Skip to content

armgit / fider

Public Imported from GitHub
0

Open platform to collect and prioritize feedback

fix(ui): hide "<" with a placeholder instead of disabling HTML tokenizers

Matt Robertscb9de449b59c
diff --git a/public/services/markdown.spec.ts b/public/services/markdown.spec.ts
index d773423e0..872d0cef2 100644
--- a/public/services/markdown.spec.ts
+++ b/public/services/markdown.spec.ts
@@ -238,10 +238,28 @@ describe("text is shown exactly as typed", () => {
expect(shown(markdown.plainText("`List<String>` and `a && b`"))).toEqual("List<String> and a && b")
expect(markdown.toText("`List<String>` and `a && b`")).toEqual("List<String> and a && b")
})
+})
+
+describe("text that starts with < keeps its block structure", () => {
+ // "<" must not reach marked's HTML-block rules, which would end paragraphs and list items early
+ const cases: [string, string][] = [
+ ["- Use a map\n<String, Int> as the key", "<ul>\n<li>Use a map<br>&lt;String, Int&gt; as the key</li>\n</ul>"],
+ ["a\n<div>b</div>", "<p>a<br>&lt;div&gt;b&lt;/div&gt;</p>"],
+ ["Steps:\n<p>para</p>", "<p>Steps:<br>&lt;p&gt;para&lt;/p&gt;</p>"],
+ ["- item\n<!-- c -->", "<ul>\n<li>item<br>&lt;!-- c --&gt;</li>\n</ul>"],
+ ]
+
+ cases.forEach(([input, expected]) => {
+ test(`full: ${JSON.stringify(input)}`, () => {
+ expect(markdown.full(input)).toEqual(expected)
+ })
+ })
- test("angle-bracket autolinks become links, as in the editor", () => {
- expect(markdown.full("<https://example.com>")).toEqual(
- '<p><a class="text-link" href="https://example.com" rel="noopener nofollow" target="_blank">https://example.com</a></p>'
+ test("angle brackets stay visible text, not <...> autolinks", () => {
+ // (the bare email is still linked by GFM's email autolinking, as before)
+ expect(markdown.full("<scheme:foo> and <[email protected]>")).toEqual(
+ '<p>&lt;scheme:foo&gt; and &lt;<a class="text-link" href="mailto:[email protected]" rel="noopener nofollow" target="_blank">[email protected]</a>&gt;</p>'
)
+ expect(markdown.toText("Contact <[email protected]> about X")).toEqual("Contact <[email protected]> about X")
})
})
diff --git a/public/services/markdown.ts b/public/services/markdown.ts
index 502da22c6..a121d966a 100644
--- a/public/services/markdown.ts
+++ b/public/services/markdown.ts
@@ -69,19 +69,9 @@ const fullRenderer: RendererObject = {
const rendered = "tokens" in token && token.tokens ? this.parser.parseInline(token.tokens) : escapeText(token.text)
return renderMentions(rendered)
},
- html({ text }) {
- // Not reached (raw HTML isn't tokenized, see noRawHTML), but never emit it unescaped.
- return escapeText(text)
- },
}
-
-// Raw HTML is never rendered: with its tokenizers off, "<b>" is just text and is escaped once
-// by the text/code renderers. (Escaping "<" before parsing instead made code escape it twice,
-// showing `List<String>` as `List&lt;String>`, and broke <https://...> autolinks.)
-const noRawHTML = { tokenizer: { html: () => undefined, tag: () => undefined } }
-
const markedFull = new Marked({ gfm: true, breaks: true })
-markedFull.use(noRawHTML, { renderer: fullRenderer })
+markedFull.use({ renderer: fullRenderer })
// Plain-text renderer: strip all formatting down to readable text.
const plainTextRenderer: RendererObject = {
@@ -123,7 +113,17 @@ const plainTextRenderer: RendererObject = {
},
}
const markedPlainText = new Marked({ gfm: true, breaks: true })
-markedPlainText.use(noRawHTML, { renderer: plainTextRenderer })
+markedPlainText.use({ renderer: plainTextRenderer })
+
+// Raw HTML is never rendered: "<" is swapped for a placeholder before parsing, so marked never
+// sees tags, HTML blocks or <...> autolinks (and they can't change where paragraphs and list items
+// end), then the placeholder becomes "&lt;" after rendering. Swapping in "&lt;" itself made code
+// escape it a second time, showing `List<String>` as `List&lt;String>`. (A placeholder character
+// already in the input is replaced with U+FFFD so it can't be mistaken for an escaped "<".)
+const LT_PLACEHOLDER = "\uE000"
+const hideLessThan = (s: string) => s.replace(/\uE000/g, "\uFFFD").replace(/</g, LT_PLACEHOLDER)
+const restoreLessThan = (html: string) => html.replace(/\uE000/g, "&lt;")
+const render = (marked: Marked, input: string) => restoreLessThan(marked.parse(hideLessThan(input)) as string).trim()
const stripTags = (input: string) => input.replace(/<[^>]*>/g, "")
const sanitize = (input: string) => (DOMPurify.isSupported ? DOMPurify.sanitize(input) : stripTags(input))
@@ -143,12 +143,12 @@ const decodeHtmlEntities = (text: string): string => {
}
export const full = (input: string): string => {
- return sanitize((markedFull.parse(input) as string).trim())
+ return sanitize(render(markedFull, input))
}
// HTML with the formatting stripped, safe to insert as HTML (<Markdown style="plainText" />)
export const plainText = (input: string): string => {
- return sanitize((markedPlainText.parse(input) as string).trim())
+ return sanitize(render(markedPlainText, input))
}
// Plain text with entities decoded (e.g. auto-generated titles). Never insert the result as HTML