Delete all Rust crates (crates/) and Cargo files. The nuwiki-ls binary
is now built in the separate nuwiki-rs repository and downloaded at
install time from the Gitea release assets, with a cargo build fallback
that clones nuwiki-rs.
Update all documentation to reflect the split repo layout.
Releasing no longer means hand-editing the version across four files. The
Release workflow is now workflow_dispatch with a `version` input:
Actions → Run workflow → version: 0.4.0
A new `prepare` job validates the semver, runs scripts/set-version.sh to
stamp it into the workspace Cargo.toml, the two internal path-dep pins,
g:nuwiki_version, and the Cargo.lock entries for our crates; gates on
`cargo test --workspace`; then commits "chore(release): X.Y.Z" and pushes
the vX.Y.Z tag. The build matrix and release job run off that freshly
pushed tag (checkout ref = the new tag), so the tagged commit carries the
real version.
scripts/set-version.sh is the single source of truth for where the version
lives — run it locally with the same arg to bump by hand. It patches
Cargo.lock via awk so it needs no cargo/toolchain.
The push:tags trigger is removed (the pipeline now creates the tag itself,
so a tag-push trigger would double-fire).
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The dein/vim-plug build hook runs `vim -e -s -c "source scripts/download_bin.vim"`,
which starts in 'compatible' mode. There, leading-backslash line
continuations aren't recognised, so the multi-line URL string raised
`E10: \ should be followed by /, ? or &`. The install actually succeeded
(binary downloaded), but in silent Ex mode any emitted error makes Vim
exit non-zero — so dein reported a build failure. :NuwikiInstall worked
because a normal session is 'nocompatible'.
Wrap the script in the standard `let s:cpo_save = &cpo | set cpo&vim` …
`let &cpo = s:cpo_save` guard so line continuations parse regardless of
how it's invoked. Verified the exact dein command now exits 0.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Gitea serves the most recent release under the `latest` tag segment, so
`/releases/download/latest/nuwiki-ls-<target>.tar.gz` resolves directly
(HTTP 200) — no API lookup or JSON parsing needed. Drop the
latest_asset_url helpers added in the previous fix; the only real bug was
the URL shape (`/releases/latest/download/` is GitHub-only and 404s on
Gitea). Both install paths now just pick the target and build the URL.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
:NuwikiInstall always fell back to building from source because the
download URL used the GitHub-only shape
`/releases/latest/download/<asset>`, which this Gitea instance serves as
404. The release assets exist and are correctly named; only the URL was
wrong.
Resolve the asset's real download URL through the Gitea API instead: GET
`/api/v1/repos/.../releases/latest`, find the asset matching
`nuwiki-ls-<target>.tar.gz`, and download its `browser_download_url`
(`/releases/download/<tag>/<asset>`, which returns 200). Fixed in both
install paths — lua/nuwiki/install.lua (vim.json.decode) and
scripts/download_bin.vim (json_decode). Verified end to end: resolve →
download → extract → executable binary.
Co-Authored-By: Claude Opus 4.8 <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>
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>
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>
Close the remaining Vim-vs-Neovim functional gaps:
* Text objects — new `autoload/nuwiki/textobjects.vim` mirroring
`lua/nuwiki/textobjects.lua`. All five pairs (ah/ih, aH/iH,
al/il, a\/i\, ac/ic) wired in the Vim path of `ftplugin/vimwiki.vim`
via the classic `:<C-u>call` idiom that works cleanly in both
operator-pending and visual modes.
* Folding — new `autoload/nuwiki/folding.vim` providing a regex
`foldexpr` over headings, plus a tidy `foldtext`. Wired in the
Vim path; opts out via `let g:nuwiki_no_folding = 1`.
* `smart_return` was using `setline()` / `append()` inside an
`<expr>` callback — fine on Neovim but plain Vim's stricter
textlock raised E565. Rewrote as pure keystrokes (matches the
Lua version): table rows insert via `<CR>...<Esc>0li`, empty
list lines break via `<Esc>0DA<CR>`.
* Function-name parity: added
`nuwiki#commands#heading_add_level`,
`nuwiki#commands#heading_remove_level`,
`nuwiki#commands#table_move_column_{left,right}` as thin
aliases over the original short names so the public
`nuwiki#commands#*` surface matches `require('nuwiki.commands').*`.
Test harness:
* `scripts/test-keymaps-vim.vim` extended from 12 to 30 cases:
4 smart_return, 3 smart_tab/<S-Tab>, 3 named-command exists,
4 text-object helpers, 1 `dah` end-to-end, 2 folding, and the
original 12 pure-VimL cases. (Visual-mode mark inspection in
`vim -e -s` stays unreliable, so text objects are exercised
at the helper level plus one operator-pending end-to-end.)
