Compare commits

...

31 Commits
0.1.14 ... main

Author SHA1 Message Date
flarum-bot bf19eed20f Bundled output for commit 2fc66b9bf7
Includes transpiled JS/TS, and Typescript declaration files (typings).

[skip ci]
2026-04-13 20:22:53 +00:00
Bart van Bragt 2fc66b9bf7 perf: eager-load relationships in UpdateSearchJob and cache permissions in DiscussionSeeder
- Add relationships() to Seeder base class; DiscussionSeeder implements it to
  return tags/recipientUsers/recipientGroups for eager loading
- UpdateSearchJob calls loadMissing(relationships()) before the map loop,
  eliminating N×R lazy-load queries per batch
- Remove refresh:true from bulk call — synchronous ES refresh was the primary
  throughput bottleneck during indexing
- Cache viewForum permissions in DiscussionSeeder per job instance, replacing
  N×2 Permission queries per document with a single cached query
- Increase seeder batch size from 1000 to 2500
2026-04-13 22:18:14 +02:00
Bart van Bragt 439c74f0a3 fix: store index compat version in ES mapping _meta for correct rollback behaviour
- Write INDEX_COMPAT_VERSION into mapping _meta.index_compat_version on every
  build/promote so the version travels with the index rather than being a
  separate settings value
- saveIndexedConfig now reads compat version back from _meta alongside the
  existing analyzer/min_gram sync, so rollback automatically restores the
  correct value (or null for pre-feature indexes, triggering the reindex warning)
- Replace InvalidArgumentException in buildIndexSettings with $this->error()
  fallback to avoid stack traces in the console
- Fix stale comment in ViewsSearchJob referencing removed CLI flags
2026-04-10 13:46:08 +02:00
Bart van Bragt f486292f15 feat: make ngram min_gram configurable via admin setting
Adds a 'Minimum search query length' setting (1–4, default 3). The admin
shows a warning when the analyzer or min-search-length have changed since
the last index build.
2026-04-10 13:19:22 +02:00
Bart van Bragt 4412583cd9 chore: fix prettier formatting and npm audit vulnerabilities
Reformat admin/index.ts to satisfy prettier checks. Update package-lock.json
via npm audit fix to resolve 17 vulnerabilities in transitive dev dependencies
(webpack, babel, lodash, ws, semver, etc.).
2026-04-10 11:47:38 +02:00
Bart van Bragt 81e7b6502d feat: Show admin dashboard notice when search index needs rebuilding
After a mapping change (e.g. the parent-child restructure), existing installs
need to rebuild their index. A DashboardWidget appears on the admin page
warning the admin and showing the command to run. Compatibility is tracked
via the blomstra-search.index-compatible setting.
2026-04-10 11:10:53 +02:00
Bart van Bragt 3039b65946 refactor: Simplify index command with better naming/options
The flag-based interface (--rebuild etc) became rather convoluted with the changes. Replace this with commands with more straight forward naming.

Also added an explicit 'rollback' action and updated the README.md
2026-04-10 09:35:49 +02:00
Bart van Bragt e2e2d1428f feat: restructure index as parent-child documents with blue-green rebuilds
- Switch to ES parent/child join_field: discussion docs hold all metadata,
  post docs hold content only. Routing ensures parent and child land on the
  same shard. Removes the old flat-document approach where post data was
  duplicated onto every comment document.

- Add HasChildQuery with inner_hits so the best-matching post ID is
  surfaced as mostRelevantPost without a second DB query.

- Blue-green index rebuilds: --recreate writes into a timestamped pending
  index, --swap atomically promotes it via alias. Interrupted builds are
  resumable with --recreate --continue. --swap requires confirmation before
  proceeding.

- Index hidden posts; non-moderators are filtered at query time via
  is_hidden on the has_child clause. Core\Hidden and Core\Restored trigger
  re-indexing so moderators with post.hide can search hidden posts.
2026-04-09 23:30:57 +02:00
Bart van Bragt 80d66c98c4 feat: index and sync view_count for fof/discussion-views sort support
- Rename SavingJob to UpdateSearchJob
- Add ViewsSearchJob for lightweight partial view_count updates
- Add probabilistic sync throttle in DiscussionSeeder::viewingOn()
  to avoid an ES update on every page view
- Add view_count to mapping, DiscussionSeeder::toDocument(), and
  translateSort
- Log a warning and skip instead of 500ing on unknown sort fields
2026-04-09 12:05:28 +02:00
Bart van Bragt ff2a252e99 fix: editing a post does not update the search index
CommentSeeder only listened to Post\Event\Posted, so post edits
left the Elasticsearch document stale.
2026-04-09 11:17:17 +02:00
Bart van Bragt 5929a191f6 fix: --seed-missing stops early when a full range is already indexed
When all documents in a range were already in the index, the empty
collection caused continueAt to be set to null, aborting before
scanning lower ID ranges. Now advances past the range bottom instead.
2026-04-09 11:17:17 +02:00
Bart van Bragt 6699c8cb58 fix: ngram min_gram=1 indexes tokens that can never be matched
Flarum's search UI enforces a minimum query length of 3 characters
(MIN_SEARCH_LEN in Search.tsx), so 1- and 2-character ngram tokens
are indexed but can never be queried. Changed min_gram from 1 to 3
and updated max_ngram_diff accordingly.
2026-04-09 11:04:27 +02:00
Bart van Bragt 369ee504f3 fix: remove unused content_partial field from seeders
content_partial was introduced alongside a partialMatch() query that
targeted it. That query was later switched to use the content field
instead, but content_partial was never removed from the seeders,
causing every document to store and index its content twice.
2026-04-09 08:39:58 +02:00
Bart van Bragt eeb091273d fix: tag filter produces empty results when browsing a tag
When navigating to /t/{tag}?q=query the tag filter from the request
was silently ignored. Added a TermsQuery on the tags field using the
tag IDs resolved from the provided slugs.
2026-04-09 08:20:42 +02:00
Daniël Klabbers d252a17586 fix: not changing recipients visibility when recipients change 2024-06-13 19:27:14 +02:00
Daniël Klabbers a9dc8bfb16 fix: byobu discussions open up while using this extension 2024-06-12 11:34:18 +02:00
StyleCI Bot 82e6ec786f
Apply fixes from StyleCI 2023-10-18 13:57:41 +00:00
flarum-bot f981c27e00 Bundled output for commit f6503f5e6a
Includes transpiled JS/TS, and Typescript declaration files (typings).

[skip ci]
2023-01-30 22:17:40 +00:00
Daniël Klabbers f6503f5e6a fix: #10, private forums have no discussions key, throws error 2023-01-30 23:17:01 +01:00
flarum-bot ad92d8856b Bundled output for commit 8b519619e5
Includes transpiled JS/TS, and Typescript declaration files (typings).

[skip ci]
2022-11-24 09:39:12 +00:00
StyleCI Bot 8b519619e5
Apply fixes from StyleCI 2022-11-24 09:38:21 +00:00
Daniël Klabbers 83afe95d61 chore: fix search and seeding 2022-11-24 10:37:57 +01:00
Daniël Klabbers 846cc348ff Merge branch 'dk/search-improvements' into main 2022-11-24 10:29:18 +01:00
Daniël Klabbers 508bb42ac9 chore: clean up buildcommand 2022-11-23 12:46:04 +01:00
StyleCI Bot db74da3e37
Apply fixes from StyleCI 2022-11-23 11:29:48 +00:00
Daniël Klabbers a02d35394b feat: store tags, rawId, support indexing missing objects 2022-11-23 12:09:10 +01:00
flarum-bot b60563f2ec Bundled output for commit 703cbbdd48
Includes transpiled JS/TS, and Typescript declaration files (typings).

[skip ci]
2022-11-21 15:33:18 +00:00
David Wheatley 703cbbdd48
fix: incorrect JS package name 2022-11-21 15:32:22 +00:00
Daniël Klabbers 9401b677a3 change weighing for exact words match 2022-11-10 12:57:46 +01:00
Daniël Klabbers 4bc89949ec complete testing 2022-11-10 12:54:26 +01:00
Daniël Klabbers 29ba864849 settings configurations 2022-11-09 18:06:48 +01:00
27 changed files with 2455 additions and 2025 deletions

143
README.md
View File

@ -5,47 +5,154 @@ fulltext search with one that is completely relying on the proven elasticsearch
## Features
- Sync discussions to elastic search using your queue, unobtrusively for the user.
- Reduces search loading times to well below 400ms (local tests with 50.000 discussion **260ms**)
- Sync discussions and posts to Elasticsearch using your queue, unobtrusively for the user.
- Reduces search loading times to well below 400ms (local tests with 50,000 discussions: **260ms**)
- Uses Flarum's group permissions and tags system.
- Compatible with Friends of Flarum Byōbu.
## Installation
## Requirements
Use composer:
- Elasticsearch 7.x or OpenSearch 1.x+
- A non-sync queue driver with a running worker (`php flarum queue:work`) is strongly recommended for production. The extension works with the default sync driver, but index jobs run inline which adds latency to user-facing changes like posting.
## Installation
```bash
composer require blomstra/search:*
```
Enable the extension inside the admin area and configure the settings.
Enable the extension in the admin area and configure the Elasticsearch endpoint and index name in the extension settings.
### Set up
## Setting up the index
Enable the extension in your admin area. Now to seed your existing discussions use the following command:
### First install
```
php flarum blomstra:search:index
Run the build command once. It creates a timestamped concrete index, immediately aliases your configured index name to it, and begins queuing documents. Search is available and improves as the queue processes:
```bash
php flarum blomstra:search:index build
php flarum queue:work
```
All mutations to discussions are automatically added and removed from the elasticsearch index.
### Subsequent rebuilds (zero-downtime)
### FAQ
When you need to rebuild the full index (e.g. after a mapping change):
*I have another question.*
Reach out to us via https://helpdesk.blomstra.net. We will get back to you as soon as we can. If you have a running subscription please mention when you started your plan and/or which plan you are on. Always add sufficient information when reporting errors. We prefer errors being reported here, but understand that sometimes you can't.
