0
0
Fork 0
mirror of https://github.com/discourse/discourse.git synced 2026-08-14 13:58:53 +08:00
discourse/plugins/spoiler-alert/test/javascripts/unit/lib/to-markdown-test.js
Renato Atilio d62cc48d9c
DEV: to-markdown prosemirror HTML parsing (#35563)
Replaces the custom ~1000-line `to-markdown` converter with ProseMirror's schema-based parsing and serialization pipeline, reusing the same extensions that power the rich editor. The old converter duplicated logic the extensions already implement (image formatting, mention/hashtag serialization, table handling, list nesting, etc.) — a single pipeline means one set of rules, fewer divergence bugs, and plugin extensions work in both directions automatically.

### Core change

- `toMarkdown()` is now async and lazy-loads ProseMirror on first use.
- Cooked→markdown serialization runs through the registered rich-editor extensions, so both conversion directions share one source of truth.

### QuoteState and quote callers

Making `toMarkdown()` async changed the quote flow. `QuoteState` now stores the selection's HTML and exposes an async `markdown()`; `buffer` is the plain-text selection. Every caller that needs markdown was updated to await `markdown()`:

- Topic controller (`selectText`, `replyToPost` including EmbedMode, `replyAsNewTopic`, `buildQuoteMarkdown`).
- Nested controller — forwards the full selection state via the new `copyFrom`.
- Fast-edit modal and `computeSupportsFastEdit`.
- discourse-ai post-helper menu — now sends markdown (not plain text) to the AI endpoints.

`markdown()` snapshots `opts` before awaiting so a concurrent selection change can't mis-pair the result.

### Plugin API

- Removed plugin-specific `addTagDecorateCallback`/`addTextDecorateCallback` usage from local-dates and spoiler-alert (the old APIs are kept as deprecated no-ops). local-dates now uses a `transformParsedHTML` rich-editor extension hook.

### Extractions and sharing

- Word paste handling moved into a dedicated `word-paste.js` extension.
- Quote-selection list-structure preservation moved into `preserve-list-structure.js`.
- `normalizeTable` is shared between the paste plugin and the serializer for consistent table output across editor modes.

### Quoting fix

- `selectedHTML` drops the empty trailing block a triple-click leaves behind. A triple-click extends the selection's end to the start of the following block, so `cloneContents()` would otherwise clone an empty `<blockquote>` that serializes to a stray `> ` in the quote.
2026-05-29 18:14:32 -03:00

33 lines
1.1 KiB
JavaScript
Vendored

import { module, test } from "qunit";
import {
registerRichEditorExtension,
resetRichEditorExtensions,
} from "discourse/lib/composer/rich-editor-extensions";
import toMarkdown from "discourse/lib/to-markdown";
import richEditorExtension from "discourse/plugins/spoiler-alert/lib/rich-editor-extension";
module("Spoiler Alert | Unit | to-markdown", function (hooks) {
hooks.beforeEach(async function () {
await resetRichEditorExtensions();
registerRichEditorExtension(richEditorExtension);
});
test("handles spoiler tags", async function (assert) {
let html = `<div>Text with a</div><div class="spoiled spoiler-blurred">spoiled</div><div>word.</div>`;
let markdown = `Text with a\n\n[spoiler]\nspoiled\n\n[/spoiler]\n\nword.`;
assert.strictEqual(
await toMarkdown(html),
markdown,
"creates block spoiler tag"
);
html = `Inline <span class="spoiled">spoiled</span> word.`;
markdown = `Inline [spoiler]spoiled[/spoiler] word.`;
assert.strictEqual(
await toMarkdown(html),
markdown,
"creates inline spoiler tag"
);
});
});