vole

Clone

Read only, no credentials:

https://rodent.software/vole.git

Write over SSH with a registered key:

git clone ssh://git@rodent.software:2222/vole.git

A free and open-source forge that lives in your Git repo.

Index the rest of the metadata branch

7f547ca3b8da218249120d3c910c73c9a1a1e52d · Keenan Nemetz · 10d ago · parent c9202f7f

Index the rest of the metadata branch

The index held issues and their numbers. Everything else on the branch —
change requests, their comments, approvals and closes, and release
artifacts — was read from git on every page that needed it, and the
repository header read the whole requests subtree just to print a count.

Add the tables and derive them from the same commit:

- requests, with the body preserved for search;
- request_closes and request_approvals, each in its own table so a
  backlink and a current-approval count are queries rather than a walk of
  every request;
- comments for both kinds of entity, with the anchor a review comment
  carries;
- artifacts, the release pointers, with the size from the pointer.

The refresh is still driven by the path delta between two commits: a
changed issue or request replaces its own rows, its comments and its sets,
and a deleted one removes them while keeping its `#42` number. Rebuild and
the first read go through IndexRepo, which is the definition the
incremental path has to reproduce; a rebuild now clears the derived tables
first so a deleted repository leaves nothing behind.

Schema version 4 drops the old rows on first open, which is the existing
recovery path: the index is disposable and built from the branch.

