Drop the personal-wiki launcher in favor of a self-contained fixture
the dev launchers seed into the cache. write_sample_wiki() lays down
one page per vimwiki feature group (syntax, lists, tables, links,
notes, diary) so every plugin feature has something to exercise
without touching a real wiki.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Move the four keymap-harness files (test-keymaps.sh/.lua and the Vim
counterparts) into a development/tests/ subfolder so they sit apart from
the interactive dev launchers. REPO_ROOT now climbs two levels, and the
internal source paths, CI invocations, and ONBOARDING tree are updated.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The root scripts/ folder now holds only download_bin.vim, the one script
the plugin itself invokes (via :NuwikiInstall and the vim-plug/Dein build
hooks). All developer tooling — the start-nvim/start-vim launchers,
test-personal-wiki, the keymap harnesses, and syntax-diag — moves to
development/ alongside the developer docs.
REPO_ROOT resolution is unchanged (development/ sits one level under root,
same as scripts/ did); only the explicit cross-references, the CI harness
invocations, the README launcher paths, and the ONBOARDING structure tree
are updated.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Relocate ONBOARDING.md and nuwiki-architecture.html out of the repo root
into a dedicated development/ folder. No file references either path, so
the move is self-contained.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Relocate start-nvim.sh, start-vim.sh and test-personal-wiki.sh from the
repo root into scripts/, alongside the existing test harnesses. Each
script now derives REPO_ROOT one level up, test-personal-wiki.sh execs
the moved start-vim.sh, and the README + usage comments point at the new
paths.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Two documented checkbox-list mappings did not match their spec:
- `glp` dispatched the same forward-only cycleCheckbox as `gln`, so it
could never "cycle the checkbox state backward". Add `ops::cycle_state_back`
and thread a `reverse` flag through the dispatcher and both clients.
- `gl<Space>` and `gL<Space>` both swept the whole buffer, making them
identical. `gl<Space>` now passes the cursor position and the server
scopes deletion to the contiguous list block under the cursor (current
list, cascading into sublists); `gL<Space>` stays whole-buffer.
Adds Rust unit tests, plus behavioral and full-registration coverage for
the documented mapping surface in the Neovim and Vim keymap harnesses, and
a command-coverage suite mirroring the doc's command groups.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
`{{{lang … }}}` fences were painted as one String group. Add automatic
nested-syntax highlighting (like vimwiki): scan the buffer for the languages
used, `:syntax include` each under a cluster, and define a contained region
per language so e.g. `{{{sh` blocks get shell highlighting and `{{{python`
gets Python. Bare `{{{` / `{{{class="…"` fences fall back to nuwikiPreformatted.
A small alias map maps labels like bash/shell → sh.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Adds STOPPED alongside TODO/DONE/STARTED/FIXME/FIXED/XXX across the stack:
lexer, AST Keyword enum, HTML renderer (class="stopped"), LSP keyword_str,
and the Vim syntax red group (nuwikiKeywordAttn). Grouped with the
attention/pending keywords (red) for both export and in-editor highlighting.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
All keywords shared one `nuwikiKeyword` group linked to `Todo`, so the editor
showed TODO, DONE and XXX in the same colour. Split into nuwikiKeywordAttn
(TODO/FIXME/XXX, red) and nuwikiKeywordDone (DONE/FIXED/STARTED, green) with
explicit colours so the distinction survives any colorscheme — matching the
export's per-keyword classes.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
All keywords previously shared `class="todo"`, so a stylesheet couldn't
distinguish e.g. red TODO from green DONE. Emit a distinct class per keyword
(todo/done/started/fixme/fixed/xxx); TODO keeps the vimwiki `todo` class so
stock vimwiki CSS still styles it.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
A silent `:!xdg-open`/`:!open` swaps to the shell's alternate screen and
returns without repainting, leaving the editor UI garbled until the next
redraw. Force a full repaint after launching the browser. Neovim already
avoids this via async jobstart.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Stock vimwiki templates and stylesheets rendered broken pages because the
HTML output diverged from vimwiki conventions on three fronts:
- Templates: the renderer only substituted nuwiki's `{{key}}` placeholders,
so vimwiki's `%title%`/`%content%`/`%root_path%`/`%date%`/`%wiki_css%`
passed through literally — dropping the body and breaking asset links.
Both delimiter styles are now recognised; `%wiki_css%` aliases the css var.
