()}
I'm sure this can be cleaned up by somebody to make it more vimmy (s:n instead of just n?), but I believe the basic functionality is there.
Edit:
Looking at this again, I really like Mikael Jansson's solution. I don't like shelling out to wc (not portable and perhaps slow). If we replace his UpdateWordCount function with the code I have above (renaming my function to UpdateWordCount), then I think we have a better solution.

Answer 10

My suggestion:
function! UpdateWordCount()
let b:word_count = eval(join(map(getline("1", "$"), "len(split(v:val, '\\s\\+'))"), "+"))
endfunction
augroup UpdateWordCount
au!
autocmd BufRead,BufNewFile,BufEnter,CursorHold,CursorHoldI,InsertEnter,InsertLeave * call UpdateWordCount()
augroup END
let &statusline='wc:%{get(b:, "word_count", 0)}'
I'm not sure how this compares in speed to some of the other solutions, but it's certainly a lot simpler than most.

Answer 11

I'm new to Vim scripting, but I might suggest
function WordCount()
redir => l:status
exe "silent normal g\<c-g>"
redir END
return str2nr(split(l:status)[11])
endfunction
as being a bit cleaner since it does not overwrite the existing status line.
My reason for posting is to point out that this function has a puzzling bug: namely, it breaks the append command. Hitting A should drop you into insert mode with the cursor positioned to the right of the final character on the line. However, with this custom status bar enabled it will put you to the left of the final character.
Anyone have any idea what causes this?

Answer 12

This is an improvement on Michael Dunn's version, caching the word count so even less processing is needed.
function! WC()
if &modified || !exists("b:wordcount")
let l:old_status = v:statusmsg
execute "silent normal g\<c-g>"
let b:wordcount = str2nr(split(v:statusmsg)[11])
let v:statusmsg = l:old_status
return b:wordcount
else
return b:wordcount
endif
endfunction

Answer 13

Since vim now supports this natively:
:echo wordcount().words

Answer 14

Using the method in the answer provided by Steve Moyer I was able to produce the following solution. It is a rather inelegant hack I'm afraid and I feel that there must be a neater solution, but it works, and is much faster than simply counting all of the words in a buffer every time the status line is updated. I should note also that this solution is platform independent and does not assume a system has 'wc' or something similar.
My solution does not periodically update the buffer, but the answer provided by Mikael Jansson would be able to provide this functionality. I have not, as of yet, found an instance where my solution becomes out of sync. However I have only tested this briefly as an accurate live word count is not essential to my needs. The pattern I use for matching words is also simple and is intended for simple text documents. If anyone has a better idea for a pattern or any other suggestions please feel free to post an answer or edit this post.
My solution:
"returns the count of how many words are in the entire file excluding the current line
"updates the buffer variable Global_Word_Count to reflect this
fu! OtherLineWordCount()
let data = []
"get lines above and below current line unless current line is first or last
if line(".") > 1
let data = getline(1, line(".")-1)
endif
if line(".") < line("$")
let data = data + getline(line(".")+1, "$")
endif
let count_words = 0
let pattern = "\\<\\(\\w\\|-\\|'\\)\\+\\>"
for str in data
let count_words = count_words + NumPatternsInString(str, pattern)
endfor
let b:Global_Word_Count = count_words
return count_words
endf
"returns the word count for the current line
"updates the buffer variable Current_Line_Number
"updates the buffer variable Current_Line_Word_Count
fu! CurrentLineWordCount()
if b:Current_Line_Number != line(".") "if the line number has changed then add old count
let b:Global_Word_Count = b:Global_Word_Count + b:Current_Line_Word_Count
endif
"calculate number of words on current line
let line = getline(".")
let pattern = "\\<\\(\\w\\|-\\|'\\)\\+\\>"
let count_words = NumPatternsInString(line, pattern)
let b:Current_Line_Word_Count = count_words "update buffer variable with current line count
if b:Current_Line_Number != line(".") "if the line number has changed then subtract current line count
let b:Global_Word_Count = b:Global_Word_Count - b:Current_Line_Word_Count
endif
let b:Current_Line_Number = line(".") "update buffer variable with current line number
return count_words
endf
"returns the word count for the entire file using variables defined in other procedures
"this is the function that is called repeatedly and controls the other word
"count functions.
fu! WordCount()
if exists("b:Global_Word_Count") == 0
let b:Global_Word_Count = 0
let b:Current_Line_Word_Count = 0
let b:Current_Line_Number = line(".")
call OtherLineWordCount()
endif
call CurrentLineWordCount()
return b:Global_Word_Count + b:Current_Line_Word_Count
endf
"returns the number of patterns found in a string
fu! NumPatternsInString(str, pat)
let i = 0
let num = -1
while i != -1
let num = num + 1
let i = matchend(a:str, a:pat, i)
endwhile
return num
endf
This is then added to the status line by:
:set statusline=wc:%{WordCount()}
I hope this helps anyone looking for a live word count in Vim. Albeit one that isn't always exact. Alternatively of course g ctrl-g will provide you with Vim's word count!