```bash
# Simple rebuild — promotes automatically once all jobs are queued
php flarum blomstra:search:index build
# Or keep a backup of the old index in case you need to roll back
php flarum blomstra:search:index build --keep-backup
```
After the queue drains, fill any gaps from content posted during the build:
```bash
php flarum blomstra:search:index fill
```
If you kept a backup and want to roll back:
```bash
php flarum blomstra:search:index rollback
```
Once satisfied with the new index, drop the backup:
```bash
php flarum blomstra:search:index discard --backup
```
### Blue-green rebuild (manual promotion)
Use `--staging` to keep the old index live until you explicitly promote:
```bash
# 1. Build into a staging index — live index is untouched
php flarum blomstra:search:index build --staging
# 2. Drain the queue
php flarum queue:work --stop-when-empty
# 3. Promote the staging index to live
php flarum blomstra:search:index promote
# Or keep the old index as a backup:
php flarum blomstra:search:index promote --keep-backup
```
### Resuming or cancelling an interrupted build
If a build is interrupted, re-run it with the appropriate flag:
```bash
# Resume each seeder from where it left off
php flarum blomstra:search:index build --resume
# Drop the staging index and start completely fresh
php flarum blomstra:search:index build --fresh
# Cancel the build without starting a new one
php flarum blomstra:search:index discard --pending
```
### Filling gaps in an existing index
If documents are missing from the live index (e.g. due to queue failures):
```bash
php flarum blomstra:search:index fill
```
### Updating the mapping only
To push a mapping change to the live index without rebuilding:
```bash
php flarum blomstra:search:index mapping
```
## Command reference
| Command | Description |
|---|---|
| `build` | Rebuild the index and promote automatically once all jobs are queued. On first install, aliases immediately so search is live during seeding. |
| `build --keep-backup` | Rebuild and promote, retaining the old index as a backup for rollback. |
| `build --staging` | Build into a staging index without promoting — use `promote` when ready (blue-green workflow). |
| `build --resume` | Resume an interrupted build from where each seeder left off. |
| `build --fresh` | Drop the staging index and start completely fresh. |
| `promote` | Atomically swap the alias to the staging index. Prompts for confirmation (blue-green workflow). |
| `promote --keep-backup` | Promote and retain the replaced live index as a backup for rollback. |
| `rollback` | Restore the backup index to live. Deletes the index that was live. |
| `discard --pending` | Drop the staging index without promoting (cancels an in-progress build). |
| `discard --backup` | Drop the backup index (cleanup after `--keep-backup`). |
| `mapping` | Push updated mapping to the live index without rebuilding or reseeding. |
| `fill` | Seed only documents missing from the live index. |
| `build --only=discussions` | Seed only the specified document type (`discussions` or `posts`). |
| `build --throttle=N` | Wait N seconds between batches (reduces queue pressure). |
| `build --max-id=N` | Limit seeding to documents with ID ≤ N. |
| `promote --i-am-sure` | Skip the promotion confirmation prompt (for scripts and CI). |
## FAQ
## Queue configuration
*"Can I dispatch indexing jobs to a specific queue?"*
*Can I dispatch the sync jobs to another queue?*
Yes:
```php
\Blomstra\Search\Observe\Job::$onQueue = 'sync';
\Blomstra\Search\Jobs\Job::$onQueue = 'search';
```
*"I have a different question"*
Reach out ot us via https://support.on-floxum.com/t/ext-search . If you have an active subscription, please mention what plan you are on.
---
- Blomstra provides managed Flarum hosting.
- https://blomstra.net
- https://blomstra.community/t/ext-search
- Floxum provides managed Flarum hosting.
- https://floxum.com
- https://support.on-floxum.com/t/ext-search
Icon made by [Freepik](https://www.freepik.com) from [Flaticon](https://www.flaticon.com/).

View File

@ -20,14 +20,19 @@ return [
(new Flarum\Frontend('forum'))
->js(__DIR__.'/js/dist/forum.js'),
(new Flarum\Frontend('admin'))
->js(__DIR__.'/js/dist/admin.js'),
->js(__DIR__.'/js/dist/admin.js')
->css(__DIR__.'/resources/less/admin.less'),
(new Flarum\Locales(__DIR__.'/resources/locale')),
new Flarum\Locales(__DIR__.'/resources/locale'),
(new Flarum\Routes('api'))
->get('/blomstra/search/{type}', 'blomstra.search', Api\Controllers\SearchController::class)
->put('/blomstra/search/index', 'blomstra.search.index', Api\Controllers\IndexController::class),
(new Flarum\Console())
->command(Commands\BuildCommand::class),
(new Flarum\Settings())
->default('blomstra-search.search-discussion-subjects', true)
->default('blomstra-search.search-post-bodies', true)
->default('blomstra-search.min-search-length', Commands\BuildCommand::DEFAULT_MIN_SEARCH_LENGTH),
];

2
js/dist/admin.js vendored
View File

@ -1,2 +1,2 @@
(()=>{var a={n:e=>{var t=e&&e.__esModule?()=>e.default:()=>e;return a.d(t,{a:t}),t},d:(e,t)=>{for(var r in t)a.o(t,r)&&!a.o(e,r)&&Object.defineProperty(e,r,{enumerable:!0,get:t[r]})},o:(a,e)=>Object.prototype.hasOwnProperty.call(a,e),r:a=>{"undefined"!=typeof Symbol&&Symbol.toStringTag&&Object.defineProperty(a,Symbol.toStringTag,{value:"Module"}),Object.defineProperty(a,"__esModule",{value:!0})}},e={};(()=>{"use strict";a.r(e);const t=flarum.core.compat["admin/app"];var r=a.n(t);r().initializers.add("blomstra-search",(function(){var a=new Map;["arabic","armenian","basque","bengali","brazilian","bulgarian","catalan","cjk","czech","danish","dutch","english","estonian","finnish","french","galician","german","greek","hindi","hungarian","indonesian","irish","italian","latvian","lithuanian","norwegian","persian","portuguese","romanian","russian","sorani","spanish","swedish","turkish","thai"].forEach((function(e){a.set(e,e)})),r().extensionData.for("blomstra-search").registerSetting({setting:"blomstra-search.elastic-endpoint",label:r().translator.trans("blomstra-search.admin.elastic-endpoint"),type:"input"}).registerSetting({setting:"blomstra-search.elastic-username",label:r().translator.trans("blomstra-search.admin.elastic-username"),type:"input"}).registerSetting({setting:"blomstra-search.elastic-password",label:r().translator.trans("blomstra-search.admin.elastic-password"),type:"password"}).registerSetting({setting:"blomstra-search.elastic-index",label:r().translator.trans("blomstra-search.admin.elastic-index"),default:"flarum",type:"input"}).registerSetting({setting:"blomstra-search.analyzer-language",label:r().translator.trans("blomstra-search.admin.analyzer.label"),help:r().translator.trans("blomstra-search.admin.analyzer.help"),type:"select",options:Object.fromEntries(a.entries()),default:"english"}).registerSetting({setting:"blomstra-search.elastic-index",label:r().translator.trans("blomstra-search.admin.elastic-index"),default:"flarum",type:"input"})}))})(),module.exports=e})();
(()=>{var t={n:a=>{var e=a&&a.__esModule?()=>a.default:()=>a;return t.d(e,{a:e}),e},d:(a,e)=>{for(var r in e)t.o(e,r)&&!t.o(a,r)&&Object.defineProperty(a,r,{enumerable:!0,get:e[r]})},o:(t,a)=>Object.prototype.hasOwnProperty.call(t,a)};(()=>{"use strict";function a(t,e){return a=Object.setPrototypeOf?Object.setPrototypeOf.bind():function(t,a){return t.__proto__=a,t},a(t,e)}const e=flarum.core.compat["admin/app"];var r=t.n(e);const n=flarum.core.compat["common/extend"],s=flarum.core.compat["admin/components/DashboardPage"];var i=t.n(s);const o=flarum.core.compat["admin/components/DashboardWidget"];var l=t.n(o);const c=flarum.core.compat["common/components/Alert"];var h=t.n(c),d=function(t){var e,n;function s(){return t.apply(this,arguments)||this}n=t,(e=s).prototype=Object.create(n.prototype),e.prototype.constructor=e,a(e,n);var i=s.prototype;return i.className=function(){return"ReindexWarningWidget"},i.content=function(){return m(h(),{type:"warning",dismissible:!1,icon:"fas fa-exclamation-triangle",title:r().translator.trans("blomstra-search.admin.reindex-required.title")},r().translator.trans("blomstra-search.admin.reindex-required.detail"))},s}(l());r().initializers.add("blomstra-search",function(){var t=r().data.settings["blomstra-search.active-index"],a=r().data.settings["blomstra-search.index-compatible"];t&&"v2"!==a&&(0,n.extend)(i().prototype,"availableWidgets",function(t){t.add("blomstra-search-reindex",m(d),110)});var e=new Map;["arabic","armenian","basque","bengali","brazilian","bulgarian","catalan","cjk","czech","danish","dutch","english","estonian","finnish","french","galician","german","greek","hindi","hungarian","indonesian","irish","italian","latvian","lithuanian","norwegian","persian","portuguese","romanian","russian","sorani","spanish","swedish","turkish","thai"].forEach(function(t){e.set(t,t)}),r().extensionData.for("blomstra-search").registerSetting(function(){var a=r().data.settings["blomstra-search.indexed-analyzer"];if(!t||!a)return null;var e=this.setting("blomstra-search.analyzer-language")()||"english",n=this.setting("blomstra-search.min-search-length")(),s=String(r().data.settings["blomstra-search.indexed-min-search-length"]||r().data.settings["blomstra-search.min-search-length"]);return e===a&&n===s?null:m(h(),{type:"warning",dismissible:!1,icon:"fas fa-exclamation-triangle"},r().translator.trans("blomstra-search.admin.index-settings-changed"))}).registerSetting({setting:"blomstra-search.elastic-endpoint",label:r().translator.trans("blomstra-search.admin.elastic-endpoint"),type:"input"}).registerSetting({setting:"blomstra-search.elastic-username",label:r().translator.trans("blomstra-search.admin.elastic-username"),type:"input"}).registerSetting({setting:"blomstra-search.elastic-password",label:r().translator.trans("blomstra-search.admin.elastic-password"),type:"password"}).registerSetting({setting:"blomstra-search.elastic-index",label:r().translator.trans("blomstra-search.admin.elastic-index"),default:"flarum",type:"input"}).registerSetting({setting:"blomstra-search.analyzer-language",label:r().translator.trans("blomstra-search.admin.analyzer.label"),help:r().translator.trans("blomstra-search.admin.analyzer.help"),type:"select",options:Object.fromEntries(e.entries()),default:"english"}).registerSetting({setting:"blomstra-search.search-discussion-subjects",label:r().translator.trans("blomstra-search.admin.search-discussion-subjects"),type:"switch"}).registerSetting({setting:"blomstra-search.search-post-bodies",label:r().translator.trans("blomstra-search.admin.search-post-bodies"),type:"switch"}).registerSetting({setting:"blomstra-search.min-search-length",label:r().translator.trans("blomstra-search.admin.min-search-length.label"),help:r().translator.trans("blomstra-search.admin.min-search-length.help"),type:"select",options:{1:"1",2:"2",3:"3",4:"4"},default:r().data.settings["blomstra-search.min-search-length"]})})})(),module.exports={}})();
//# sourceMappingURL=admin.js.map

File diff suppressed because one or more lines are too long

3
js/dist/forum.js vendored

File diff suppressed because one or more lines are too long

File diff suppressed because one or more lines are too long

2752
js/package-lock.json generated

File diff suppressed because it is too large Load Diff

View File

@ -1,16 +1,14 @@
{
"name": "@blomstra/realtime",
"name": "@blomstra/search",
"private": true,
"version": "0.0.0",
"prettier": "@flarum/prettier-config",
"dependencies": {
"@flarum/prettier-config": "^1.0.0",
"build": "^0.1.4",
"flarum-tsconfig": "^1.0.2",
"flarum-webpack-config": "^2.0.0",
"pusher-js": "^7.0.3",
"webpack": "^5.75.0",
"webpack-cli": "^4.10.0"
"webpack-cli": "^5.0.0"
},
"scripts": {
"dev": "webpack --mode development --watch",

View File

@ -1,6 +1,40 @@
import app from 'flarum/admin/app';
import { extend } from 'flarum/common/extend';
import DashboardPage from 'flarum/admin/components/DashboardPage';
import DashboardWidget from 'flarum/admin/components/DashboardWidget';
import Alert from 'flarum/common/components/Alert';
const REQUIRED_INDEX_COMPAT = 'v2';
class ReindexWarningWidget extends DashboardWidget {
className() {
return 'ReindexWarningWidget';
}
content() {
return m(
Alert,
{
type: 'warning',
dismissible: false,
icon: 'fas fa-exclamation-triangle',
title: app.translator.trans('blomstra-search.admin.reindex-required.title'),
},
app.translator.trans('blomstra-search.admin.reindex-required.detail')
);
}
}
app.initializers.add('blomstra-search', () => {
const activeIndex = app.data.settings['blomstra-search.active-index'];
const compatVersion = app.data.settings['blomstra-search.index-compatible'];
if (activeIndex && compatVersion !== REQUIRED_INDEX_COMPAT) {
extend(DashboardPage.prototype, 'availableWidgets', function (items) {
items.add('blomstra-search-reindex', m(ReindexWarningWidget), 110);
});
}
const languages = new Map();
[
'arabic',
@ -44,6 +78,25 @@ app.initializers.add('blomstra-search', () => {
app.extensionData
.for('blomstra-search')
.registerSetting(function (this: any) {
// No index yet, or first build predates this tracking — stay silent.