- Links: `[[todo.wiki]]` exported to `todo.wiki.html`. render_page_html now
takes the wiki file extension and strips it from wiki/interwiki targets
(via index::strip_wiki_extension) before the `.html` URL is built, matching
in-editor navigation. file:/local: links keep their literal extension.
- Tasks: checkbox items used bespoke `task-*` classes plus an `<input>`,
which double-rendered against vimwiki stylesheets. Emit `done0..done4`
(`[ ] [.] [o] [O] [X]`) and `rejected` (`[-]`) classes with no input — the
stylesheet draws the box.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
The picker fetched the wiki list via workspace/executeCommand, but the LSP
only starts once a vimwiki buffer exists — so a wiki could not be selected
until one was already open (chicken-and-egg). Read the list straight from
config instead (g:nuwiki_wikis / scalar fallback, the same source as
open_wiki_path) and open the chosen index directly; that auto-starts the
server via the FileType autocmd.
- Vim: new public nuwiki#commands#wiki_list(); wiki_ui_select() uses it +
inputlist() + :edit. Global :VimwikiUISelect / :NuwikiUISelect commands.
- Neovim: config.wiki_cfg(n) + config.wiki_list() (init.lua delegates);
commands.wiki_ui_select() uses them + vim.ui.select + :edit; global
VimwikiUISelect / NuwikiUISelect user commands registered in setup().
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
concealcursor=nc kept the cursor's own line concealed in normal mode, so
the hidden characters of a wikilink ([[, target, ]]) still occupied buffer
columns — each l/h moved one hidden position with no visible movement,
making the cursor feel stuck. Drop the `n` flag (keep `c` for incsearch) so
the cursor line shows raw markup in normal mode while other lines stay
concealed.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Pure rustfmt churn — the anchor-match, source-relative resolution,
and config-unwrap commits introduced lines that needed rustfmt
reformatting. CI's `cargo fmt --check` job was blocking on this.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Heading-block folds were computed but all collapsed on file open,
so users had to `zR` after every `:edit`. Set `foldlevel=99` once at
ftplugin attach so the fold structure is still there (closeable with
`zc`/`zM`) but the buffer renders fully expanded.
`foldlevel` is window-local; we set it on initial attach only — not
in the Lua BufWinEnter re-apply — so closing folds with `zc` in a
window survives navigating away and back to that buffer. Split
windows still get the same value because Vim copies window-local
options to the new window when splitting.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Pressing <Tab> past the last cell unconditionally appended a new row
below, even when there was already a next row to jump into. Tabbing
through a header → separator → body table from the header row
inserted a blank row between header and separator instead of landing
in the first body cell.
After the final cell, walk forward from the current row to the end
of the table, skip any separator row, and jump to the first cell of
the next data row. Only insert a fresh row when the cursor is on the
table's last row (or only separator rows follow). Applied to both
the Vim (autoload/nuwiki/commands.vim) and Neovim (lua/nuwiki/
commands.lua) sides so the two harnesses stay in sync; covered by
new keymap tests in scripts/test-keymaps-vim.vim and test-keymaps.lua.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
`didOpen` for the wiki's index page fires almost immediately after
`initialized`, but the background `index_workspace` task hasn't yet
parsed the other pages. Link-health diagnostics computed at that
point look up every cross-page link target in a near-empty index and
flag them all as missing-page — and the diagnostics are never
recomputed once the scan catches up, so the editor shows stale red
squiggles on every link until the buffer is edited.
After each wiki's initial scan finishes, walk every open document
that lives in that wiki, recompute its diagnostics against the now-
populated index, and re-publish. Snapshots both the URI list and
each doc's `(ast, text, version)` so we don't hold DashMap iterators
across awaits.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
test-personal-wiki.sh wraps start-vim.sh to open the user's real wiki
at ~/.vimwiki/personal_wiki (override with NUWIKI_PERSONAL_WIKI) — used
when reproducing user-reported bugs against actual content instead of
the seeded smoke-test wiki.
Also adds a NUWIKI_DEV_NO_SEED env var to start-vim.sh so seed_wiki
doesn't drop a scratch Notes.wiki into the user's real notes when we
point WIKI_DIR at an already-populated wiki.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
`[[llm-wiki-pattern]]` in `tips/index.wiki` should resolve to
`tips/llm-wiki-pattern.wiki` — vimwiki's default is source-relative
for non-absolute targets. The index keyed every page by its
workspace-relative path, so the link looked up `llm-wiki-pattern`
(root-relative) and missed the sibling. Goto-definition then
synthesised a non-existent root URL, and `nuwiki.link` diagnostics
flagged every such link as missing-page.
