Previously vimwiki#tbl#get_cells() would act on `line` as a normal
string (i.e. using `[idx]` and strpart()). This breaks when the column
separator (s:s_rep() or rxTableSep) is a multibyte character like the
vertical box drawing character │. In that case the separator is never
found and only one empty cell is returned. This also affects
vimwiki#tbl#format() which in turn clears the whole table each time the
insert mode is left.
Fix this issue by initially splitting the line into a list of the
characters. Getting the elements with `[idx]` now returns each character
instead of a byte. In addition len() is used in place of strlen() as
well as slicing together with join() instead of strpart().
Fixes: #1297 ("Table formatting breaks on multibyte separator")
If a user is happy with the default vimwiki location, they may not set
the g:vimwiki_list global variable. In this case, the
len(g:vimwiki_list) vimwiki#diary#calendar_sign function throws an
error every time it is called. This function is called for every day
displayed in the calendar which is very noisy. Checking if the variable
exists before doing the length check prevents the error and the rest of
the function works fine if the variable is not set.
Intended to address #1267 - "get_diary_files() takes very long to
execute on Windows - this leads to very slow link-creation using the
Enter key".
In theory, this should be a performance improvement for link creation on
systems with very slow filesystem access (this came up under Windows, but
ought to be a general optimization). It does nothing for generating the
entire list of diary files, which by nature has to get the whole list
anyway.
- Uses a filereadable() call against a path instead of checking the
results of get_diary_files() for most calls.
- Introduces is_among_diary_files() for the case where you want to
check against a get_diary_files() list.
- is_diary_file() will pass through to is_among_diary_files() if it gets
invoked with a second parameter, to retain backwards compatibility on
the off chance anyone has scripted against this behavior. I'm not sure
this is necessary, but trying to be cautious with the implicit API
here.
- Includes a test/api_base_is_diary_file.vader
My original version of this change assumed that input would be in the
form of "2023-01-01", instead of a fully-qualifed filename. As it is,
it expects a full path to the file, in keeping with the list returned by
vimwiki#diary#get_diary_files().
Bug: #1267
Problem:
The loading of a .md file was slow due to multiple nested lookback.
It is advised in vim help to prefer \zs, see :h /\@<
Anyway, it is useless to nest them.
089.080 000.085 000.085: sourcing /home/mtourneb/.vim/after/syntax/vimwiki.vim
089.264 005.184: opening buffers
089.322 000.058: BufEnter autocommands
089.323 000.001: editing files in windows
089.414 000.091: VimEnter autocommands
089.415 000.001: before starting main loop
303.698 214.283: first screen update
303.700 000.002: --- VIM STARTED ---
Solution:
Remove the function s:expand_delimiter in vars.vim.
It was doing a duplicate job with vimwiki#u#hi_expand_regex
And ... well avoid lookback as much as possible
091.784 000.087 000.087: sourcing /home/mtourneb/.vim/after/syntax/vimwiki.vim
091.991 006.034: opening buffers
092.050 000.059: BufEnter autocommands
092.052 000.002: editing files in windows
092.172 000.120: VimEnter autocommands
092.173 000.001: before starting main loop
112.574 020.401: first screen update
112.578 000.004: --- VIM STARTED ---
Setting a global for g:vimwiki_version doesn't seem too egregious, and
lets us reference it cleanly elsewhere.
Removes version header from vimwiki.txt and hardcoding in tagfile
generation.
There's still one instance of that left in test/tag.vader that we should
clean up eventually.
This:
- Changes version number everywhere I could find it.
- Reworks get_version() to dispense with stable / dev division.
- Rewrites CONTRIBUTING.md to describe a new release workflow.
The development/release workflow stuff could use some thought. It's
clunky at best. There's sort of an inherent tension between semver-style
version numbers and the rolling release thing - maybe we should just use
commit hash or something date-based?
Still, it's a start.
It turns out that the target extension here in `ext` has a dot prepended.
Under manual testing, this seems to fix the issue where a wiki with
`markdown_link_ext` set and using `.md` (for example) in links will wind up
taking people to pages with a repeated extension, like `foo.md.md`.
I feel like the repeated use of:
vimwiki#vars#get_wikilocal('ext', link_infos.index)
here should probably be replaced with a named variable like `target_ext`,
but I wanted to touch the minimum lines of code possible for this fix.
If a link contains `.`, then this:
fnamemodify(link_text, ':e')
will return the last .-separated part of the name as extension, and nothing
else will get appended. Instead, we check if the extension is empty _or_
doesn't match the extension for the target wiki, in which case we append
the correct one.
This seems like it works, and might not break anything else, given the
branch of the function that it's in.
Co-authored-by: Brennen Bearnes <code@p1k3.com>
I find this is one of the more common reasons I wind up with way too many
copies of the same file open.
This adds 2 new magic numbers to vimwiki#diary#make_note()'s second
parameter: 4 for `tab drop` and 5 for `drop`. It also checks to make sure
:drop exists in the current build.
Replaces :e in `vimwiki#base#follow_link()` with :drop, making this the
default behavior for pressing <CR> on a link. Checks for the existence
of :drop first, since this isn't available in some builds.
Adds a new :VimwikiTabDropLink and makes this a default for the keybindings
formerly occupied by :VimwikiTabnewLink; leaving :VimwikiTabnewLink
available for backwards compatibility and anyone who still wants the old
behavior.
Doesn't touch the split window reuse functionality, or the
:VimwikiGoBackLink behavior, although I can see an argument for adding
:drop to the latter.
I've wanted this for a while and happened to notice @davidlmontgomery's
patch from 2016 in #238. Code has moved around a little since, but I
think this is correct.
Before updating a listing, check if the update differs from the existing state.
Only then update the buffer, otherwise return early.
Initial reasoning: If `let g:vimwiki_auto_toc = 1` is set, saving a buffer would always update
the `contents`-section, even if there were no updates in this section. This lead to undesired
undo-behavior.
NOTE: this fix was only tested for the toc-listing.
When using vim's native mechanism for `packages`, wildcards (*) are used
in the `runtimepath` since package folders can have arbitrary names.
They were not expanded, thus a path to e.g. `default.tpl` could not be
resolved when calling Vimwiki2HTML.
Resolvesvimwiki/vimwiki#1193
There is an extra space when calling the `basename` command, which causes the following error:
```
Vimwiki: 'basename: extra operand ‘md’
Try ''basename --help'' for more information.
```
Jiaming Chen
Tinmarino
Julian Prein
Joe Planisky
Luke Atkinson
R. David Dunphy
Brennen Bearnes
dmitry kim
Brennen Bearnes
Philipp Oberdiek
Philipp Oberdiek
Jean-Luc Bastarache
Tobias Mersmann
ysftaha
Thomas Leyh
Tulio Brunoro