const indexedAnalyzer = app.data.settings['blomstra-search.indexed-analyzer'];
if (!activeIndex || !indexedAnalyzer) return null;
const currentAnalyzer = this.setting('blomstra-search.analyzer-language')() || 'english';
const currentMinLength = this.setting('blomstra-search.min-search-length')();
const indexedMinLength = String(
app.data.settings['blomstra-search.indexed-min-search-length'] || app.data.settings['blomstra-search.min-search-length']
);
if (currentAnalyzer === indexedAnalyzer && currentMinLength === indexedMinLength) return null;
return m(
Alert,
{ type: 'warning', dismissible: false, icon: 'fas fa-exclamation-triangle' },
app.translator.trans('blomstra-search.admin.index-settings-changed')
);
})
.registerSetting({
setting: 'blomstra-search.elastic-endpoint',
label: app.translator.trans('blomstra-search.admin.elastic-endpoint'),
@ -74,9 +127,21 @@ app.initializers.add('blomstra-search', () => {
default: 'english',
})
.registerSetting({
setting: 'blomstra-search.elastic-index',
label: app.translator.trans('blomstra-search.admin.elastic-index'),
default: 'flarum',
type: 'input',
setting: 'blomstra-search.search-discussion-subjects',
label: app.translator.trans('blomstra-search.admin.search-discussion-subjects'),
type: 'switch',
})
.registerSetting({
setting: 'blomstra-search.search-post-bodies',
label: app.translator.trans('blomstra-search.admin.search-post-bodies'),
type: 'switch',
})
.registerSetting({
setting: 'blomstra-search.min-search-length',
label: app.translator.trans('blomstra-search.admin.min-search-length.label'),
help: app.translator.trans('blomstra-search.admin.min-search-length.help'),
type: 'select',
options: { '1': '1', '2': '2', '3': '3', '4': '4' },
default: app.data.settings['blomstra-search.min-search-length'],
});
});

View File

@ -9,8 +9,17 @@ import DiscussionsSearchSource from './SearchSources/DiscussionsSearchSource';
import extendDiscussionState from './PaginatedListStates/extendDiscussionState';
app.initializers.add('blomstra-search', () => {
const minLength = parseInt(app.data.settings['blomstra-search.min-search-length'] || String(Search.MIN_SEARCH_LEN), 10);
if (minLength !== Search.MIN_SEARCH_LEN) {
// Flarum provides no extension point for MIN_SEARCH_LEN, so we overwrite the
// static property directly. TypeScript `readonly` is compile-time only — at
// runtime this is a plain property assignment and is safe as long as no code
// reads MIN_SEARCH_LEN before this initializer runs.
(Search as any).MIN_SEARCH_LEN = minLength;
}
extend(Search.prototype, 'sourceItems', function (this: Search<SearchAttrs>, items: ItemList<SearchSource>) {
items.setContent('discussions', new DiscussionsSearchSource());
items.replace('discussions', new DiscussionsSearchSource());
});
});

View File

@ -0,0 +1,3 @@
.ReindexWarningWidget {
padding: 0;
}

View File

@ -8,4 +8,18 @@ blomstra-search:
label: Analyzer language
help: |
The analyzer makes search understand stop words and undertakes language
specific improvements for indexing.
specific improvements for indexing. Changing this requires a full index rebuild to take effect.
index-settings-changed: "Search index settings have changed. Run: php flarum blomstra:search:index build"
reindex-required:
title: Search index rebuild required
detail: "The search index is not compatible with this version of the extension. Run: php flarum blomstra:search:index build"
search-discussion-subjects: Search inside discussion titles
search-post-bodies: Search inside comments
match-sentences: Match search term against full sentence
match-words: Match search term against full words
min-search-length:
label: Minimum search query length
help: |
Minimum number of characters required before a search is triggered. Lower this to 1 or 2
for CJK (Chinese, Japanese, Korean) communities where single characters carry full meaning.
Changing this requires a full index rebuild to take effect.

View File

@ -12,10 +12,13 @@
namespace Blomstra\Search\Api\Controllers;
use Blomstra\Search\Elasticsearch\HasChildQuery;
use Blomstra\Search\Elasticsearch\MatchPhraseQuery;
use Blomstra\Search\Elasticsearch\MatchQuery;
use Blomstra\Search\Elasticsearch\TermsQuery;
use Blomstra\Search\Save\Document as ElasticDocument;
use Blomstra\Search\Searchers\CommentPostSearcher;
use Blomstra\Search\Searchers\DiscussionSearcher;
use Blomstra\Search\Searchers\Searcher;
use Elasticsearch\Client;
use Flarum\Api\Controller\ListDiscussionsController;
use Flarum\Api\Serializer\DiscussionSerializer;
@ -24,15 +27,17 @@ use Flarum\Extension\ExtensionManager;
use Flarum\Group\Group;
use Flarum\Http\RequestUtil;
use Flarum\Http\UrlGenerator;
use Flarum\Settings\SettingsRepositoryInterface;
use Flarum\Tags\Tag;
use Flarum\User\User;
use Illuminate\Contracts\Container\Container;
use Illuminate\Support\Arr;
use Illuminate\Support\Collection;
use Illuminate\Support\Str;
use Psr\Http\Message\ServerRequestInterface;
use Psr\Log\LoggerInterface;
use Spatie\ElasticsearchQueryBuilder\Builder;
use Spatie\ElasticsearchQueryBuilder\Queries\BoolQuery;
use Spatie\ElasticsearchQueryBuilder\Queries\Query;
use Spatie\ElasticsearchQueryBuilder\Queries\TermQuery;
use Spatie\ElasticsearchQueryBuilder\Sorts\Sort;
use Tobscure\JsonApi\Document;
@ -45,92 +50,103 @@ class SearchController extends ListDiscussionsController
'lastPostedAt' => 'updated_at',
'createdAt' => 'created_at',
'commentCount' => 'comment_count',
'view_count' => 'view_count',
];
public function __construct(protected Client $elastic, protected UrlGenerator $uri)
protected bool $matchSentences;
protected bool $matchWords;
protected ?Searcher $discussionSearcher;
protected ?Searcher $postSearcher;
public function __construct(protected Client $elastic, protected UrlGenerator $uri, Container $container, SettingsRepositoryInterface $settings)
{
$this->matchSentences = true;
$this->matchWords = true;
$searchers = collect($container->tagged('blomstra.search.searchers'));
$this->discussionSearcher = $searchers->first(fn ($s) => $s instanceof DiscussionSearcher);
$this->postSearcher = $searchers->first(fn ($s) => $s instanceof CommentPostSearcher);
}
protected function data(ServerRequestInterface $request, Document $document)
{
// Not used for now.
$type = Arr::get($request->getQueryParams(), 'type');
$actor = RequestUtil::getActor($request);
$filters = $this->extractFilter($request);
$search = $this->getSearch($filters);
$limit = $this->extractLimit($request);
$offset = $this->extractOffset($request);
$include = array_merge($this->extractInclude($request), ['state']);
$filterQuery = BoolQuery::create();
$query = BoolQuery::create()
// Always restrict to discussion documents; posts are only searched via has_child.
