feat: heph migrate-links — rewrite legacy [[Name]] links to [[id]] (§8.4)
Some checks failed
Build / validate (pull_request) Failing after 11s

`wikilink::to_ids` rewrites name-addressed links to the canonical id
(id-first resolve: an already-id target is left alone, a name → its id
with any label preserved). `Store::migrate_wikilinks_to_ids` runs it over
every body and re-saves through update_node (which collapses + materializes
by id); idempotent. Surfaced as the `migrate.wikilinks` RPC + RemoteStore
forward + the `heph migrate-links` CLI command (not auto-run — the owner
runs it once per store).

Name-resolution + the canonical-context hack stay for now so legacy links
keep resolving pre-migration; retiring them is a later tidy. Tests:
to_ids unit + a heph-core migrate integration (rewrite + materialize +
idempotency).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Erich Blume 2026-06-03 12:40:57 -07:00
commit b112b0d7c1
9 changed files with 101 additions and 3 deletions

View file

@ -321,6 +321,29 @@ impl Store for LocalStore {
tags::of(&self.conn, node_id)
}
fn migrate_wikilinks_to_ids(&mut self) -> Result<usize> {
let owner = self.owner_id.clone();
let candidates: Vec<(String, String)> = {
let mut stmt = self.conn.prepare(
"SELECT id, body FROM nodes
WHERE owner_id = ?1 AND tombstoned = 0 AND body IS NOT NULL",
)?;
let rows = stmt.query_map([&owner], |r| Ok((r.get(0)?, r.get(1)?)))?;
rows.collect::<rusqlite::Result<Vec<_>>>()?
};
let mut changed = 0;
for (id, body) in candidates {
let new_body = crate::wikilink::to_ids(&body, |t| {
links::resolve_id(&self.conn, &owner, t).ok().flatten()
});
if new_body != body {
self.update_node(&id, None, Some(new_body))?;
changed += 1;
}
}
Ok(changed)
}
fn log_append(&mut self, task_id: &str, text: &str) -> Result<()> {
let now = self.clock.now_ms();
let tx = self.conn.transaction()?;

View file

@ -157,6 +157,12 @@ pub trait Store {
/// [`Store::list_nodes`] with [`NodeKind::Tag`].)
fn tags_of(&self, node_id: &str) -> Result<Vec<String>>;
/// One-time migration (tech-spec §8.4): rewrite legacy name-addressed body
/// links `[[Name]]` to the canonical `[[NODEID]]`, resolving each name and
/// re-materializing the `wiki` links by id. Idempotent (already-id links are
/// left alone). Returns the number of nodes whose body changed.
fn migrate_wikilinks_to_ids(&mut self) -> Result<usize>;
// --- per-task log ([[design]] §6.4) ---
/// Append a line to a task's append-only log (creating the log on first

View file

@ -71,6 +71,22 @@ pub fn collapse(body: &str, title_of: impl Fn(&str) -> Option<String>) -> String
})
}
/// Rewrite legacy name-addressed links `[[Name]]`/`[[Name|text]]` to the
/// canonical `[[NODEID]]`/`[[NODEID|text]]` (the one-time migration, §8.4).
/// `resolve(target)` returns the node id a target resolves to (id-first, then
/// name): a target that already *is* an id resolves to itself and is left
/// alone; a name resolves to a different id and is rewritten (its display text,
/// if any, preserved); an unresolvable target is left untouched.
pub fn to_ids(body: &str, resolve: impl Fn(&str) -> Option<String>) -> String {
rewrite_spans(body, |target, display| match resolve(target) {
Some(id) if id != target => Some(match display {
Some(d) => format!("[[{id}|{d}]]"),
None => format!("[[{id}]]"),
}),
_ => None,
})
}
#[cfg(test)]
mod tests {
use super::*;
@ -110,6 +126,21 @@ mod tests {
assert_eq!(collapse(&shown, &t), stored);
}
#[test]
fn to_ids_rewrites_names_keeps_ids_and_preserves_labels() {
// `resolve`: a known name → its id; an id → itself; else None.
let resolve = |t: &str| match t {
"Roof" => Some("01ID".to_string()),
"01ID" => Some("01ID".to_string()),
"02ID" => Some("02ID".to_string()),
_ => None,
};
assert_eq!(to_ids("see [[Roof]]", resolve), "see [[01ID]]");
assert_eq!(to_ids("[[Roof|my label]]", resolve), "[[01ID|my label]]");
assert_eq!(to_ids("[[01ID]]", resolve), "[[01ID]]"); // already an id
assert_eq!(to_ids("[[Unknown]]", resolve), "[[Unknown]]"); // unresolvable
}
#[test]
fn preserves_surrounding_text_and_handles_unterminated() {
let t = titles();