- `WorkspaceIndex::resolve_wiki_path` and `page_for_wiki_target`
centralise the lookup: try `{source_dir}/{path}` first, fall back
to root-relative. Used by goto-definition, diagnostics, and the
workspace `checkLinks` walker.
- `collapse_dots` resolves `.` / `..` segments before lookup so
`[[../Other]]` from `posts/foo` correctly hits `Other` at the
wiki root (and is reported broken if no such page exists, rather
than silently going to nowhere).
- `canonical_link_name` applies the same rule when keying the
backlinks index, so backlinks queried by the resolved page name
pick up references that were written source-relative.
- `OutgoingLink` carries `is_absolute` now so `classify_outgoing`
(driven from cached index data, not the live AST) gets the same
treatment without re-parsing the source.
- `classify_outgoing`'s File/Local branch now reuses
`resolve_file_path` instead of duplicating the `//absolute` /
relative-to-source logic inline.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
`[[Page#Heading text]]` parsed the anchor as the literal string after
`#`, but the index stored each heading's anchor as `slugify(title)`.
The two never compared equal, so every TOC-style anchor link
(`[[#1. Topologia atual]]`, etc.) was reported as a broken anchor and
goto-definition couldn't jump to the right heading either.
Slugify the request before comparing — `slugify` is idempotent on
valid-slug input, so the same code path now accepts both
`[[Page#some-heading]]` (slug form) and `[[Page#Some Heading]]`
(raw text). Applied to diagnostics, navigation, and hover preview so
all three agree on what "an anchor matches a heading" means.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Drop-in script that dumps the syntax/highlight state of the current
buffer to /tmp/nuwiki-syndiag.log: filetype, b:current_syntax, every
syntax/vimwiki.vim found in &runtimepath (to spot duplicates from
other plugins), the resolved attributes of our highlight groups +
the colorscheme groups they link to, a per-segment synID walk over
the buffer, and the filtered scriptnames list. Useful when a user
reports broken highlighting and we need to know whether the patterns
aren't matching, the colorscheme is the culprit, or another plugin's
syntax/vimwiki.vim is winning the runtimepath race.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The single `nuwikiHeading` group linked all six levels to `Title`.
Many popular colorschemes (codedark, gruvbox-light, etc.) define
`Title` as bold-only with no foreground colour, so every heading
rendered as bold text in the Normal colour — H1 and H3 were
visually identical.
Split into nuwikiHeading1..6 and link each to a different accent
(Title / Function / Identifier / Type / Constant / PreProc), matching
the per-level Tree-sitter mapping the Neovim semantic-token side
already uses. The patterns are mutually exclusive: each requires
whitespace immediately after the equals run, so `=` won't match `==`.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two undefined targets meant bold/italic rendered as plain text:
- Neovim semantic-token defaults linked @vimwikiBold/Italic to
markdownBold/Italic, which only exist when the markdown treesitter
parser is loaded — never in a .wiki buffer. Use @markup.strong /
@markup.italic / @markup.raw.inline / @markup.link instead; those
are defined by every nvim 0.9+ colorscheme.
- The static fallback linked nuwikiBold/Italic to Bold/Italic groups
that exist in Neovim but not plain Vim, so on Vim the links
resolved to nothing. Define them with explicit gui=bold/italic
attributes so they render correctly everywhere.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Set conceallevel=2 and concealcursor=nc on .wiki buffers and add
contained `conceal` matches for the bold/italic/code delimiters and
wikilink brackets/targets. The static-fallback rendering now mirrors
what the Neovim semantic-token path produced: *bold* shows as bold
without the asterisks, [[target|desc]] shows just "desc", etc.
The wikilink region uses an inline `contains=` (no line continuation)
because some Vim builds (notably Homebrew Vim 9.2) raise E10 on the
leading `\\ contains=...` form even though the surrounding region
parses fine.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
`[[page.wiki]]` should resolve identically to `[[page]]`. The index
keyed by stem, so links with the explicit extension fell through to
the broken-link diagnostic and to-page navigation failed. Strip the
configured extension before every page lookup (resolve, definition,
backlinks, classify_link, classify_outgoing) and stop appending the
extension in synthesise_page_uri when it's already present.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
workspace/didChangeConfiguration settings from Neovim arrive as
{ "nuwiki": { "wiki_root": ..., "wikis": [...] } } because
server_settings() in lua/nuwiki/lsp.lua wraps the payload in a
"nuwiki" namespace key. apply_change() was trying to deserialise that
outer object directly as InitOptions, which always succeeded but produced
all-None fields (unknown keys are ignored by serde), so every settings
refresh was silently dropped.