->add(TermQuery::create('join_field', 'discussion'), 'filter');
if (!empty($search)) {
$filterQuery
// @todo commented out to use only partial matching for now
->add($this->sentenceMatch($search))
->add($this->wordMatch($search, 'and'))
->add($this->wordMatch($search, 'or'))
// ->add($this->partialMatch($search))
;
$query->add($this->buildTextQuery($search, $actor));
}
$this->addFilters($query, $actor, $filters);
$builder = (new Builder($this->elastic))
->index(resolve('blomstra.search.elastic_index'))
->size($limit + 1)
->from($this->extractOffset($request))
->addQuery(
$this->addFilters($filterQuery, $actor, $filters)
);
->from($offset)
->addQuery($query);
$knownSortFields = array_merge(array_values($this->translateSort), ['rawId']);
$logger = resolve(LoggerInterface::class);
$phpSortField = null;
$phpSortDir = 'desc';
foreach ($this->extractSort($request) as $field => $direction) {
$field = $this->translateSort[$field] ?? $field;
$builder->addSort(new Sort($field, $direction));
$translated = $this->translateSort[$field] ?? $field;
if (!in_array($translated, $knownSortFields)) {
$logger->warning("blomstra/search: unknown sort field \"{$field}\", ignoring.");
continue;
}
$builder->addSort(new Sort($translated, $direction));
if ($phpSortField === null && $translated !== 'rawId') {
$phpSortField = $translated;
$phpSortDir = $direction;
}
}
$response = $builder->search();
Discussion::setStateUser($actor);
// Eager load groups for use in the policies (isAdmin check)
if (in_array('mostRelevantPost.user', $include)) {
$include[] = 'mostRelevantPost.user.groups';
// If the first level of the relationship wasn't explicitly included,
// add it so the code below can look for it
if (!in_array('mostRelevantPost', $include)) {
$include[] = 'mostRelevantPost';
}
}
// we need to retrieve all discussion ids and when the results are posts,
// their ids as most relevant post id
// All hits are discussion documents. Extract the best-matching post ID from
// inner_hits when a has_child clause matched (i.e. the match came from a post).
$results = Collection::make(Arr::get($response, 'hits.hits'))
->map(function ($hit) {
$type = $hit['_source']['type'];
$id = Str::after($hit['_source']['id'], "$type:");
// _id is "discussions:123" — parse the numeric part directly.
$discussionId = Str::after($hit['_id'], 'discussions:');
// rawId on the inner hit gives us the integer post ID.
$bestPostId = Arr::get($hit, 'inner_hits.best_post.hits.hits.0._source.rawId');
if ($type === 'posts') {
return [
'most_relevant_post_id' => $id,
'weight' => Arr::get($hit, 'sort.0'),
'discussion_id' => $discussionId,
'most_relevant_post_id' => $bestPostId,
'weight' => Arr::get($hit, 'sort.0', Arr::get($hit, '_score', 0)),
];
} else {
return [
'discussion_id' => $id,
'weight' => Arr::get($hit, 'sort.0'),
];
}
});
$document->addPaginationLinks(
$this->uri->to('api')->route('blomstra.search', [
'type' => 'discussions',
]),
$this->uri->to('api')->route('blomstra.search', ['type' => 'discussions']),
$request->getQueryParams(),
$offset,
$limit,
@ -140,28 +156,25 @@ class SearchController extends ListDiscussionsController
$results = $results->take($limit);
$discussions = Discussion::query()
->select('discussions.*')
->join('posts', 'posts.discussion_id', 'discussions.id')
// Extra safety to prevent leaking hidden discussion (titles) towards search results.
->when($actor->isGuest() || !$actor->hasPermission('discussion.hide'), fn ($query) => $query->whereNull('discussions.hidden_at'))
->where(function ($query) use ($results) {
$query
->whereIn('discussions.id', $results->pluck('discussion_id')->filter())
->orWhereIn('posts.id', $results->pluck('most_relevant_post_id')->filter());
})
->when(
$actor->isGuest() || !$actor->hasPermission('discussion.hide'),
fn ($q) => $q->whereNull('hidden_at')
)
->whereIn('id', $results->pluck('discussion_id')->filter())
->get()
->each(function (Discussion $discussion) use ($results) {
if (in_array($discussion->id, $results->pluck('discussion_id')->toArray())) {
$discussion->most_relevant_post_id = $discussion->first_post_id;
$discussion->weight = $results->firstWhere('discussion_id', $discussion->id)['weight'] ?? 0;
} else {
$post = $discussion->posts()->whereIn('id', $results->pluck('most_relevant_post_id'))->first();
$discussion->most_relevant_post_id = $post?->id ?? $discussion->first_post_id;
$discussion->weight = $results->firstWhere('most_relevant_post_id', $post?->id)['weight'] ?? 0;
}
$result = $results->firstWhere('discussion_id', $discussion->id);
$discussion->most_relevant_post_id = $result['most_relevant_post_id']
?? $discussion->first_post_id;
$discussion->weight = $result['weight'] ?? 0;
})
->keyBy('id')
->sortByDesc('weight')
->when(
$phpSortField,
fn ($c) => $phpSortDir === 'desc' ? $c->sortByDesc($phpSortField) : $c->sortBy($phpSortField),
fn ($c) => $c->sortByDesc('weight')
)
->unique();
$this->loadRelations($discussions, $include);
@ -179,33 +192,77 @@ class SearchController extends ListDiscussionsController
return $discussions;
}
protected function getDocument(string $type): ?ElasticDocument
/**
* Build the text-matching portion of the query.
*
* Discussion titles are matched directly (filtered to join_field=discussion).
* Post bodies are matched via has_child with score_mode=sum so discussions
* with many matching posts score higher than those with a single strong match.
* inner_hits returns the best-scoring post for use as mostRelevantPost.
* Hidden posts are only included in matching for users with post.hide permission.
*/
protected function buildTextQuery(string $search, User $actor): BoolQuery
{
$documents = resolve(Container::class)->tagged('blomstra.search.documents');
$textQuery = BoolQuery::create();
return collect($documents)->first(function (ElasticDocument $document) use ($type) {
return $document->type() === $type;
});
if ($this->discussionSearcher?->enabled()) {
$textQuery->add($this->buildShouldClauses($search, $this->discussionSearcher->boost()), 'should');
}
if ($this->postSearcher?->enabled()) {
$postQuery = $this->buildShouldClauses($search, $this->postSearcher->boost());
// Guests and non-moderators may not see hidden posts; exclude them from child matching.
if ($actor->isGuest() || !$actor->hasPermission('post.hide')) {
$postQuery->add(TermQuery::create('is_hidden', false), 'filter');
}
$textQuery->add(
HasChildQuery::create('post', $postQuery)->withInnerHits(),
'should'
);
}
return $textQuery;
}
protected function buildShouldClauses(string $search, float $boost): BoolQuery
{
$should = BoolQuery::create();
if ($this->matchSentences) {
$should->add((new MatchPhraseQuery('content', $search))->boost(2 * $boost), 'should');
}
if ($this->matchWords) {
$should->add((new MatchQuery('content', $search))->operator('and')->boost(1.8 * $boost), 'should');
$should->add((new MatchQuery('content', $search))->operator('or')->boost(0.8 * $boost), 'should');
}
return $should;
}
protected function extensionEnabled(string $extension): bool
{
/** @var ExtensionManager $manager */
$manager = resolve(ExtensionManager::class);
return $manager->isEnabled($extension);
return resolve(ExtensionManager::class)->isEnabled($extension);
}
protected function addFilters(BoolQuery $query, User $actor, array $filters = []): BoolQuery
protected function addFilters(BoolQuery $query, User $actor, array $filters = []): void
{
$groups = $this->getGroups($actor);
$onlyPrivate = Str::contains($filters['q'] ?? '', 'is:private');
$subQuery = BoolQuery::create()
->add(TermQuery::create('is_private', 'false'))
->add(TermsQuery::create('groups', $groups->toArray()));
if ($this->extensionEnabled('flarum-tags') && !empty($filters['tag'])) {
$slugs = is_array($filters['tag']) ? $filters['tag'] : explode(',', $filters['tag']);
$tagIds = Tag::query()->whereIn('slug', $slugs)->pluck('id')->toArray();
if (!empty($tagIds)) {
$query->add(TermsQuery::create('tags', $tagIds), 'filter');
}
}
if ($this->extensionEnabled('fof-byobu') && $actor->exists) {
$byobuQuery = BoolQuery::create()
->add(TermQuery::create('is_private', 'true'))
@ -224,85 +281,12 @@ class SearchController extends ListDiscussionsController
}
}
$query->add(
$subQuery,
'filter'
);
return $query;
}
protected function sentenceMatch(string $q): Query
{
$query = (new MatchPhraseQuery('content', $q));
return BoolQuery::create()
// Discussion titles
->add(
BoolQuery::create()
->add(TermQuery::create('type', 'discussions'), 'filter')
->add($query->boost(2)),
'should'
)
// Post bodies
->add(
BoolQuery::create()
->add(TermQuery::create('type', 'posts'), 'filter')
->add($query->boost(1.9)),
'should'
);
}
protected function wordMatch(string $q, string $operator = 'or')
{
$query = (new MatchQuery('content', $q))
->operator($operator);
$boost = $operator === 'and' ? 1 : .8;
return BoolQuery::create()
// Discussion titles
->add(
BoolQuery::create()
->add(TermQuery::create('type', 'discussions'), 'filter')
->add($query->boost($boost * 1.8)),
'should'
)
// Post bodies
->add(
BoolQuery::create()
->add(TermQuery::create('type', 'posts'), 'filter')
->add($query->boost($boost * 1.8)),
'should'
);
}
protected function partialMatch(string $q)
{
$query = (new MatchQuery('content', $q));
return BoolQuery::create()
// Discussion titles
->add(
BoolQuery::create()
->add(TermQuery::create('type', 'discussions'), 'filter')
->add(clone $query->boost(1.6)),
'should'
)
// Post bodies
->add(
BoolQuery::create()
->add(TermQuery::create('type', 'posts'), 'filter')
->add(clone $query->boost(1.3)),
'should'
);
$query->add($subQuery, 'filter');
}
protected function getGroups(User $actor): Collection
{
/** @var Collection $groups */
$groups = $actor->groups->pluck('id');
$groups->add(Group::GUEST_ID);
if ($actor->is_email_confirmed) {
@ -318,9 +302,7 @@ class SearchController extends ListDiscussionsController
if ($search) {
$q = collect(explode(' ', $search))
->filter(function (string $part) {
return $part !== 'is:private';
})
->filter(fn (string $part) => $part !== 'is:private')
->filter()
->join(' ');

View File

@ -13,7 +13,7 @@
namespace Blomstra\Search\Commands;
use Blomstra\Search\Jobs\Job;
use Blomstra\Search\Jobs\SavingJob;
use Blomstra\Search\Jobs\UpdateSearchJob;
use Blomstra\Search\Seeders\Seeder;
use Elasticsearch\Client;
use Flarum\Settings\SettingsRepositoryInterface;
@ -21,97 +21,314 @@ use Illuminate\Console\Command;
use Illuminate\Contracts\Container\Container;
use Illuminate\Contracts\Queue\Queue;
use Illuminate\Database\Eloquent\Collection;
use Illuminate\Support\Arr;
use Spatie\ElasticsearchQueryBuilder\Builder;
use Spatie\ElasticsearchQueryBuilder\Queries\BoolQuery;
use Spatie\ElasticsearchQueryBuilder\Queries\RangeQuery;
use Spatie\ElasticsearchQueryBuilder\Queries\TermQuery;
class BuildCommand extends Command
{
/**
* Bump this when a mapping change requires a full reindex.