View file

@ -37,3 +37,28 @@ fn update_collapses_name_matching_labels_and_materializes_by_id() {
let u2 = s.update_node(&src.id, None, Some(custom.clone())).unwrap();
assert_eq!(u2.body.as_deref(), Some(custom.as_str()));
}
#[test]
fn migrate_rewrites_legacy_name_links_to_ids() {
let mut s = store();
let target = s.create_node(NewNode::doc("Roof", "")).unwrap();
// A legacy body authored with a name-addressed link (pre-§8.4).
let src = s
.create_node(NewNode::doc("Daily", "fix the [[Roof]] soon"))
.unwrap();
let changed = s.migrate_wikilinks_to_ids().unwrap();
assert_eq!(changed, 1);
// The name was rewritten to the canonical bare id, and the wiki link is by id.
let body = s.get_node(&src.id).unwrap().unwrap().body.unwrap();
assert_eq!(body, format!("fix the [[{}]] soon", target.id));
assert!(s
.outgoing_links(&src.id)
.unwrap()
.iter()
.any(|l| l.link_type == LinkType::Wiki && l.dst_id == target.id));
// Idempotent: a second run finds nothing to change.
assert_eq!(s.migrate_wikilinks_to_ids().unwrap(), 0);
}

View file

@ -230,6 +230,9 @@ enum Command {
/// Destination directory (created if needed).
dir: PathBuf,
},
/// One-time: rewrite legacy `[[Name]]` body links to the canonical
/// `[[node-id]]` form (tech-spec §8.4). Idempotent.
MigrateLinks,
/// Manage the hephd daemon as an OS service (launchd / systemd).
Daemon {
#[command(subcommand)]
@ -726,6 +729,11 @@ fn main() -> Result<()> {
let count = result.get("count").and_then(Value::as_u64).unwrap_or(0);
println!("Exported {count} nodes to {}", dir.display());
}
Command::MigrateLinks => {
let result = client.call("migrate.wikilinks", json!({}))?;
let n = result.as_u64().unwrap_or(0);
println!("Rewrote legacy [[Name]] links to [[id]] in {n} node(s).");
}
Command::Auth { .. } => unreachable!("auth is handled before connecting"),
Command::Daemon { .. } => unreachable!("daemon is handled before connecting"),
}

View file

@ -248,6 +248,10 @@ impl Store for RemoteStore {
self.call_as("tag.list", json!({ "node_id": node_id }))
}
fn migrate_wikilinks_to_ids(&mut self) -> Result<usize> {
self.call_as("migrate.wikilinks", json!({}))
}
fn log_append(&mut self, task_id: &str, text: &str) -> Result<()> {
self.call("log.append", json!({ "task_id": task_id, "text": text }))
.map(|_| ())

View file

@ -422,6 +422,7 @@ pub fn dispatch(store: &mut dyn Store, method: &str, params: Value) -> Result<Va
let p: TagParams = parse(params)?;
json!(store.tags_of(&p.node_id)?)
}
"migrate.wikilinks" => json!(store.migrate_wikilinks_to_ids()?),
"log.append" => {
let p: LogAppendParams = parse(params)?;
store.log_append(&p.task_id, &p.text)?;

