xref: /plugin/annotations/DESIGN.md (revision 5fa3d1851dabf8560591f7c142012774077e7767)
1# Annotations Plugin — Design & Architecture
2
3A developer reference for the annotations plugin. For installation and end-user
4behaviour see [README.md](README.md); for the wider review/environment
5conventions see `CLAUDE.md` in the plugins root.
6
7## Concept
8
9Word- and sentence-level annotations on wiki pages, in the spirit of
10Hypothes.is and `ep_comments_page`:
11
12- **Out-of-band.** Annotations live in a separate per-page JSON file, never in
13  the page text or the wiki changelog. Creating one needs only `AUTH_READ`, so
14  a group whose page *edit* access is blocked can still annotate.
15- **Text-quote anchored.** Each annotation is tied to the quoted text plus a
16  little surrounding context, not to a character position, so it survives minor
17  edits and is re-found in the rendered DOM on each page load.
18- **Threaded.** Annotations carry replies; both have open/resolved status at the
19  annotation level.
20- **Orphan-aware.** When the quoted text disappears from the page the annotation
21  becomes an *orphan* — still stored, surfaced through a counter, and bulk-
22  removable by an admin.
23
24## Components
25
26| File | Owns |
27|------|------|
28| `helper.php` | The per-page store, all CRUD, server-side orphan detection, and the **permission rules as the single source of truth**. Pure logic — permission methods take facts (user, admin flag, ACL level) as parameters and read no globals. |
29| `action.php` | Event registration; injecting the page payload into `JSINFO`; the AJAX endpoint and **permission enforcement** (gathers facts from DokuWiki globals, calls the helper). |
30| `script.js` | All front-end behaviour: boot/gate, load + re-anchor, highlights, gutter markers, counter, selection→new-annotation flow, thread panels, and AJAX. Plain IIFE, vanilla JS. |
31| `style.css` | Styling via DokuWiki theme tokens (`__background__`, `__text__`, …). Only the amber (open) / green (resolved) highlight colours are hard-coded. |
32| `lang/en/lang.php` | The usersettings toggle label/description (used) plus a set of UI strings that are **not yet wired into the JS** — see *Known gaps*. |
33
34## Data model & storage
35
36One pretty-printed JSON file per page at `metaFN($id, '.annotations')`
37(`data/meta/<namespace>/<page>.annotations`):
38
39```json
40{
41  "version": 1,
42  "annotations": [
43    {
44      "id": "a1b2c3d4e5f6g7h8",
45      "anchor": { "exact": "...", "prefix": "...", "suffix": "...", "start": 123 },
46      "author": "alice",
47      "created": 1716336000,
48      "modified": 1716336000,
49      "body": "Does this cover remuxes?",
50      "status": "open",
51      "resolved_by": "",
52      "resolved_at": 0,
53      "replies": [
54        {
55          "id": "x1y2z3a4b5c6d7e8",
56          "author": "bob",
57          "created": 1716336100,
58          "modified": 1716336100,
59          "body": "Yes, remuxes count."
60        }
61      ]
62    }
63  ]
64}
65```
66
67Limits and identifiers (`helper.php` constants): `SCHEMA_VERSION = 1`,
68`MAX_QUOTE = 1000`, `MAX_CONTEXT = 64`, `MAX_BODY = 10000`. IDs are
69`bin2hex(random_bytes(8))` — 16 hex chars. Writes go through `io_lock()` →
70modify → `io_saveFile()` → `io_unlock()` (the `mutate()` helper); a modifier
71returning `false` aborts the write (used for "target not found").
72
73## Text-quote anchoring
74
75An anchor is `{exact, prefix, suffix, start}`:
76
77- `exact` — the selected text, whitespace-normalised (runs collapsed to one
78  space, trimmed). The same normalisation is applied on capture (JS), on
79  storage (PHP), and on matching, so client and server agree.
80- `prefix` / `suffix` — context on each side to disambiguate a quote that
81  appears more than once. Client captures ~30 chars; server caps at 64.
82- `start` — a character-offset hint into the page text, used only as a
83  tie-breaker.
84
85**Re-anchoring (client, `findRange`)**: collect the content text with a
86`TreeWalker`, search for the normalised `exact`, disambiguate repeats with
87`prefix`/`suffix`, tie-break with the `start` hint, then map the chosen
88character offset back to a DOM `Range` and wrap it in a highlight `<span>`. A
89quote that cannot be located is an orphan (no highlight, no gutter marker).
90
91## Orphan detection (two layers)
92
93- **Client (live UI).** Anything `findRange` cannot anchor on page load is
94  counted as orphaned; the count feeds the counter bar, and the orphaned link
95  opens a drawer at the bottom of the content area with those threads.
96- **Server (authoritative, `findOrphaned`).** For the admin "clear orphaned"
97  action the page is rendered with `p_wiki_xhtml`, block-closing tags are turned
98  into spaces, tags/entities are stripped, whitespace normalised, and each
99  annotation's `exact` is searched with `mb_strpos`. This re-check is the source
100  of truth for deletion, so a stale client can't cause data loss.