* Written into the mapping's _meta.index_compat_version on every build and
* read back into blomstra-search.index-compatible by saveIndexedConfig so the
* value always reflects the live index including after a promote/rollback.
*/
public const INDEX_COMPAT_VERSION = 'v2';
/** Matches Flarum's Search::MIN_SEARCH_LEN — the default minimum query length. */
public const DEFAULT_MIN_SEARCH_LENGTH = 3;
protected $signature = 'blomstra:search:index
{--max-id= : Limits for each object the number of items to seed}
{--chunk-size= : Size of the chunks to dispatch into jobs}
{--throttle= : Number of seconds to wait between pushing to the queue}
{--only= : type to run seeder for, eg discussions or posts}
{--recreate : create or recreate the index}
{--mapping : recreate the mapping}
{--continue : continue each object type where you left off}';
protected $description = 'Rebuilds the complete search server with its documents.';
{action? : build | promote | rollback | discard | mapping | fill}
{--fresh}
{--resume}
{--staging}
{--keep-backup}
{--pending}
{--backup}
{--only=}
{--max-id=}
{--throttle=}
{--i-am-sure}';
public function handle(Container $container)
protected $description = 'Build and manage the Elasticsearch search index.';
protected $help = <<<'HELP'
<comment>Actions and their options:</comment>
<info>build</info> Queue documents and promote automatically.
<comment>--resume</comment> Resume an interrupted build from where it left off
<comment>--fresh</comment> Drop the staging index and start completely fresh
<comment>--staging</comment> Keep in staging requires explicit <info>promote</info> (blue-green)
<comment>--keep-backup</comment> Retain the replaced index for rollback
<info>promote</info> Swap alias to staging index (blue-green workflow).
<comment>--keep-backup</comment> Retain the replaced index for rollback
<comment>--i-am-sure</comment> Skip the confirmation prompt
<info>rollback</info> Restore the backup index to live.
<info>discard</info> Drop an index without promoting.
<comment>--pending</comment> Drop the staging index (cancel a build)
<comment>--backup</comment> Drop the backup index (cleanup after --keep-backup)
<info>mapping</info> Push updated mapping to the live index only.
<info>fill</info> Seed only documents missing from the live index.
<comment>Shared seeding options (build / fill):</comment>
<comment>--only=TYPE</comment> Seed only this type: <info>discussions</info> or <info>posts</info>
<comment>--throttle=N</comment> Seconds to wait between batches
<comment>--max-id=N</comment> Limit seeding to IDs up to this value
HELP;
public function handle(Container $container): void
{
$index = $container->make('blomstra.search.elastic_index');
/** @var array $seeders */
$seeders = $container->tagged('blomstra.search.seeders');
/** @var Queue $queue */
$queue = $container->make(Queue::class);
/** @var Client $client */
$client = $container->make(Client::class);
/** @var SettingsRepositoryInterface $settings */
$settings = $container->make(SettingsRepositoryInterface::class);
$properties = [
'properties' => [
'content' => ['type' => 'text', 'analyzer' => 'flarum_analyzer_partial', 'search_analyzer' => 'flarum_analyzer'],
'created_at' => ['type' => 'date'],
'updated_at' => ['type' => 'date'],
'is_private' => ['type' => 'boolean'],
'is_sticky' => ['type' => 'boolean'],
'groups' => ['type' => 'integer'],
'recipient_groups' => ['type' => 'integer'],
'recipient_users' => ['type' => 'integer'],
'comment_count' => ['type' => 'integer'],
],
];
/** @var string $alias */
$alias = $container->make('blomstra.search.elastic_index');
if ($this->option('recreate')) {
// Flush the index.
$client->indices()->delete([
'index' => $index,
'ignore_unavailable' => true,
]);
// Create a new index.
$client->indices()->create([
'index' => $index,
'body' => [
'settings' => [
'index.max_ngram_diff' => 10,
'analysis' => [
'analyzer' => [
'flarum_analyzer' => [
'type' => $settings->get('blomstra-search.analyzer-language') ?: 'english',
],
'flarum_analyzer_partial' => [
'type' => 'custom',
'tokenizer' => 'standard',
'filter' => [
'lowercase',
'partial_search_filter',
],
],
],
'filter' => [
'partial_search_filter' => [
'type' => 'ngram',
'min_gram' => 1,
'max_gram' => 10,
'token_chars' => ['letter', 'digit', 'symbol'],
],
],
],
],
],
]);
if (!$this->argument('action')) {
$this->call('help', ['command_name' => $this->getName()]);
return;
}
if ($this->option('recreate') || $this->option('mapping')) {
switch ($this->argument('action')) {
case 'build':
$this->runBuild($client, $alias, $settings, $container);
break;
case 'promote':
$this->runPromote($client, $alias, $settings);
break;
case 'rollback':
$this->runRollback($client, $alias, $settings);
break;
case 'discard':
$this->runDiscard($client, $settings);
break;
case 'mapping':
$client->indices()->putMapping([
'index' => $index,
'body' => $properties,
'index' => $alias,
'body' => $this->mappingProperties(),
]);
$this->info('Mapping updated on live index.');
break;
case 'fill':
$this->runSeeders(
collect($container->tagged('blomstra.search.seeders')),
$container->make(Queue::class),
$client,
$settings,
$alias,
seedMissing: true
);
break;
default:
$this->error("Unknown action '{$this->argument('action')}'. Valid actions: build, promote, rollback, discard, mapping, fill.");
}
}
protected function runBuild(Client $client, string $alias, SettingsRepositoryInterface $settings, Container $container): void
{
/** @var Seeder[] $seeders */
$seeders = collect($container->tagged('blomstra.search.seeders'));
// Guard: staging build exists without explicit intent given.
$staging = $settings->get('blomstra-search.staging-index');
if ($staging && $client->indices()->exists(['index' => $staging])
&& !$this->option('resume') && !$this->option('fresh')
) {
$this->warn("An in-progress build already exists: $staging");
$this->line('');
$this->line('Choose one of:');
$this->line(' blomstra:search:index build --resume Continue from where it left off');
$this->line(' blomstra:search:index build --fresh Drop this build and start completely fresh');
return;
}
$aliasExists = (bool) $client->indices()->existsAlias(['name' => $alias]);
$indexExists = !$aliasExists && (bool) $client->indices()->exists(['index' => $alias]);
if (!$aliasExists && !$indexExists) {
// First install: create a timestamped index, alias it immediately, seed live.
$targetIndex = $this->prepareFirstInstall($client, $alias, $settings, $seeders);
$stagingBuild = false;
} else {
// Live system: build into a staging index without touching the live alias.
$targetIndex = $this->prepareStagingIndex($client, $alias, $settings, $seeders);
$stagingBuild = true;
}
$client->indices()->putMapping([
'index' => $targetIndex,
'body' => $this->mappingProperties(),
]);
$this->runSeeders($seeders, $container->make(Queue::class), $client, $settings, $targetIndex);
if ($stagingBuild) {
if ($this->option('staging')) {
$staging = $settings->get('blomstra-search.staging-index');
$this->info("Staging build complete. Drain the queue, then promote '$staging' to live:");
$this->line(' php flarum queue:work --stop-when-empty');
$this->line(' php flarum blomstra:search:index promote');
} else {
$this->info('Jobs queued. Promoting now (search improves as the queue drains)...');
$this->performPromote($client, $alias, $settings);
}
}
}
protected function runPromote(Client $client, string $alias, SettingsRepositoryInterface $settings): void
{
if (!$this->option('i-am-sure')) {
$this->warn('promote switches the live index. Ensure all queued jobs have finished first:');
$this->line(' php flarum queue:work --stop-when-empty');
$this->line('');
if (!$this->confirm('Queue drained? Proceed?')) {
return;
}
}
$this->performPromote($client, $alias, $settings);
}
protected function performPromote(Client $client, string $alias, SettingsRepositoryInterface $settings): void
{
$staging = $settings->get('blomstra-search.staging-index');
if (!$staging || !$client->indices()->exists(['index' => $staging])) {
$this->error('No staging index ready to promote. Run: blomstra:search:index build');
return;
}
$aliasExists = (bool) $client->indices()->existsAlias(['name' => $alias]);
$indexExists = !$aliasExists && (bool) $client->indices()->exists(['index' => $alias]);
if ($aliasExists) {
$result = $client->indices()->getAlias(['name' => $alias]);
$activeIndex = array_key_first($result);
$client->indices()->updateAliases([
'body' => ['actions' => [
['remove' => ['index' => $activeIndex, 'alias' => $alias]],
['add' => ['index' => $staging, 'alias' => $alias]],
]],
]);
$this->info("'$alias' now points to '$staging'. Promote complete.");
if ($this->option('keep-backup')) {
$settings->set('blomstra-search.backup-index', $activeIndex);
$this->info("Kept '$activeIndex' as backup — run rollback to revert, discard --backup to drop it.");
} else {
$client->indices()->delete(['index' => $activeIndex, 'ignore_unavailable' => true]);
$this->info("Deleted old index: $activeIndex");
}
} else {
// One-time migration: concrete index → alias (legacy installs only).