View file

@ -28,7 +28,7 @@ Begin the v1 prototype (Phase 1, tech-spec §11.1), built in TDD slices:
- `heph-tui` task-list visuals (§8.1): each row now leads with an attention **flag** (`⚑`, colored red/orange/blue; blank for white) and a **project-colored bullet** — the bullet's color is derived stably from the project id (so it survives projects being added/removed), letting you scan a mixed list by project at a glance. The list also grows a **scrollbar** and keeps the selected task scrolled into view when there are more tasks than fit.
- `heph-tui` sort toggle (§8.1): **`s`** flips the task list between two orders — **default** (attention → most-overdue → project → creation) and **by-project** (grouped under dimmed `──── Project ────` separators, then the same sub-order). The view's filter still applies first. (To free `s`, **skip** moved to **`S`**.)
- `heph.nvim` task-view rows (§8): `:Heph next`/`:Heph list` rows now show a compact **do/late date chip** (and a recurrence `↻`), so you can see scheduling at a glance; `<CR>` still jumps to a task's context doc.
- Wiki-links by node id (§8.4): node resolution is now **id-first** (`[[NODEID]]` resolves to its node ahead of any name match, so links can't be shadowed by a like-named node), and heph.nvim grows a **`[[` picker** — type `[[` (or `:Heph link`) to search your nodes and insert a canonical `[[NODEID]]` link, with a "+ Create new doc" entry that mints one on the spot. Following such a link (`<CR>`) jumps straight by id. Those id links are kept **readable**: on read a bare `[[NODEID]]` is expanded to `[[NODEID|Current Name]]` (so it follows renames, in both the nvim buffer and the TUI preview), and on save it collapses back to the canonical bare id — a custom `|label` you write is preserved as an override. In the editor the id is **concealed** — a link renders as just its name, styled like a hyperlink, with the raw `[[id|Name]]` revealed on the line your cursor is on. (Legacy `[[Name]]` links still resolve until a one-time migration rewrites them.)
- Wiki-links by node id (§8.4): node resolution is now **id-first** (`[[NODEID]]` resolves to its node ahead of any name match, so links can't be shadowed by a like-named node), and heph.nvim grows a **`[[` picker** — type `[[` (or `:Heph link`) to search your nodes and insert a canonical `[[NODEID]]` link, with a "+ Create new doc" entry that mints one on the spot. Following such a link (`<CR>`) jumps straight by id. Those id links are kept **readable**: on read a bare `[[NODEID]]` is expanded to `[[NODEID|Current Name]]` (so it follows renames, in both the nvim buffer and the TUI preview), and on save it collapses back to the canonical bare id — a custom `|label` you write is preserved as an override. In the editor the id is **concealed** — a link renders as just its name, styled like a hyperlink, with the raw `[[id|Name]]` revealed on the line your cursor is on. Legacy `[[Name]]` links keep working, and **`heph migrate-links`** rewrites them to the canonical id form in one pass when you're ready (idempotent).
- Frontmatter editing in heph.nvim (§8.3): opening a node now shows an editable **YAML frontmatter** block on top of the body (`id`/`kind`/`title`/`tags`, and for a task or its context doc the task's `state`/`attention`/`do_date`/`late_on`/`recurrence`/`project`). On save, the plugin diffs the block and issues the right RPC per changed field — rename, set-attention, reschedule (dates as `YYYY-MM-DD`), move-to-project (by name), and tag add/remove — then saves the body; the store strips the block so it never persists. A mistyped `state` surfaces a validation error; a buffer with no block changes no metadata (so deleting the block can't wipe your tags). Inline **`#hashtags`** typed in the body are also added as tags on save (a `# heading` doesn't count) and are rendered in **italics** so they stand out. Link-follow and promotion are unaffected (they're content-relative, not line-absolute).
- Frontmatter projection (§8.3): a node can now be fetched with an editable **YAML frontmatter** block prepended — `node.get {frontmatter: true}` renders `id`/`kind`/`title`/`tags`, and for a task (or its context doc) the owning task's `state`/`attention`/`do_date`/`late_on`/`recurrence`/`project` plus a `task:` ref. Dates are local `YYYY-MM-DD`. On write, the store **strips and ignores** any leading frontmatter (conservatively — a real `---` hrule in prose survives) before the CRDT diff, so frontmatter never persists and an unchanged read→write is a no-op; a naive editor can't corrupt metadata. This is the read/write groundwork for editing a node's metadata as frontmatter in heph.nvim (the diff-into-RPCs layer is next).
- Tags (§4, §8.3): nodes can now be **tagged**. A tag is a `tag`-kind node whose id is deterministic in `(owner, name)`, so the same name is **one canonical tag** shared across everything it's applied to (and replicas converge — no duplicate tags). Tagging is an OR-set link, so adding/removing is idempotent and merge-safe. Surfaced as `tag.add`/`tag.remove`/`tag.list` RPCs and `heph tag add|rm|list` (list a node's tags, or every tag with no node). Tag names are trimmed; a canonical case/spelling normalization is deferred to the future zk import. This is the groundwork for the `tags:` line of the upcoming frontmatter edit surface.