Answer 15

In case someone else is coming here from Google, I modified Abslom Daak's answer to work with Airline. I saved the following as
~/.vim/bundle/vim-airline/autoload/airline/extensions/pandoc.vim
and added
call airline#extensions#pandoc#init(s:ext)
to extensions.vim
let s:spc = g:airline_symbols.space
function! airline#extensions#pandoc#word_count()
if mode() == "s"
return 0
else
let s:old_status = v:statusmsg
let position = getpos(".")
let s:word_count = 0
exe ":silent normal g\<c-g>"
let stat = v:statusmsg
let s:word_count = 0
if stat != '--No lines in buffer--'
let s:word_count = str2nr(split(v:statusmsg)[11])
let v:statusmsg = s:old_status
end
call setpos('.', position)
return s:word_count
end
endfunction
function! airline#extensions#pandoc#apply(...)
if &ft == "pandoc"
let w:airline_section_x = "%{airline#extensions#pandoc#word_count()} Words"
endif
endfunction
function! airline#extensions#pandoc#init(ext)
call a:ext.add_statusline_func('airline#extensions#pandoc#apply')
endfunction

Answer 16

A variation of Guy Gur-Ari's refinement that
only counts words if spell checking is enabled,
counts the number of selected words in visual mode
keeps mute outside of insert and normal mode, and
hopefully is more agnostic to the system language (when different from english)
function! StatuslineWordCount()
if !&l:spell
return ''
endif
if empty(getline(line('$')))
return ''
endif
let mode = mode()
if !(mode ==# 'v' || mode ==# 'V' || mode ==# "\<c-v>" || mode =~# '[ni]')
return ''
endif
let s:old_status = v:statusmsg
let position = getpos('.')
let stat = v:statusmsg
let s:word_count = 0
exe ":silent normal g\<c-g>"
try
if mode ==# 'v' || mode ==# 'V'
let s:word_count = split(split(v:statusmsg, ';')[1])[0]
elseif mode ==# "\<c-v>"
let s:word_count = split(split(v:statusmsg, ';')[2])[0]
elseif mode =~# '[ni]'
let s:word_count = split(split(v:statusmsg, ';')[2])[3]
end
" index out of range
catch /^Vim\%((\a\+)\)\=:E\%(684\|116\)/
return ''
endtry
let v:statusmsg = s:old_status
call setpos('.', position)
return "\ \|\ " . s:word_count . 'w'
endfunction
that can be appended to the statusline by, say,
set statusline+=%.10{StatuslineWordCount()} " wordcount

Answer 17

Building upon https://stackoverflow.com/a/60310471/11001018, my suggestion is:
"new in vim 7.4.1042
let g:word_count=wordcount().words
function WordCount()
if has_key(wordcount(),'visual_words')
let g:word_count=wordcount().visual_words."/".wordcount().words
else
let g:word_count=wordcount().cursor_words."/".wordcount().words
endif
return g:word_count
endfunction
And then:
set statusline+=\ w:%{WordCount()},