Gates: 456 Rust / 39 Neovim / 30 Vim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`<Tab>` / `<S-Tab>` in insert mode navigate between table cells when
the cursor is on a `|…|` row, otherwise they pass through to their
default insert-mode behaviour (literal tab / shift-tab). Same
`<expr>`-mapping idiom as Cluster 5's smart_return.
<Tab> next cell; creates a fresh row below if past the last cell
<S-Tab> previous cell; no-op past the first
Cursor lands immediately after the destination cell's leading `|`,
matching upstream vimwiki's `vimwiki#tbl#go_*_cell` behaviour.
Both Lua and VimL counterparts wired into ftplugin's existing
table_editing block (gated alongside `gqq`/`<A-Left>`).
Tests: 4 new in scripts/test-keymaps.lua — next cell, new-row
overflow, previous cell, and pass-through outside a table.
Gates: 421 Rust / 39 Neovim / 12 Vim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Bind `<CR>` in insert mode as an `<expr>` map to a `smart_return`
helper. Behaviour matches upstream vimwiki's `:VimwikiReturn`:
- On a list line with content → continue the list with the same
marker on a new line (preserving leading checkbox `[ ]` if any).
- On an empty list line (marker only) → clear the marker and break
out of the list with a plain newline.
- Inside a `|…|` table row → insert a fresh empty row below with
the same column count, cursor lands inside the first cell.
- Otherwise → plain `<CR>`.
Implementation note: the helper is `<expr>`-bound, which means
textlock is active and the buffer can't be mutated from inside the
callback. The function returns key sequences only — including
`<Esc>0DA<CR>` for the empty-list break and `<Esc>0li` for the
table-row cursor jump — so every effect flows through Vim's normal
keystroke pipeline and stays undo-coherent.
Both Lua and VimL counterparts shipped.
Tests: 4 new in scripts/test-keymaps.lua covering list continuation,
checkbox preservation, empty-marker break-out, and new table row.
Gates: 421 Rust / 35 Neovim / 12 Vim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five operator-pending + visual text-object pairs, matching upstream
vimwiki:
ah / ih heading section only (stops at any next heading)
aH / iH heading section + sub-tree (stops at same-or-shallower)
al / il list item (line, with / without marker prefix + checkbox)
a\ / i\ table cell (with / without surrounding `|` separators)
ac / ic table column (with / without the header separator row)
Pre-existing `ah` used aH semantics — fixed to match vimwiki: `ah`
now stops at the next heading regardless of level. `aH` is the new
descendants-inclusive variant.
Implementation notes:
- Pure regex; no LSP round-trip. Each helper computes a `(line, col)`
rectangle and drives the selection with feedkeys.
- Operator-pending and visual need different feed sequences. In visual
the previous anchor is sticky, so bounce through `<Esc>` first. In
operator-pending an `<Esc>` would CANCEL the pending operator —
feed the visual keys directly so Vim treats them as the motion.
- Re-exported helpers (`_heading_block`, `_cell_ranges`,
`_current_cell_for_line`, `_table_bounds`) keep the integration
surface testable without going through Vim's visual-mode pipeline,
whose `'<`/`'>` mark semantics fight headless test harnesses.
Tests: 6 new in scripts/test-keymaps.lua — five pure-helper cases
plus one end-to-end `dah` deletion to verify the operator-pending
wiring.
Gates: 421 Rust / 31 Neovim / 12 Vim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Match upstream vimwiki's insert-mode list editing:
<C-D> list_change_level(-1) dedent current item
<C-T> list_change_level(+1) indent current item
<C-L><C-J> list_cycle_symbol(+1) cycle marker forward
<C-L><C-K> list_cycle_symbol(-1) cycle marker backward
<C-L><C-M> list_toggle_or_add_chk toggle if has checkbox, else add `[ ]`
Cycle order matches vimwiki's canonical run:
- → * → # → 1. → 1) → a) → A) → i) → I) → (wrap)
Both Lua and VimL paths wired. Lua callbacks fire directly (no `<C-o>`
dance) since list_change_level / changeSymbol mutate the buffer via
the LSP applyEdit pipeline, which works cleanly in insert mode. VimL
keeps the `<C-o>:call …<CR>` idiom since that's the standard there.
Harness adds 5 cases covering the new insert-mode bindings plus one
direct LSP roundtrip for changeSymbol — surfaced a long-standing
stale-binary footgun in test-keymaps.sh (rebuilt only when bin was
missing; now rebuilds every run since incremental cargo is fast).
Harness also captures server log_messages and dumps them on failure
so future swallowed-error bugs (Err → log + Ok(None)) are visible.
Gates: 421 Rust / 25 Neovim / 12 Vim.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Closing out §13.1's smallest cluster. Three commands move from
"not yet implemented" stubs to real behaviour:
- `nuwiki.link.pasteWikilink` (server, executeCommand) — derives
the current page name from the source URI + wiki root, returns a
`WorkspaceEdit` inserting `[[<page>]]` at the requested cursor
position.