View file

@ -338,7 +338,7 @@ Field rules: `id`/`kind` are **read-only** (display only); `title`, `attention`,
- **At rest (target):** `[[NODEID]]`, or `[[NODEID|custom text]]` when the author wrote explicit display text. The id before the `|` is the target.
- ✅ **Projection (same philosophy as §8.3):** `heph-core::wikilink` (pure, injected id→title) — `node.get` **expands** a bare `[[NODEID]]``[[NODEID|Current Name]]` (every read, so the nvim buffer *and* the TUI preview are readable), and `update_node` **collapses** a `|text` equal to the target's current name back to bare before the CRDT diff (a custom label is preserved as an override). Transform order: read = expand links → prepend frontmatter; write = strip frontmatter → collapse links → store. An unchanged read→write round-trips to the canonical bare id. *(`heph export` still emits raw ids — a later polish.)*
- ✅ **`heph.nvim` display:** `conceal.lua` hides the `[[id|` prefix and the `]]` suffix with conceal extmarks (refreshed on edit), leaving the label as a styled `HephLink` hyperlink; `conceallevel=2` + empty `concealcursor` reveal the raw link on the cursor's line so it stays editable.
- **Migration:** a **one-time fixup** rewrites existing `[[Title]]` bodies to `[[NODEID]]` (resolve→id; flag the unresolvable), after which name-resolution + the canonical-context hack are removed. No special care is warranted (no critical data yet); a first-class migrations feature stays **deferred**.
- **Migration:** **`heph migrate-links`** (the `migrate.wikilinks` RPC → `Store::migrate_wikilinks_to_ids``wikilink::to_ids`) rewrites legacy `[[Name]]` bodies to `[[NODEID]]` and re-materializes the `wiki` links by id; idempotent (already-id links untouched). It is **not auto-run** — the owner runs it once per store. Name-resolution and the canonical-context hack **stay for now** (legacy links keep working until the migration has been run everywhere); removing them is a later tidy. A first-class migrations feature stays **deferred**.
## 9. Testing strategy (TDD, layered)
@ -470,7 +470,7 @@ See [[design]] §5§7 for the constraints later phases impose on present choi
- ✅ **nvim task-navigation polish (§8) — DONE:** `:Heph next`/`list` rows now carry a compact **do/late date chip** (and a recurrence `↻`); `<CR>` already jumps to a row's canonical-context doc (read/navigate, not field-edit).
3. ✅ **Tags (§4, §8.3) — DONE:** a tag is a `tag`-kind node whose id is **deterministic in `(owner, name)`** (`tag:<owner>:<name>`, like the journal), so a name is one canonical tag and replicas converge — no duplicate tag nodes. Tagging is an **OR-set `tagged` link** (mirroring `in-project`): `Store::add_tag` (get-or-create the tag node, idempotent link), `remove_tag` (tombstone the link), `tags_of` (sorted names); enumerate all tags via `list_nodes(Tag)`. RPCs `tag.add`/`tag.remove`/`tag.list` (+ RemoteStore forward); CLI `heph tag add|rm|list`. Names are trimmed, case preserved (canonical normalization deferred to the zk import). Unblocks the `tags:` line of the frontmatter surface (§8.3) and the eventual zk import; inline `#hashtags` remain a heph.nvim concern (§8.3).