if ($indexExists) {
$client->indices()->delete(['index' => $alias]);
$this->info("Deleted legacy concrete index: $alias");
}
$client->indices()->putAlias(['index' => $staging, 'name' => $alias]);
$this->info("'$alias' now points to '$staging'. Promote complete.");
}
$settings->set('blomstra-search.active-index', $staging);
$settings->set('blomstra-search.staging-index', null);
$this->saveIndexedConfig($client, $settings, $staging);
}
protected function runRollback(Client $client, string $alias, SettingsRepositoryInterface $settings): void
{
$backup = $settings->get('blomstra-search.backup-index');
if (!$backup || !$client->indices()->exists(['index' => $backup])) {
$this->error('No backup index to roll back to. Was promote run with --keep-backup?');
return;
}
$result = $client->indices()->getAlias(['name' => $alias]);
$activeIndex = array_key_first($result);
$client->indices()->updateAliases([
'body' => ['actions' => [
['remove' => ['index' => $activeIndex, 'alias' => $alias]],
['add' => ['index' => $backup, 'alias' => $alias]],
]],
]);
$client->indices()->delete(['index' => $activeIndex, 'ignore_unavailable' => true]);
$this->info("Rolled back: '$alias' now points to '$backup'. Deleted '$activeIndex'.");
$settings->set('blomstra-search.active-index', $backup);
$settings->set('blomstra-search.backup-index', null);
$this->saveIndexedConfig($client, $settings, $backup);
}
protected function runDiscard(Client $client, SettingsRepositoryInterface $settings): void
{
$pending = $this->option('pending');
$backup = $this->option('backup');
if (!$pending && !$backup) {
$this->error('Specify what to discard: --pending (staging build) or --backup (rollback index).');
return;
}
if ($pending) {
$staging = $settings->get('blomstra-search.staging-index');
if (!$staging) {
$this->warn('No staging index to discard.');
} else {
if ($client->indices()->exists(['index' => $staging])) {
$client->indices()->delete(['index' => $staging]);
}
$settings->set('blomstra-search.staging-index', null);
$this->info("Discarded staging index: $staging");
}
}
if ($backup) {
$backupIndex = $settings->get('blomstra-search.backup-index');
if (!$backupIndex) {
$this->warn('No backup index to discard.');
} else {
if ($client->indices()->exists(['index' => $backupIndex])) {
$client->indices()->delete(['index' => $backupIndex]);
}
$settings->set('blomstra-search.backup-index', null);
$this->info("Discarded backup index: $backupIndex");
}
}
}
protected function runSeeders(
iterable $seeders,
Queue $queue,
Client $client,
SettingsRepositoryInterface $settings,
string $targetIndex,
bool $seedMissing = false
): void {
$only = $this->option('only');
/** @var Seeder $seeder */
@ -122,33 +339,63 @@ class BuildCommand extends Command
$total = 0;
$continueAt = $this->option('continue')
? ($this->continueAt($seeder->type()) ?? $seeder->query()->max('id'))
: $seeder->query()->max('id');
if ($this->option('resume')) {
$saved = $this->getContinueAt($settings, $seeder->type());
if ($saved === 0) {
$this->info("Seeder '{$seeder->type()}' already completed in a previous run, skipping.");
continue;
}
$continueAt = $saved ?? $seeder->query()->max('id');
} else {
$continueAt = $seeder->query()->max('id');
}
$seeded = null;
while ($continueAt !== null) {
$rangeFrom = max(1, $continueAt - 2500);
$rangeTo = $continueAt;
if ($seedMissing) {
$response = (new Builder($client))
->index($targetIndex)
->size(2500)
->addQuery(
(new BoolQuery())
->add((new RangeQuery('rawId'))->gte($rangeFrom)->lte($rangeTo))
->add(TermQuery::create('join_field', $seeder->joinRelation()))
)
->search();
$seeded = Arr::pluck(Arr::get($response, 'hits.hits'), '_source.rawId');
}
/** @var Collection $collection */
$collection = $seeder->query()
->latest('id')
->whereBetween('id', [$continueAt - 1000, $continueAt])
->when($this->option('max-id'), function ($query, $id) {
$query->where('id', '<=', $id);
})
->whereBetween('id', [$rangeFrom, $rangeTo])
->when($this->option('max-id'), fn ($q, $id) => $q->where('id', '<=', $id))
->when($seeded, fn ($q, $seeded) => $q->whereNotIn('id', $seeded))
->get();
$min = $collection->min('id');
if ($seedMissing && $collection->isEmpty()) {
$continueAt = $rangeFrom > 2 ? $rangeFrom - 1 : null;
} else {
$continueAt = $min && $min > 2 ? $min - 1 : null;
}
$queue->pushOn(Job::$onQueue, new SavingJob($collection, $seeder));
if ($collection->isNotEmpty()) {
$queue->pushOn(Job::$onQueue, new UpdateSearchJob($collection, $seeder, $targetIndex));
}
$this->info("Pushed into the index, type: {$seeder->type()}, amount: {$collection->count()}.");
$this->info("IDs {$rangeFrom}{$rangeTo} | type: {$seeder->type()} | queued: {$collection->count()}.");
$total += $collection->count();
$this->continueAt(
$seeder->type(),
$continueAt
);
$this->setContinueAt($settings, $seeder->type(), $continueAt);
if ($throttle = $this->option('throttle')) {
$this->info("Throttling for $throttle seconds");
@ -156,21 +403,181 @@ class BuildCommand extends Command
}
}
$this->info("Pushed a total of $total into the index.");
$this->setContinueAt($settings, $seeder->type(), 0);
$this->info("Queued a total of $total {$seeder->type()} for indexing.");
}
}
protected function continueAt(string $type, int $at = null)
/**
* First install: create a timestamped concrete index, alias the configured name to it
* immediately, and return the alias so seeding goes through it and is live from the start.
*/
protected function prepareFirstInstall(
Client $client,
string $alias,
SettingsRepositoryInterface $settings,
iterable $seeders
): string {
$concrete = $alias . '_' . date('YmdHis');
$client->indices()->create([
'index' => $concrete,
'body' => ['settings' => $this->buildIndexSettings($settings)],
]);
$client->indices()->putAlias(['index' => $concrete, 'name' => $alias]);
$settings->set('blomstra-search.active-index', $concrete);
$this->saveIndexedConfig($client, $settings, $concrete);
foreach ($seeders as $seeder) {
$this->setContinueAt($settings, $seeder->type(), null);
}
$this->info("Created '$concrete', aliased '$alias' → '$concrete'.");
$this->info("Index is live — documents become searchable as the queue processes.");
return $alias;
}
/**
* Prepare the staging index for a blue-green build.
*
* - If a staging build exists and --fresh is not set, resume it.
* - Otherwise create a fresh timestamped index and save it as staging.
*/
protected function prepareStagingIndex(
Client $client,
string $alias,
SettingsRepositoryInterface $settings,
iterable $seeders
): string {
$staging = $settings->get('blomstra-search.staging-index');
if ($this->option('fresh') && $staging) {
if ($client->indices()->exists(['index' => $staging])) {
$client->indices()->delete(['index' => $staging]);
$this->info("Dropped staging index: $staging");
}
$staging = null;
$settings->set('blomstra-search.staging-index', null);
}
if ($staging && $client->indices()->exists(['index' => $staging])) {
$this->info("Resuming staging index build: $staging");
return $staging;
}
$staging = $alias . '_' . date('YmdHis');
$client->indices()->create([
'index' => $staging,
'body' => ['settings' => $this->buildIndexSettings($settings)],
]);
$settings->set('blomstra-search.staging-index', $staging);
foreach ($seeders as $seeder) {
$this->setContinueAt($settings, $seeder->type(), null);
}
$this->info("Created staging index: $staging");
return $staging;
}
/**
* Persist the analysis config and compat version that are actually live in ES for the
* given index. Reading from ES (rather than from Flarum settings) means rollbacks are
* also covered: the stored values always reflect the index that is currently aliased,
* not the settings at the time the command ran. Indexes built before _meta tracking
* existed will yield a null compat version, which correctly triggers the reindex warning.
*/
protected function saveIndexedConfig(Client $client, SettingsRepositoryInterface $settings, string $indexName): void
{
/** @var SettingsRepositoryInterface $settings */
$settings = resolve(SettingsRepositoryInterface::class);
$settingsResponse = $client->indices()->getSettings(['index' => $indexName]);
$analysis = Arr::get($settingsResponse, "$indexName.settings.index.analysis", []);
$key = "blomstra-search.continued-at.$type";
$analyzer = Arr::get($analysis, 'analyzer.flarum_analyzer.type', 'english');
$minGram = (int) Arr::get($analysis, 'filter.partial_search_filter.min_gram', self::DEFAULT_MIN_SEARCH_LENGTH);
if ($at) {
$settings->set($key, $at);
} else {
return $settings->get($key);
$mappingResponse = $client->indices()->getMapping(['index' => $indexName]);
$compatVersion = Arr::get($mappingResponse, "$indexName.mappings._meta.index_compat_version");
$settings->set('blomstra-search.indexed-analyzer', $analyzer);
$settings->set('blomstra-search.indexed-min-search-length', $minGram);
$settings->set('blomstra-search.index-compatible', $compatVersion);
}
protected function buildIndexSettings(SettingsRepositoryInterface $settings): array
{
$language = $settings->get('blomstra-search.analyzer-language') ?: 'english';
$minGram = max(1, (int) ($settings->get('blomstra-search.min-search-length') ?: self::DEFAULT_MIN_SEARCH_LENGTH));
$maxGram = 10;
if ($minGram >= $maxGram) {
$this->error("min_gram ($minGram) must be less than max_gram ($maxGram). Using default.");
$minGram = self::DEFAULT_MIN_SEARCH_LENGTH;
}
return [
'index.max_ngram_diff' => $maxGram - $minGram,
'analysis' => [
'analyzer' => [
'flarum_analyzer' => [
'type' => $language,
],
'flarum_analyzer_partial' => [
'type' => 'custom',
'tokenizer' => 'standard',
'filter' => ['lowercase', 'partial_search_filter'],
],
],
'filter' => [
'partial_search_filter' => [
'type' => 'ngram',
'min_gram' => $minGram,
'max_gram' => $maxGram,
'token_chars' => ['letter', 'digit', 'symbol'],
],
],
],
];
}
protected function mappingProperties(): array
{
return [
'_meta' => ['index_compat_version' => self::INDEX_COMPAT_VERSION],
'properties' => [
'join_field' => ['type' => 'join', 'relations' => ['discussion' => 'post']],
'discussion_id' => ['type' => 'integer'],
'content' => ['type' => 'text', 'analyzer' => 'flarum_analyzer_partial', 'search_analyzer' => 'flarum_analyzer'],
'rawId' => ['type' => 'integer'],
'created_at' => ['type' => 'date'],
'updated_at' => ['type' => 'date'],
'is_private' => ['type' => 'boolean'],
'is_sticky' => ['type' => 'boolean'],
'groups' => ['type' => 'integer'],
'tags' => ['type' => 'integer'],
'recipient_groups' => ['type' => 'integer'],
'recipient_users' => ['type' => 'integer'],
'comment_count' => ['type' => 'integer'],
'view_count' => ['type' => 'integer'],
'is_hidden' => ['type' => 'boolean'],
],
];
}
protected function getContinueAt(SettingsRepositoryInterface $settings, string $type): ?int
{
$raw = $settings->get("blomstra-search.continued-at.$type");
return $raw !== null ? (int) $raw : null;
}
protected function setContinueAt(SettingsRepositoryInterface $settings, string $type, ?int $at): void
{
$settings->set("blomstra-search.continued-at.$type", $at);
}
}

View File

@ -0,0 +1,57 @@
<?php
/*
* This file is part of blomstra/search.