forge/index/index.go +91−8

@@ -108,7 +108,75 @@ CREATE TABLE IF NOT EXISTS entity_numbers (
	108	108	id TEXT NOT NULL,
	109	109	PRIMARY KEY (repo_id, number)
	110	110	);
	111		CREATE INDEX IF NOT EXISTS entity_numbers_id ON entity_numbers(repo_id, id);`); err != nil {
		111	CREATE INDEX IF NOT EXISTS entity_numbers_id ON entity_numbers(repo_id, id);
		112
		113	CREATE TABLE IF NOT EXISTS requests (
		114	repo_id TEXT NOT NULL,
		115	id TEXT NOT NULL,
		116	title TEXT NOT NULL,
		117	state TEXT NOT NULL,
		118	author TEXT NOT NULL,
		119	created INTEGER NOT NULL,
		120	base TEXT NOT NULL DEFAULT '',
		121	head TEXT NOT NULL DEFAULT '',
		122	head_ref TEXT NOT NULL DEFAULT '',
		123	merged_base TEXT NOT NULL DEFAULT '',
		124	merged_head TEXT NOT NULL DEFAULT '',
		125	merged_by TEXT NOT NULL DEFAULT '',
		126	merged_at INTEGER NOT NULL DEFAULT 0,
		127	body TEXT NOT NULL DEFAULT '',
		128	comments INTEGER NOT NULL DEFAULT 0,
		129	PRIMARY KEY (repo_id, id)
		130	);
		131	CREATE TABLE IF NOT EXISTS request_closes (
		132	repo_id TEXT NOT NULL,
		133	request_id TEXT NOT NULL,
		134	issue_id TEXT NOT NULL,
		135	PRIMARY KEY (repo_id, request_id, issue_id)
		136	);
		137	CREATE INDEX IF NOT EXISTS request_closes_issue ON request_closes(repo_id, issue_id);
		138	CREATE TABLE IF NOT EXISTS request_approvals (
		139	repo_id TEXT NOT NULL,
		140	request_id TEXT NOT NULL,
		141	principal TEXT NOT NULL,
		142	head TEXT NOT NULL,
		143	base TEXT NOT NULL,
		144	created INTEGER NOT NULL DEFAULT 0,
		145	PRIMARY KEY (repo_id, request_id, principal)
		146	);
		147
		148	-- comments holds the discussion on both kinds of entity. The entity row carries
		149	-- the count a list shows; these rows are what search reads and what a future
		150	-- thread renderer can page through.
		151	CREATE TABLE IF NOT EXISTS comments (
		152	repo_id TEXT NOT NULL,
		153	kind TEXT NOT NULL,
		154	parent TEXT NOT NULL,
		155	id TEXT NOT NULL,
		156	author TEXT NOT NULL,
		157	created INTEGER NOT NULL,
		158	body TEXT NOT NULL DEFAULT '',
		159	reply_to TEXT NOT NULL DEFAULT '',
		160	deleted INTEGER NOT NULL DEFAULT 0,
		161	path TEXT NOT NULL DEFAULT '',
		162	side TEXT NOT NULL DEFAULT '',
		163	line INTEGER NOT NULL DEFAULT 0,
		164	blob TEXT NOT NULL DEFAULT '',
		165	PRIMARY KEY (repo_id, kind, parent, id)
		166	);
		167	CREATE INDEX IF NOT EXISTS comments_parent ON comments(repo_id, kind, parent);
		168
		169	-- artifacts are the release pointers: releases/<tag>/<name> on the metadata
		170	-- branch. The size is the pointer's, which is the artifact's real size, so a
		171	-- release page lists its files without reading the branch.
		172	CREATE TABLE IF NOT EXISTS artifacts (
		173	repo_id TEXT NOT NULL,
		174	tag TEXT NOT NULL,
		175	name TEXT NOT NULL,
		176	oid TEXT NOT NULL,
		177	size INTEGER NOT NULL DEFAULT 0,
		178	PRIMARY KEY (repo_id, tag, name)
		179	);`); err != nil {
	112	180	db.Close()
	113	181	return nil, err
	114	182	}
@@ -120,11 +188,11 @@ CREATE INDEX IF NOT EXISTS entity_numbers_id ON entity_numbers(repo_id, id);`);
	120	188	}
	121	189
	122	190	// schemaVersion is the shape the rows in this file were derived from. It is not
	123		// a migration chain: the index is disposable, so a bump drops the issue rows and
	124		// leaves every repo with an empty issues_sha, which makes the next read rebuild
	125		// that repo from git. A stale row is worse than a rebuild, and a rebuild is the
		191	// a migration chain: the index is disposable, so a bump drops the derived rows
		192	// and leaves every repo with an empty issues_sha, which makes the next read
		193	// rebuild that repo from git. A stale row is worse than a rebuild, and a rebuild is the
	126	194	// operation the index already has to survive.
	127		const schemaVersion = 3
		195	const schemaVersion = 4
	128	196
	129	197	func (s *Store) migrate() error {
	130	198	var v int
@@ -139,6 +207,11 @@ DELETE FROM issues;
	139	207	DELETE FROM issue_labels;
	140	208	DELETE FROM issue_assignees;
	141	209	DELETE FROM entity_numbers;
		210	DELETE FROM requests;
		211	DELETE FROM request_closes;
		212	DELETE FROM request_approvals;
		213	DELETE FROM comments;
		214	DELETE FROM artifacts;
	142	215	UPDATE repos SET issues_sha = '';`); err != nil {
	143	216	return err
	144	217	}
@@ -179,6 +252,16 @@ func (s Store) Rebuild(st repo.Store) (int, error) {
	179	252	if _, err := tx.Exec(`DELETE FROM repos`); err != nil {
	180	253	return 0, err
	181	254	}
		255	// A rebuild is a replacement, not a merge: rows for a repository that no
		256	// longer exists must not survive it.
		257	for _, table := range []string{
		258	"issues", "issue_labels", "issue_assignees", "requests", "request_closes",
		259	"request_approvals", "comments", "artifacts", "entity_numbers",
		260	} {
		261	if _, err := tx.Exec(`DELETE FROM ` + table); err != nil {
		262	return 0, err
		263	}
		264	}
	182	265	for _, m := range repos {
	183	266	e, err := entryFor(m)
	184	267	if err != nil {
@@ -191,10 +274,10 @@ func (s Store) Rebuild(st repo.Store) (int, error) {
	191	274	if err := tx.Commit(); err != nil {
	192	275	return 0, err
	193	276	}
	194		// Issue rows are indexed per repo after the repo rows exist, because the
	195		// issues table is keyed by repo id.
		277	// Derived rows are indexed per repo after the repo rows exist, because every
		278	// table is keyed by repo id.
	196	279	for _, m := range repos {
	197		if err := s.IndexIssues(m); err != nil {
		280	if err := s.IndexRepo(m); err != nil {
	198	281	return 0, err
	199	282	}
	200	283	}