4. ✅ **YAML frontmatter as an edit surface (§8.3) — DONE:** the projection — `heph-core::frontmatter::strip` (conservative, runs in `update_node` before the CRDT diff) + `hephd::frontmatter::render` (local-tz dates via `datespec::fmt_iso`) behind `node.get {frontmatter: true}`; a task's context-doc surfaces the owning task's scalars + a `task:` ref; round-trip is a no-op and inbound frontmatter is always stripped (safe vs any client). And the `heph.nvim` smart client (`frontmatter.lua`): the buffer opens with the editable block, and `BufWriteCmd` diffs it → `title`→rename / `attention`→set_attention / dates→set_schedule / `project`→set_project / `tags`→tag.add·remove (a no-block buffer touches no metadata), and inline `#hashtags` in the body are unioned into the tag set on save.
5. **Wiki-links by node id (§8.4) — authoring + display DONE, migration next:** id-first resolution; the `heph.nvim` `[[` picker (`search` → insert `[[NODEID|Name]]`, "+ Create" mints a doc) + id-direct follow; the **expand-on-read / collapse-on-write** projection (`heph-core::wikilink`); and **conceal display** (`conceal.lua` hides the id, shows the label as a `HephLink`, reveals on the cursor line). ⏳ Remaining: the one-time `[[Title]]``[[NODEID]]` migration (then retire name-resolution + the canonical-context hack). See §8.4.
5. **Wiki-links by node id (§8.4) — DONE:** id-first resolution; the `heph.nvim` `[[` picker (`search` → insert `[[NODEID|Name]]`, "+ Create" mints a doc) + id-direct follow; the **expand-on-read / collapse-on-write** projection (`heph-core::wikilink`); **conceal display** (`conceal.lua` hides the id, shows the label as a `HephLink`, reveals on the cursor line); and the one-time **`heph migrate-links`** migration of legacy `[[Name]]``[[NODEID]]`. *(Follow-up tidy, once the migration is run on every store: retire name-resolution + the canonical-context hack — kept for now so legacy links work pre-migration.)* See §8.4.
6. ⏳ **`heph.nvim` slice 11d (§6/§8) — DEFERRED, post-parity:** daemon **server-push** notification framing (no-`id` lines on the socket; the client read-loop already demuxes them) + the **dirty-buffer reconcile** (the §8 "known-hard" case) + the "update-arrives-while-open" e2e (§9).
7. ⏳ **Split `heph.nvim` to its own forge repo (§8) — UX polish:** generated from this monorepo (subtree-split in CI) so the lazy spec becomes `{ "eblume/heph.nvim" }` instead of a local-clone `dir` (see [[install-heph]]).
8. ⏳ **Adoption refinement + multi-tenant (§13) — before v1 done, low priority:** local→authed **adoption** currently rewrites `owner_id` (`adopt_owner`) but not yet the owner-embedded deterministic ids (journal/tag) + their links; and the hub is single-tenant (one owner per store) — owner-per-token storage is a future extension.