*
* Copyright (c) 2022 Blomstra Ltd.
*
* For the full copyright and license information, please view the LICENSE.md
* file that was distributed with this source code.
*
*/
namespace Blomstra\Search\Elasticsearch;
use Spatie\ElasticsearchQueryBuilder\Queries\Query;
class HasChildQuery implements Query
{
protected bool $innerHits = false;
public function __construct(
protected string $type,
protected Query $query,
protected string $scoreMode = 'sum'
) {}
public static function create(string $type, Query $query, string $scoreMode = 'sum'): static
{
return new static($type, $query, $scoreMode);
}
/**
* Include inner_hits so the best-matching post ID is available in the response.
* ES returns the top-scoring child document per parent hit under inner_hits.best_post.
*/
public function withInnerHits(): static
{
$this->innerHits = true;
return $this;
}
public function toArray(): array
{
$hasChild = [
'type' => $this->type,
'score_mode' => $this->scoreMode,
'query' => $this->query->toArray(),
];
if ($this->innerHits) {
$hasChild['inner_hits'] = ['name' => 'best_post', 'size' => 1];
}
return ['has_child' => $hasChild];
}
}

View File

@ -26,9 +26,10 @@ class DeletingJob extends Job
// Preparing body for storing.
$body = $this->models->map(function (Model $model) {
$document = $this->seeder->toDocument($model);
$routing = $this->seeder->routing($model);
return [
['delete' => ['_index' => $this->index, '_id' => $document->id]],
['delete' => ['_index' => $this->index, '_id' => $document->id, 'routing' => $routing]],
];
})->flatten(1);

View File

@ -22,9 +22,13 @@ abstract class Job extends AbstractJob
public static ?string $onQueue = null;
public function __construct(protected Collection $models, protected Seeder $seeder)
/**
* @param string|null $targetIndex Explicit index name for blue-green builds.
* Defaults to the configured alias when null.
*/
public function __construct(protected Collection $models, protected Seeder $seeder, ?string $targetIndex = null)
{
$this->index = resolve('blomstra.search.elastic_index');
$this->index = $targetIndex ?? resolve('blomstra.search.elastic_index');
if (static::$onQueue) {
$this->onQueue(static::$onQueue);

View File

@ -17,7 +17,7 @@ use Elasticsearch\Client;
use Illuminate\Database\Eloquent\Model;
use Illuminate\Support\Arr;
class SavingJob extends Job
class UpdateSearchJob extends Job
{
public function handle(Client $client)
{
@ -25,12 +25,15 @@ class SavingJob extends Job
return;
}
$this->models->loadMissing($this->seeder->relationships());
// Preparing body for storing.
$body = $this->models->map(function (Model $model) {
$document = $this->seeder->toDocument($model);
$routing = $this->seeder->routing($model);
return [
['index' => ['_index' => $this->index, '_id' => $document->id]],
['index' => ['_index' => $this->index, '_id' => $document->id, 'routing' => $routing]],
$document->toArray(),
];
})
@ -39,7 +42,6 @@ class SavingJob extends Job
$response = $client->bulk([
'index' => $this->index,
'body' => $body->toArray(),
'refresh' => true,
]);
if (Arr::get($response, 'errors') !== true) {

View File

@ -0,0 +1,59 @@
<?php
/*
* This file is part of blomstra/search.
*
* Copyright (c) 2022 Blomstra Ltd.
*
* For the full copyright and license information, please view the LICENSE.md
* file that was distributed with this source code.
*
*/
namespace Blomstra\Search\Jobs;
use Elasticsearch\Client;
use Flarum\Api\Serializer\DiscussionSerializer;
use Flarum\Discussion\Discussion;
use Flarum\Queue\AbstractJob;
class ViewsSearchJob extends AbstractJob
{
protected string $index;
protected string $documentType;
public function __construct(protected int $discussionId)
{
$this->index = resolve('blomstra.search.elastic_index');
$this->documentType = resolve(DiscussionSerializer::class)->getType(new Discussion());
if (Job::$onQueue) {
$this->onQueue(Job::$onQueue);
}
}
public function handle(Client $client): void
{
$discussion = Discussion::find($this->discussionId);
if (!$discussion) {
return;
}
$type = $this->documentType;
try {
$client->update([
'index' => $this->index,
'id' => "$type:{$this->discussionId}",
'routing' => (string) $this->discussionId,
'retry_on_conflict' => 3,
'body' => [
'doc' => ['view_count' => (int) $discussion->view_count],
],
]);
} catch (\Elasticsearch\Common\Exceptions\Missing404Exception $e) {
// Document not yet indexed; will be picked up on next: blomstra:search:index fill
}
}
}

View File

@ -14,7 +14,8 @@ namespace Blomstra\Search;
use Blomstra\Search\Jobs\DeletingJob;
use Blomstra\Search\Jobs\Job;
use Blomstra\Search\Jobs\SavingJob;
use Blomstra\Search\Jobs\UpdateSearchJob;
use Blomstra\Search\Jobs\ViewsSearchJob;
use Elasticsearch\Client as Elastic;
use Elasticsearch\ClientBuilder;
use Flarum\Api\Client;
@ -87,6 +88,11 @@ class Provider extends AbstractServiceProvider
return new Api\Client($pipe);
}
);
$this->container->tag([
Searchers\DiscussionSearcher::class,
Searchers\CommentPostSearcher::class,
], 'blomstra.search.searchers');
}
public function boot()
@ -103,12 +109,16 @@ class Provider extends AbstractServiceProvider
/** @var string|Seeders\Seeder $seeder */
foreach ($seeders as $seeder) {
$seeder::savingOn($events, function ($model) use ($queue, $seeder) {
$queue->pushOn(Job::$onQueue, new SavingJob(Collection::make([$model]), $seeder));
$queue->pushOn(Job::$onQueue, new UpdateSearchJob(Collection::make([$model]), $seeder));
});
$seeder::deletingOn($events, function ($model) use ($queue, $seeder) {
$queue->pushOn(Job::$onQueue, new DeletingJob(Collection::make([$model]), $seeder));
});
$seeder::viewingOn($events, function (int $discussionId) use ($queue) {
$queue->pushOn(Job::$onQueue, new ViewsSearchJob($discussionId));
});
}
}
}

View File

@ -13,11 +13,12 @@
namespace Blomstra\Search\Save;
use Carbon\Carbon;
use Illuminate\Support\Arr;
use Illuminate\Support\Fluent;
/**
* @property string $type
* @property string $id
* @property int $rawId
* @property string $content
* @property Carbon $created_at
* @property Carbon $updated_at
@ -30,4 +31,13 @@ use Illuminate\Support\Fluent;
*/
class Document extends Fluent
{
/**
* Exclude `id` from the body sent to Elasticsearch. The document ID is
* already stored as `_id` by the bulk API action; duplicating it in
* `_source` wastes space and is redundant.
*/
public function toArray(): array
{
return Arr::except($this->attributes, ['id']);
}
}

View File

@ -0,0 +1,27 @@
<?php
/*
* This file is part of blomstra/search.
*
* Copyright (c) 2022 Blomstra Ltd.
*
* For the full copyright and license information, please view the LICENSE.md
* file that was distributed with this source code.
*
*/
namespace Blomstra\Search\Searchers;
use Blomstra\Search\Seeders\CommentSeeder;
class CommentPostSearcher extends Searcher
{
protected string|null $seeder = CommentSeeder::class;
public function enabled(): bool
{
$enabled = $this->setting('blomstra-search.search-post-bodies', true);
return boolval($enabled);
}
}

View File

@ -0,0 +1,32 @@
<?php
/*
* This file is part of blomstra/search.
*
* Copyright (c) 2022 Blomstra Ltd.
*
* For the full copyright and license information, please view the LICENSE.md
* file that was distributed with this source code.
*
*/
namespace Blomstra\Search\Searchers;
use Blomstra\Search\Seeders\DiscussionSeeder;
class DiscussionSearcher extends Searcher
{
protected string|null $seeder = DiscussionSeeder::class;
public function enabled(): bool
{
$enabled = $this->setting('blomstra-search.search-discussion-subjects', true);
return boolval($enabled);
}
public function boost(): float
{
return 1.5;
}
}

View File

@ -0,0 +1,48 @@
<?php
/*
* This file is part of blomstra/search.
*
* Copyright (c) 2022 Blomstra Ltd.
*
* For the full copyright and license information, please view the LICENSE.md
* file that was distributed with this source code.
*
*/
namespace Blomstra\Search\Searchers;
use Blomstra\Search\Seeders\Seeder;
use Flarum\Settings\SettingsRepositoryInterface;
abstract class Searcher
{
protected string|null $seeder = null;
public function type(): string
{
/** @var Seeder $seeder */
$seeder = $this->seeder;
if (empty($seeder)) {
throw new \InvalidArgumentException('Implement type or add $seeder');
}
return (new $seeder())->type();
}
public function enabled(): bool
{
return true;
}
public function boost(): float
{
return 1;
}
protected function setting(string $key, $default = null)
{
return resolve(SettingsRepositoryInterface::class)->get($key, $default);
}
}

View File

@ -13,10 +13,11 @@
namespace Blomstra\Search\Seeders;
use Blomstra\Search\Save\Document;
use Flarum\Api\Serializer\DiscussionSerializer;
use Flarum\Api\Serializer\PostSerializer;
use Flarum\Discussion\Discussion;
use Flarum\Post\CommentPost;
use Flarum\Post\Event\Deleted;
use Flarum\Post\Event\Posted;
use Flarum\Post\Event as Core;
use Illuminate\Contracts\Events\Dispatcher;
use Illuminate\Database\Eloquent\Builder;
use Illuminate\Database\Eloquent\Model;
@ -28,64 +29,68 @@ class CommentSeeder extends Seeder
return resolve(PostSerializer::class)->getType(new CommentPost());
}
public function joinRelation(): string
{
return 'post';
}
public function routing(Model $model): string
{
return (string) $model->discussion_id;
}
public function query(): Builder
{
$includes = ['discussion'];
if ($this->extensionEnabled('flarum-tags')) {
$includes[] = 'discussion.tags';
}
if ($this->extensionEnabled('fof-byobu')) {
$includes[] = 'discussion.recipientUsers';
$includes[] = 'discussion.recipientGroups';
}
return CommentPost::query()
->whereNull('hidden_at')
->where('type', CommentPost::$type)
->with($includes);
->where('type', CommentPost::$type);
}
public static function savingOn(Dispatcher $events, callable $callable)
{
$events->listen(Posted::class, function (Posted $event) use ($callable) {
$events->listen([
Core\Posted::class,
Core\Revised::class,
Core\Hidden::class,
Core\Restored::class,
], function ($event) use ($callable) {
$callable($event->post);
});
}
public static function deletingOn(Dispatcher $events, callable $callable)
{
$events->listen(Deleted::class, function (Deleted $event) use ($callable) {
$events->listen([
Core\Deleted::class
], function ($event) use ($callable) {
$callable($event->post);
});
}
/** Cached discussion type string (e.g. "discussions") — resolved once per job. */
private ?string $discussionType = null;
private function discussionType(): string
{
return $this->discussionType ??= resolve(DiscussionSerializer::class)->getType(new Discussion());
}
/**
* @param CommentPost $model
* Post documents only need content for has_child matching and the join
* field to establish the parent-child relationship. All discussion-level
* fields (groups, tags, comment_count, etc.) live on the discussion
* document and are irrelevant here.