101
102## JSINFO injection (important gotcha)
103
104`script.js` needs per-page facts at boot without an extra round-trip, but you
105**cannot** add them by writing `$JSINFO` inside `TPL_METAHEADER_OUTPUT`:
106`tpl_metaheaders()` calls `jsinfo()` and serialises `$JSINFO` into the inline
107`var JSINFO = …;` script **before** firing that event. Instead `handleMetaHeader`
108finds that inline `<script>` in `$event->data['script']` and appends a
109`JSINFO.annotations = {…};` statement so it runs in the same scope. Injection is
110gated to `show` / `export_xhtml` views.
111
112Payload: `{ enabled, pageId, stats, user, isAdmin, token }`. `user`, `isAdmin`
113and `token` are included because stock `JSINFO` exposes no user identity and no
114security token — the script reads them from `JSINFO.annotations`, not from
115`JSINFO.userinfo` (which does not exist) or the `#dw__token` field.
116
117## Per-user toggle
118
119Registered with the **usersettings** plugin via `PLUGIN_USERSETTINGS_REGISTER`
120(key `annotations_enabled`, checkbox, default on). `isEnabledForUser()` reads the
121preference through the usersettings helper; if that plugin is absent, or the
122toggle has not been registered yet, the feature defaults to **on**. When a user
123turns it off, `boot()` returns early and nothing is rendered (annotations are
124still stored).
125
126## Permission model
127
128The rules live in `helper.php` and are pure; `action.php` gathers the facts and
129calls them. `isAdmin` is true for the `admin` group or DokuWiki's `$INFO['isadmin']`.
130
131| Action | Rule (helper method) |
132|--------|----------------------|
133| Create annotation / reply / resolve / reopen | logged in **and** `AUTH_READ` on the page — *not* `AUTH_EDIT` (`canAnnotate`) |
134| Edit / delete own annotation | author (`canEditAnnotation`) |
135| Edit / delete own reply | author (`canEditReply`) |
136| Edit / delete **any** annotation or reply | admin (`canEditAnnotation` / `canEditReply`) |
137| Clear resolved / clear orphaned (per page) | admin (`canClear`) |
138| Load (read) annotations | `AUTH_READ` on the page |
139
140## Security
141
142- **CSRF.** Every state-changing action requires a valid DokuWiki security
143  token. The token is injected into `JSINFO.annotations.token` and sent back as
144  `sectok` in the JSON body. Because `checkSecurityToken()` reads `$_REQUEST`
145  (empty for a JSON body), `handleAjax` copies `sectok` into `$_POST`/`$_REQUEST`
146  before validating. The read-only `load` action is exempt (GET, no token) but
147  still ACL-checked.
148- **ACL.** `auth_quickaclcheck($id)` gates both reading and writing.
149- **Output.** Bodies are stored as plain text (newlines kept, length-capped) and
150  rendered client-side via `textContent`, so user content is never interpolated
151  as HTML.
152
153## AJAX endpoint
154
155`…/lib/exe/ajax.php?call=annotations` (handled on `AJAX_CALL_UNKNOWN`). The
156`load` action is a GET with query params; everything else is `POST` with an
157`application/json` body. Every response is `{ "success": true, … }` or
158`{ "success": false, "error": "…" }`.
159
160| Action | Method | Token | Extra fields |
161|--------|--------|-------|--------------|
162| `load` | GET | — | — |
163| `create` | POST | ✓ | `anchor`, `body` |
164| `reply` | POST | ✓ | `annId`, `body` |
165| `edit_annotation` | POST | ✓ | `annId`, `body` |
166| `edit_reply` | POST | ✓ | `annId`, `replyId`, `body` |
167| `delete_annotation` | POST | ✓ | `annId` |
168| `delete_reply` | POST | ✓ | `annId`, `replyId` |
169| `resolve` | POST | ✓ | `annId`, `status` (`open`\|`resolved`) |
170| `clear_resolved` | POST | ✓ | — |
171| `clear_orphaned` | POST | ✓ | — |
172
173All actions also take the page `id`.
174
175## Constraints
176
177- **JS/CSS floor: Firefox 78 ESR.** No `#private` fields, `??=`/`||=`/`&&=`,
178  `Array.at`, `structuredClone`, `Object.hasOwn`, native `<dialog>`; no CSS
179  `:has()`, selector `:not()`, `aspect-ratio`, container queries, or nesting.
180  `async`/`await`, `fetch`, classes, `?.`, `??`, `Map`/`Set` are fine.
181- **PHP:** developed against 8.3; requires the `mbstring` extension.
182
183## Known gaps / next steps
184
185- **UI localisation.** `script.js` renders hardcoded English; `annInfo.lang` is
186  never populated, so the `counter_*`, `btn_*`, `status_*`, `placeholder_*`,
187  `tooltip_*`, `orphaned_*`, `error_*` and `confirm_*` strings in
188  `lang/en/lang.php` are currently dead. To localise: inject `lang` into the
189  `JSINFO.annotations` payload in `handleMetaHeader` and read `_lang` in the JS
190  string sites. Only `toggle_label` / `toggle_desc` are wired (via `getLang`).
191- **Translations.** No `de` / `ru` / `ja` yet (depends on the localisation work
192  above).
193- **Tests.** No `_test/` suite. Candidates: helper CRUD, input cleaning,
194  permission rules, and `findOrphaned` against a rendered page.
195- **Config.** No `conf/` — nothing is configurable (highlight colours, context
196  length, body cap are all constants/CSS).
197- **Cleanup.** The `ann-highlight-orphaned` JS constant has no CSS rule and no
198  call site (orphans have no in-page range to highlight).
199