89eff9d · favela-markdown

Changed files (4) +47 -5

lib/strip.js +5 -2

…	…	@@ -1,7 +1,7 @@
1	1	// strip(source) -> { text, facets } — spec/markdown-strip.md
2	2	//
3	3	// The strip's output is what bsky sees and what a CID commits to, so every rule here cites the
4		// spec section it implements and nothing is left to "whatever the parser does" (§1). The 36
	4	// spec section it implements and nothing is left to "whatever the parser does" (§1). The conformance
5	5	// vectors are the oracle; this file exists to reproduce them byte-for-byte.
6	6	//
7	7	// Structure, and why:
…	…	@@ -32,7 +32,10 @@ import { directive } from 'micromark-extension-directive'
32	32	import { directiveFromMarkdown } from 'mdast-util-directive'
33	33
34	34	const MICROMARK = {
35		extensions: [gfmStrikethrough(), gfmTable(), gfmFootnote(), directive()],
	35	// ⛔ singleTilde:false — micromark defaults it TRUE, which would consume `~x~` as strikethrough.
	36	// §2.1 pins the extension as `~~x~~`, and the dialect is CLOSED: syntax outside it is text, not
	37	// markup. Caught by the `single-tilde-is-literal` vector, which exists because of that rule.
	38	extensions: [gfmStrikethrough({ singleTilde: false }), gfmTable(), gfmFootnote(), directive()],
36	39	mdastExtensions: [
37	40	gfmStrikethroughFromMarkdown(),
38	41	gfmTableFromMarkdown(),

spec/markdown-strip.md +18 -0

…	…	@@ -47,6 +47,24 @@ strip(source: string) -> { text: string, facets: Facet[] }
47	47	An implementation MUST NOT enable further extensions. A new one is a spec change, because it
48	48	changes the bytes of every record that uses it.
49	49
	50	⛔ The dialect is CLOSED, so syntax outside it is TEXT, not markup. Decided 2026-08-11
	51	(hyper-social #43) as "unknown syntax → literal text", written 2026-09-05. `==highlight==`,
	52	`$$x$$`, `{: .cls}`, a task-list checkbox and a definition list are all literal text, because
	53	they are not in the table above. This is not leniency — it is what "the dialect is exactly these six"
	54	means, and it is the property that lets a future extension be added without reinterpreting a single
	55	existing record.
	56
	57	⚠ It also constrains how the parser is CONFIGURED, not just which extensions are enabled. A
	58	parser option that widens a pinned syntax silently breaks the rule: micromark's `gfmStrikethrough`
	59	defaults `singleTilde: true`, which consumes `~x~` even though §2.1 pins strikethrough as `~~x~~`.
	60	That defect shipped in `lib/strip.js` and was caught only by writing the vector for this rule
	61	(`single-tilde-is-literal`). **Every parser option that affects accepted syntax is part of the
	62	dialect and must be pinned explicitly.**
	63
	64	One exception, and only one: an unknown directive name still yields its contents (§4.8), because
	65	`:::name` itself IS in the dialect — only the name is unrecognised. `unknown-directive-keeps-contents`
	66	pins it.
	67
50	68	---
51	69
52	70	## 3. Inline rules

spec/vectors/markdown-strip.json +22 -1

…	…	@@ -329,6 +329,27 @@
329	329	"value": "https://two.example"
330	330	}
331	331	]
	332	},
	333	{
	334	"name": "unknown-syntax-is-literal",
	335	"why": "§2.1 the dialect is closed, so syntax outside it is TEXT, never markup (hyper-social #43)",
	336	"source": "==highlight== and $$x$$ and {: .cls}",
	337	"text": "==highlight== and $$x$$ and {: .cls}",
	338	"facets": []
	339	},
	340	{
	341	"name": "single-tilde-is-literal",
	342	"why": "§2.1 strikethrough is pinned as ~~x~~; a SINGLE tilde is outside the dialect and stays text",
	343	"source": "~sub~ not strikethrough",
	344	"text": "~sub~ not strikethrough",
	345	"facets": []
	346	},
	347	{
	348	"name": "unknown-directive-keeps-contents",
	349	"why": "§4.8 the ONE exception — an unknown directive NAME still yields its contents, not literal text",
	350	"source": ":::unknownName\nbody\n:::",
	351	"text": "body",
	352	"facets": []
332	353	}
333	354	]
334		}
	355	}

test/strip.test.js +2 -2

…	…	@@ -32,8 +32,8 @@ function expandFacet (f) {
32	32
33	33	const utf8 = (s) => Buffer.from(s, 'utf8')
34	34
35		test('strip: all 36 vectors reproduce byte-for-byte', (t) => {
36		t.is(VECTORS.length, 36, 'the vector set is the one the spec pins')
	35	test('strip: every vector reproduces byte-for-byte', (t) => {
	36	t.ok(VECTORS.length >= 36, 'the pinned vector set may grow, never shrink')
37	37
38	38	for (const v of VECTORS) {
39	39	const got = strip(v.source)

favela-markdown commits / 89eff9d

Changed files (4) +47 -5