Unwrap single-key objects whose value is also an object before
deserialising — this handles the Neovim wrapper while keeping the flat
VimL shape working unchanged.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
VimL dicts serialise boolean-like settings such as `auto_export = 1`
as the JSON number `1`, not `true`. serde's default `Option<bool>`
deserializer rejects that, causing `serde_json::from_value::<InitOptions>`
to return `Err` and `.ok()` to silently discard the entire wikis list.
The server then fell back to `wiki_root` and could not resolve per-wiki
links, marking every cross-wiki link as broken.
Add an `opt_bool_or_int` serde helper that accepts JSON bool, `null`,
or integers (0 = false, non-zero = true), and apply it to `auto_export`,
`html_filename_parameterization`, `listsyms_propagate`, and `auto_toc`
in `RawWiki`.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The server received only wiki_root (the parent directory) and had no
knowledge of the per-wiki sub-roots configured in g:nuwiki_wikis / the
setup() wikis list. It therefore resolved all links relative to the
parent root, causing every link in a sub-wiki to appear broken.
Vim path (autoload/nuwiki/lsp.vim):
s:settings() now includes a 'wikis' key when g:nuwiki_wikis is set,
so the initialization_options and settings payloads carry the full
per-wiki config (root, diary_rel_path, file_extension, …).
Lua path (lua/nuwiki/lsp.lua):
Add resolved_opts() which merges vim.g.nuwiki_wikis into
config.options.wikis when the user configures via VimL rather than
setup({wikis=…}). init_options(), server_settings(), and the
pre-0.11 root_dir_for() fallback all use resolved_opts() so the
server receives the correct per-wiki roots in every code path.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Commands were hardwired to vim-lsp (lsp#send_request, :LspDefinition,
:LspReferences, :LspRename). Users running coc.nvim got "vim-lsp not
loaded" on every command despite the plugin correctly detecting coc.
Changes:
autoload/nuwiki/commands.vim
- Add s:has_coc() (checks CocActionAsync exists)
- Add s:coc_wrap() adapter — converts coc's (err, result) callback into
the vim-lsp notification shape {response:{result:…}} so all existing
on_notification handlers work unchanged
- s:exec(): try vim-lsp first, fall back to CocActionAsync('runCommand',
…) with arguments spread as positional params, then error if neither
- Add s:jump_definition() helper — :LspDefinition vs CocActionAsync
- Add nuwiki#commands#backlinks() — :LspReferences vs CocActionAsync
- badd_link(): coc path uses CocAction('getDefinition') to resolve the
target URI without navigating
- follow_link_or_create(): replace inline :LspDefinition checks with
s:jump_definition()
- rename_file(): :LspRename vs CocActionAsync('rename')
ftplugin/vimwiki.vim
- VimwikiFollowLink / NuwikiFollowLink: dispatch through
nuwiki#commands#follow_link_or_create() instead of :LspDefinition
- VimwikiBacklinks / NuwikiBacklinks / VWB: dispatch through
nuwiki#commands#backlinks() instead of :LspReferences
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
<sfile> inside a command! body is evaluated at execution time, not at
definition time. When :NuwikiInstall is run interactively there is no
sourcing context, so <sfile> expands to nothing and the path collapses
to ./scripts/download_bin.vim — producing E484.
Fix: capture the path in s:plugin_root immediately while the script is
being sourced, then reference that variable from the command body.
Also hoist :NuwikiInstall above the early `finish` for Neovim so the
command is available in both editors (Neovim routes to
require('nuwiki.install').install(), Vim sources the download script).
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The global <Leader>ww / diary helpers read g:nuwiki_wiki_root directly,
so they always opened the root-level index.wiki even when the user had
configured per-wiki roots via g:nuwiki_wikis (e.g. personal_wiki.root =
'~/.vimwiki/personal_wiki').
Vim path (autoload): replace the three open_*_path functions with a
shared s:wiki_cfg(n) helper that checks g:nuwiki_wikis[n] first, then
falls back to the scalar g:nuwiki_* vars and built-in defaults.