*
* @return Document
* @param CommentPost $model
*/
public function toDocument(Model $model): Document
{
$document = new Document([
'type' => $this->type(),
return new Document([
'join_field' => ['name' => $this->joinRelation(), 'parent' => "{$this->discussionType()}:{$model->discussion_id}"],
'discussion_id' => $model->discussion_id,
'id' => $this->type().':'.$model->id,
'rawId' => $model->id,
'content' => $model->content,
'content_partial' => $model->content,
'created_at' => $model->created_at?->toAtomString(),
'updated_at' => $model->edited_at?->toAtomString(),
'is_private' => $model->is_private,
'user_id' => $model->user_id,
'groups' => $this->groupsForDiscussion($model->discussion),
'comment_count' => $model->discussion->comment_count,
'is_hidden' => $model->hidden_at !== null,
]);
if ($this->extensionEnabled('fof-byobu')) {
$document['recipient_users'] = $model->discussion->recipientUsers->pluck('id')->toArray();
$document['recipient_groups'] = $model->discussion->recipientGroups->pluck('id')->toArray();
}
return $document;
}
}

View File

@ -15,12 +15,16 @@ namespace Blomstra\Search\Seeders;
use Blomstra\Search\Save\Document;
use Flarum\Api\Serializer\DiscussionSerializer;
use Flarum\Discussion\Discussion;
use Flarum\Discussion\Event\Deleted;
use Flarum\Discussion\Event\Hidden;
use Flarum\Discussion\Event\Restored;
use Flarum\Discussion\Event\Started;
use Flarum\Discussion\Event as Core;
use Flarum\Extension\ExtensionManager;
use Flarum\Group\Group;
use Flarum\Group\Permission;
use Flarum\Tags\Tag;
use FoF\Byobu\Events as Byobu;
use FoF\DiscussionViews\Events\DiscussionWasViewed;
use Illuminate\Contracts\Events\Dispatcher;
use Illuminate\Database\Eloquent\Builder;
use Illuminate\Database\Eloquent\Collection;
use Illuminate\Database\Eloquent\Model;
class DiscussionSeeder extends Seeder
@ -30,7 +34,23 @@ class DiscussionSeeder extends Seeder
return resolve(DiscussionSerializer::class)->getType(new Discussion());
}
public function joinRelation(): string
{
return 'discussion';
}
public function routing(Model $model): string
{
return (string) $model->id;
}
public function query(): Builder
{
return Discussion::query()
->whereNull('hidden_at');
}
public function relationships(): array
{
$includes = [];
@ -43,21 +63,48 @@ class DiscussionSeeder extends Seeder
$includes[] = 'recipientGroups';
}
return Discussion::query()
->whereNull('hidden_at')
->with($includes);
return $includes;
}
public static function savingOn(Dispatcher $events, callable $callable)
{
$events->listen([Started::class, Restored::class], function ($event) use ($callable) {
$events->listen([
// flarum/core events
Core\Started::class, Core\Restored::class,
// fof/byobu discussion recipients events.
Byobu\DiscussionMadePublic::class, Byobu\RemovedSelf::class, Byobu\RecipientsChanged::class,
], function ($event) use ($callable) {
return $callable($event->discussion);
});
}
public static function viewingOn(Dispatcher $events, callable $callable): void
{
if (!resolve(ExtensionManager::class)->isEnabled('fof-discussion-views')) {
return;
}
$events->listen(DiscussionWasViewed::class, function (DiscussionWasViewed $event) use ($callable) {
$viewCount = $event->discussion->view_count;
$shouldSync = match (true) {
$viewCount < 15 => true,
$viewCount < 100 => rand(1, 3) === 1,
default => rand(1, 19) === 1,
};
if ($shouldSync) {
$callable($event->discussion->id);
}
});
}
public static function deletingOn(Dispatcher $events, callable $callable)
{
$events->listen([Deleted::class, Hidden::class], function ($event) use ($callable) {
$events->listen([
// flarum/core events.
Core\Deleted::class, Core\Hidden::class
], function ($event) use ($callable) {
return $callable($event->discussion);
});
}
@ -70,10 +117,10 @@ class DiscussionSeeder extends Seeder
public function toDocument(Model $model): Document
{
$document = new Document([
'type' => $this->type(),
'join_field' => $this->joinRelation(),
'id' => $this->type().':'.$model->id,
'rawId' => $model->id,
'content' => $model->title,
'content_partial' => $model->title,
'created_at' => $model->created_at?->toAtomString(),
'updated_at' => $model->last_posted_at?->toAtomString(),
'is_private' => $model->is_private,
@ -82,15 +129,81 @@ class DiscussionSeeder extends Seeder
'comment_count' => $model->comment_count,
]);
if ($this->extensionEnabled('flarum-tags')) {
$document['tags'] = $model->tags->pluck('id')->toArray();
}
if ($this->extensionEnabled('fof-byobu')) {
$document['recipient_users'] = $model->recipientUsers->pluck('id')->toArray();
$document['recipient_groups'] = $model->recipientGroups->pluck('id')->toArray();
$document['recipient_users'] = $model->recipientUsers
->whereNull('removed_at')
->pluck('id')
->toArray();
$document['recipient_groups'] = $model->recipientGroups
->whereNull('removed_at')
->pluck('id')
->toArray();
}
if ($this->extensionEnabled('flarum-sticky')) {
$document['is_sticky'] = (bool) $model->is_sticky;
}
if ($this->extensionEnabled('fof-discussion-views')) {
$document['view_count'] = (int) ($model->view_count ?? 0);
}
return $document;
}
/**
* All viewForum permissions keyed by permission string, loaded once per seeder instance.
* Avoids N×2 Permission queries inside the per-document map loop.
*/
private ?Collection $cachedPermissions = null;
private ?Collection $cachedGlobalPermission = null;
private function allPermissions(): Collection
{
if ($this->cachedPermissions === null) {
$this->cachedPermissions = Permission::query()
->where(function ($q) {
$q->where('permission', 'viewForum')
->orWhere('permission', 'like', 'tag%.viewForum');
})
->get();
$this->cachedGlobalPermission = $this->cachedPermissions
->where('permission', 'viewForum')
->pluck('group_id');
}
return $this->cachedPermissions;
}
protected function groupsForDiscussion(Discussion $discussion): array
{
$allPerms = $this->allPermissions();
$permissions = collect();
if ($this->extensionEnabled('flarum-tags')) {
/** @var Collection $tags */
$tags = $discussion->tags;
$permissions = $tags->map(function (Tag $tag) use ($allPerms) {
$tagPerms = $allPerms->where('permission', "tag$tag->id.viewForum");
if ($tag->is_restricted) {
$tagPerms = $tagPerms->add(['group_id' => Group::ADMINISTRATOR_ID]);
}
return $tagPerms->pluck('group_id');
})->flatten();
}
if (!$discussion->is_private && $permissions->isEmpty()) {
$permissions = $this->cachedGlobalPermission;
}
return $permissions->toArray();
}
}

View File

@ -13,65 +13,42 @@
namespace Blomstra\Search\Seeders;
use Blomstra\Search\Save\Document;
use Flarum\Discussion\Discussion;
use Flarum\Extension\ExtensionManager;
use Flarum\Group\Group;
use Flarum\Group\Permission;
use Flarum\Tags\Tag;
use Illuminate\Contracts\Events\Dispatcher;
use Illuminate\Database\Eloquent\Builder;
use Illuminate\Database\Eloquent\Collection;
use Illuminate\Database\Eloquent\Model;
abstract class Seeder
{
abstract public function type(): string;
/**
* The join relation name for this document type ('discussion' or 'post').
* Used in the join_field mapping and for filtering in seed-missing checks.
*/
abstract public function joinRelation(): string;
/**
* The ES routing key for this model. Parent and child documents for the same
* discussion must share a routing key so ES places them on the same shard.
*/
abstract public function routing(Model $model): string;
abstract public function query(): Builder;
abstract public static function savingOn(Dispatcher $events, callable $callable);
abstract public static function deletingOn(Dispatcher $events, callable $callable);
/** No-op default; override in seeders that need to react to view events. */
public static function viewingOn(Dispatcher $events, callable $callable): void {}
abstract public function toDocument(Model $model): Document;
protected function groupsForDiscussion(Discussion $discussion): array
/** Relationships to eager-load on the collection before calling toDocument(). */
public function relationships(): array
{
$permissions = collect();
$globalPermission = Permission::query()
->where('permission', 'viewForum')
->pluck('group_id');
if ($this->extensionEnabled('flarum-tags')) {
/** @var Collection $tags */
$tags = $discussion->tags;
$filters['tags'] = $tags->pluck('id')->toArray();
$tagPermissions = Permission::query()
->whereIn(
'permission',
$tags->pluck('id')->map(function (int $id) {
return "tag$id.viewForum";
})
)->get();
$permissions = $tags->map(function (Tag $tag) use ($tagPermissions) {
$permissions = $tagPermissions->where('permission', "tag$tag->id.viewForum");
if ($tag->is_restricted) {
$permissions = $permissions->add(['group_id' => Group::ADMINISTRATOR_ID]);
}
return $permissions->pluck('group_id');
})->flatten();
}
if (!$discussion->is_private && $permissions->isEmpty()) {
$permissions = $globalPermission;
}
return $permissions->toArray();
return [];
}
protected function extensionEnabled(string $extension): bool