- `nuwiki.link.pasteUrl` (server) — same lookup, but the inserted
text is the page's relative HTML output URL
(`<page>.html`, including subdir segments) so the snippet survives
when the export root moves.
- `nuwiki.link.normalize` (client) — wraps the word at cursor as
`[[word]]` without following. Reuses the `wrap_cword_as_wikilink`
helper that already powers the `<CR>` two-step. Pure-VimL on the
Vim path; pure-Lua on the Neovim path. No LSP round-trip.
Keymaps:
- `+` (normal + visual) now actually calls `normalize_link` on both
editor paths instead of stubbing with a "deferred" notification.
Tests:
- 5 new Rust unit tests in `cluster_c_link_helpers.rs` covering
command-list presence + the page-name derivation for root and
subdirectory pages + the URL / wikilink shape strings.
- Neovim keymap harness gains a `links.normalize_via_+` case (20
passing now, up from 19). Vim harness inherits the same
command-presence check via its existing smoke tests.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
CI log for the keymaps job showed Lua crashing with
`attempt to call field 'get_clients' (a nil value)` plus a flood of
`W18: Invalid character in group name` on the `@vimwiki*` highlight
links. Both signal that apt's `neovim` package is too old (Ubuntu
LTS ships 0.6/0.7) — `vim.lsp.get_clients` only landed in 0.10, and
Tree-sitter-style `@group.modifier` highlight names only became
valid sometime in the 0.8-ish era.
Switched the installer to pull the official static tarball
(`neovim/neovim/releases/v0.11.3/nvim-linux-x86_64.tar.gz`) and
symlink it into `/usr/local/bin/nvim`. Vim from apt is kept — it
only runs the pure-VimL harness so its age doesn't matter.
Also added a defensive fallback in the harness:
`get_lsp_clients = vim.lsp.get_clients or vim.lsp.get_active_clients`,
plus a `server_capabilities` nil guard. Won't matter on the pinned
0.11 runner, but means a future regression that drops the deps will
fail with a clear `lsp.attached` FAIL message instead of a Lua
stack trace.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
User reported the editor-keymaps CI job hung for 30 minutes before
"Has been cancelled" — twice in a row. Two changes to keep that from
happening again:
1. Wrap both harness invocations with `timeout` (60s Neovim, 30s
Vim). A misbehaving LSP attach, a stuck `vim.defer_fn`, or
ex-mode hung on stdin can no longer hold the runner indefinitely;
`timeout` kills the process and the wrapper exits non-zero so CI
shows the real failure shape.
2. `</dev/null` on `vim -e -s` so ex-mode doesn't sit waiting for
input on CI's non-TTY stdin (the local zsh run inherits the
user's tty, masking this).
3. CI: the keymaps job now has `needs: test` and
`timeout-minutes: 10`. `needs: test` serialises behind the cargo
jobs so we don't fight for the single Gitea runner, and the
workflow-level timeout is a belt to the script-level `timeout`'s
braces — if for some reason both fail the same way, the runner
reclaims itself in 10 minutes instead of 30.
Local timing: `test-keymaps.sh` 10.4s, `test-keymaps-vim.sh` 0.1s.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
User asked about the Vim path of the keymap suite. Building it
surfaced a real bug: 25 of our autoload functions were written as
`function! foo() abort | call bar() | endfunction` one-liners, which
isn't valid Vim syntax — `function!` requires a multi-line body, and
Vim parses the `|` after `abort` as an unexpected trailing character
(E488). Most invocations of the buggy autoload functions errored
out the moment Vim tried to parse them, which is why the user saw
broken keymaps and confusing diagnostics.
Rewrote all 25 one-liners (`autoload/nuwiki/commands.vim`) into the
standard three-line form. Affected groups: `diary_*`, `toggle_list_item`,
`cycle_list_item`, `reject_list_item`, `heading_add`, `heading_remove`,
`toc_generate`, `links_generate`, `export_current` / `_all` /
`_all_force` / `_rss`, and the §13.1 deferred stubs (`list_change_lvl`,
`list_remove_done`, `table_*`, `colorize`, `paste_link`, `paste_url`).
Added `scripts/test-keymaps-vim.{sh,vim}` — a Vim-side counterpart of
the Neovim harness covering the pure-VimL bindings (header nav,
link nav, `o`/`O` bullet continuation, `<CR>` wrap step). The
LSP-roundtrip bindings (`<C-Space>`, `=`, `-`, …) stay on the Neovim
side because vim-lsp's async layer uses timers that don't fire
inside `vim -e -s` headless mode — their server-side codepath is
already exercised by the Neovim harness and the cargo test suite.