Lua path (init.lua): replace the separate _wiki_index_path /_diary_path
locals with a unified _wiki_cfg() that checks setup() opts.wikis, then
vim.g.nuwiki_wikis (for users who configure via VimL rather than
setup()), then scalar opts / g: vars.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Third-party plugins such as vimwiki-sync and vim-zettel call
vimwiki#vars#get_wikilocal() — a function only provided by the original
vimwiki plugin. Without it, the after/ftplugin loaded by vimwiki-sync
errors on every wiki buffer open:
E117: Unknown function: vimwiki#vars#get_wikilocal
E121: Undefined variable: g:zettel_dir
Add autoload/vimwiki/vars.vim with lightweight stubs for the two keys
vimwiki-sync uses ('path', 'is_temporary_wiki'), reading values from
nuwiki's own g:nuwiki_wiki_root / g:nuwiki_file_extension config so
users don't have to duplicate settings. get_global() is stubbed as a
no-op for completeness.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
Two bugs prevented keymaps from working after a Dein install:
1. <Leader>ww and <Leader>wt both called diary_today() instead of
wiki_index() / wiki_tab_index(). Fixed in keymaps.lua (Neovim)
and ftplugin/vimwiki.vim (Vim).
2. All keymaps were buffer-local (<buffer> / buffer=bufnr), so they
only activated inside an already-open .wiki file. Users had no way
to reach the wiki from any other buffer, making the plugin appear
broken on a fresh Vim start.
Fix: register the eight <Leader>w* entry-point mappings globally —
in setup() for Neovim (lua/nuwiki/init.lua) and in plugin/nuwiki.vim
for plain Vim. To avoid a chicken-and-egg LSP dependency, the global
mappings open files directly from config (wiki_root, diary_rel_path,
file_extension); the LSP auto-starts via the FileType autocmd once
the vimwiki buffer loads. Three new autoload helpers
(open_wiki_path, open_diary_path, open_diary_index_path) provide the
LSP-free path for the Vim side.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
gcc-aarch64-linux-gnu ships the cross compiler and binutils only;
linking glibc binaries also requires the target libc crt/dev files
(Scrt1.o, crti.o), which live in libc6-dev-arm64-cross.
Without it, the aarch64-unknown-linux-gnu build failed with:
ld: cannot find Scrt1.o: No such file or directory
ld: cannot find crti.o: No such file or directory
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
actions/upload-artifact@v4 and download-artifact@v4 require the
GitHub-hosted artifact backend, which Gitea Actions does not
implement (GHESNotSupportedError on v2.0.0+). Downgrade to v3.
v3's download-artifact has no merge-multiple flag and nests each
artifact under artifacts/<artifact-name>/, so the upload loop now
recurses to find the archives.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
cross spawns Docker containers per target, but the Gitea Actions
runner doesn't expose /var/run/docker.sock, so every cross-compile
step died with "Cannot connect to the Docker daemon".
Replace with rustup-installed targets plus per-target linker setup:
- x86_64-gnu: native (no extra deps)
- aarch64-gnu: gcc-aarch64-linux-gnu from apt
- x86_64-musl: musl-tools from apt
- aarch64-musl: rust-lld + link-self-contained (rust-std bundles
musl libc.a + crt, no apt deps needed)
Verified aarch64-musl path locally: produces a statically linked
ARM aarch64 ELF.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
The prior commit (e849ac9) intentionally removed the h1-only
restriction from find_section_range so VimwikiGenerateTagLinks stays
idempotent under non-h1 tag headings. The accompanying test still
asserted the old h1-only behavior and now fails.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CI's cargo fmt --check flagged a few lines from the propagation patch
that exceeded the wrapping threshold. No behavioural change.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
nvim defaults autoindent=on, so the indent inserted by Vim/Nvim after
the <CR> we return was being doubled by our own indent prefix —
pressing Enter on a sub-item produced a deeper-nested item instead of
a same-level one. Detect any auto-indent mechanism (autoindent,
smartindent, cindent, indentexpr) and let it own the indentation;
otherwise add it ourselves so vim's noautoindent default still works.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
ensure_binary used to short-circuit whenever bin/nuwiki-ls existed,
which left vim/nvim running an outdated LSP after every source change.
Now always invoke cargo build (incremental, so a no-op when nothing
changed) and refresh the symlink + log the build timestamp. Add
NUWIKI_DEV_SKIP_BUILD=1 for the rare opt-out.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Toggling/cycling/rejecting a nested checkbox now recomputes each
ancestor's progress marker (vimwiki's listsyms_propagate). Mirrors
vimwiki's averaging — rejected items count as 100% unless every child
is rejected, in which case the parent itself becomes rejected. Walks
up while each ancestor has its own checkbox.
Also tightens find_checkbox_span to scan only the marker line, since a
parent item's span covers its nested sublist and was matching child
markers as if they belonged to the parent.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
gffranco