Vim harness covers 12 cases: filetype + 2 command-presence smoke
tests, 4 header-nav (`]]`/`[[`/`]=`/`]u`), `<Tab>` link-nav, `<CR>`
wrap-on-first-press, and 3 bullet-continuation flows.
CI: extended the existing `keymaps` job to install both nvim + vim
and run both harnesses. Verified locally:
Neovim harness: 19 passed, 0 failed
Vim harness: 12 passed, 0 failed
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
`scripts/test-keymaps.sh` boots `nvim --headless` against a scratch
wiki, sources `scripts/test-keymaps.lua`, and asserts the post-state
of each buffer-local mapping registered by `ftplugin/vimwiki.vim` +
`lua/nuwiki/keymaps.lua`.
The harness covers the bindings that actually have observable effects
on the buffer or cursor:
- **Lists** (LSP round-trip): `<C-Space>` / `<C-@>` / `<Nul>` toggle,
`gln` cycle, `glx` reject, plus a `[X]→[ ]` round-trip.
- **Tasks**: `gnt` jumps to the next unfinished `[ ]`.
- **Headings** (LSP round-trip): `=` adds a level, `-` removes one,
`=` on h6 clamps without growing.
- **Header nav** (pure Lua): `]]` next, `[[` prev, `]u` parent.
- **Link nav** (pure Lua): `<Tab>` to the next `[[…]]`.
- **`<CR>` two-step**: first press wraps a bare word as `[[word]]`
and stops (matches vimwiki's review-then-follow).
- **Bullet continuation**: `o` / `O` keep the marker, and `o` on a
`[ ]` line carries the checkbox forward.
Each case sets buffer lines, places the cursor (1-based), fires
`nvim_feedkeys` with `replace_termcodes`, waits long enough for the
LSP `executeCommand` round-trip (800 ms by default; 100 ms for the
pure-client maps), and asserts either the resulting lines, a specific
line, the cursor's new line, or the buffer's new name.
Output goes via `$NUWIKI_KEYMAP_RESULTS` (env-var IPC keeps it
robust to `--clean -u` argument parsing). Exit code mirrors the
harness: 0 on green, 1 on any FAIL. Local run yields:
SUMMARY: 19 passed, 0 failed
CI: new `keymaps` job in `.gitea/workflows/ci.yaml` installs Neovim
and runs the script, so a regression in `lua/nuwiki/keymaps.lua`,
`autoload/nuwiki/commands.vim`, or the LSP command surface is caught
on every push.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
P5 + P6 resolved (SPEC §11): Neovim 0.11+ and Vim 9.1+.
Universal Vim entry:
- plugin/nuwiki.vim dispatches Vim vs Neovim. Neovim path waits for
the user's `require('nuwiki').setup()` call; Vim path starts the
LSP client on `FileType vimwiki` and exposes `:NuwikiInstall`.
- ftdetect/nuwiki.vim associates `*.wiki` with the `vimwiki` filetype.
- ftplugin/nuwiki.vim sets commentstring + comments + formatoptions +
suffixesadd + iskeyword, with a clean `b:undo_ftplugin`.
- syntax/nuwiki.vim ships default `@vimwiki*` highlight links for the
semantic-token legend (with `level1..6` + `centered` modifiers),
plus a minimal regex fallback so static highlighting works before
the LSP attaches.
Lua glue (Neovim 0.11+):
- init.lua exposes setup() / install() / health() — wiring only per
SPEC §6.2.
- config.lua holds the §7.5 schema (wiki_root / file_extension /
syntax / log_level) and a tbl_deep_extend apply().
- lsp.lua registers via `vim.lsp.config{} + vim.lsp.enable` on 0.11+;
falls back to a FileType autocmd calling `vim.lsp.start` on older
builds. Root dir walks up for `.git` / `.nuwiki`, then uses
wiki_root.
- install.lua does target-triple detection (Linux gnu/musl x x86_64/
aarch64, macOS arm/x86, Windows), curl+tar download from the
release URL, cp/chmod into bin/, with a cargo fallback. Honours
`g:nuwiki_build_from_source`.
- health.lua implements `:checkhealth nuwiki` per §7.6.
VimL glue (Vim 9.1+):
- autoload/nuwiki/lsp.vim follows §7.4 preference order:
vim-lsp (calls `lsp#register_server`) → coc.nvim (prints the
coc-settings.json snippet) → error.
- scripts/download_bin.vim mirrors the Lua installer for Dein /
vim-plug build hooks — same target triples, same download →
fallback → cargo build chain.
Help: doc/nuwiki.txt with tags for install (lazy.nvim / vim-plug /
Dein), config options, health, LSP feature list, `:NuwikiInstall`.
Rust workspace unchanged this phase; 172 tests still green locally.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
gffranco
gffranco