fix(catalog): explicit section ownership, no body-sniffing heuristic

carriedSectionsFrom() decided whether a ## Usage section was generated by matching
its first line against a list of historical opener phrases — so a hand-written
Usage section that happened to open that way was classified as generated and
silently deleted on regeneration. Ownership is now purely set membership: a
section is generated iff its heading is one the template emits, and ambiguous
'usage' is no longer in that set (the template never emits it), so any ## Usage is
carried. Exported carriedSectionsFrom behind an entrypoint guard and added two
executable preservation fixtures. Flagged by Magi (#5).
This commit is contained in:
ukimsanov
2026-08-05 10:23:37 -07:00
parent c9fd7465c9
commit d373c3f4a0
2 changed files with 95 additions and 25 deletions
@@ -1,7 +1,9 @@
import { readdirSync, readFileSync } from "node:fs"; import { mkdtempSync, readdirSync, readFileSync, rmSync, writeFileSync } from "node:fs";
import { tmpdir } from "node:os";
import { dirname, join, resolve } from "node:path"; import { dirname, join, resolve } from "node:path";
import { fileURLToPath } from "node:url"; import { fileURLToPath } from "node:url";
import { describe, expect, it } from "vitest"; import { afterAll, describe, expect, it } from "vitest";
import { carriedSectionsFrom } from "../../../../scripts/generate-catalog-pages.ts";
const here = dirname(fileURLToPath(import.meta.url)); const here = dirname(fileURLToPath(import.meta.url));
const repoRoot = resolve(here, "../../../.."); const repoRoot = resolve(here, "../../../..");
@@ -39,3 +41,77 @@ describe("catalog pages keep the required reader continuation", () => {
expect(lastHeading).toBe("## Related topics"); expect(lastHeading).toBe("## Related topics");
}); });
}); });
// The carry-forward path is the subtler half: a regeneration must preserve every
// hand-written section, including one whose heading (## Usage) the generator once
// emitted and whose first line looks like generated prose, and one appended below
// the generated footer marker.
describe("carriedSectionsFrom preserves hand-written sections", () => {
const dir = mkdtempSync(join(tmpdir(), "hf-catalog-carry-"));
afterAll(() => rmSync(dir, { recursive: true, force: true }));
const write = (name: string, body: string) => {
const p = join(dir, name);
writeFileSync(p, body, "utf-8");
return p;
};
it("keeps a custom Usage section that opens like generated prose", () => {
const page = write(
"usage.mdx",
[
"## Install",
"",
"```bash Terminal",
"npx hyperframes add sample",
"```",
"",
"## Usage",
"",
// deliberately opens with a phrase an old generated Usage used
"After installing, add the block — but here is our own hand-written guidance the author cares about.",
"",
"{/* hf:generated-footer */}",
"",
"Tagged `sample`.",
"",
].join("\n"),
);
const carried = carriedSectionsFrom(page);
const joined = carried.sections.join("\n");
expect(joined).toContain("## Usage");
expect(joined).toContain("hand-written guidance the author cares about");
expect(carried.hasCustomUsage).toBe(true);
// The generated Install section is owned by the template and not carried.
expect(joined).not.toContain("## Install");
});
it("keeps a hand-written section appended below the generated footer", () => {
const page = write(
"below-footer.mdx",
[
"## Install",
"",
"text",
"",
"{/* hf:generated-footer */}",
"",
"Tagged `sample`.",
"",
"## Related topics",
"",
"- [Browse the complete Catalog](/catalog)",
"",
"## Field notes",
"",
"A section a human added after the generated tail.",
"",
].join("\n"),
);
const joined = carriedSectionsFrom(page).sections.join("\n");
expect(joined).toContain("## Field notes");
expect(joined).toContain("A section a human added after the generated tail.");
// The generator's own Related topics is not carried (it re-emits it).
expect(joined).not.toContain("## Related topics");
});
});
+16 -22
View File
@@ -15,7 +15,7 @@
import { readFileSync, existsSync, mkdirSync, writeFileSync, rmSync } from "node:fs"; import { readFileSync, existsSync, mkdirSync, writeFileSync, rmSync } from "node:fs";
import { join, resolve, dirname } from "node:path"; import { join, resolve, dirname } from "node:path";
import { fileURLToPath } from "node:url"; import { fileURLToPath, pathToFileURL } from "node:url";
// Import from source — bun workspace linking doesn't resolve for scripts outside packages/. // Import from source — bun workspace linking doesn't resolve for scripts outside packages/.
import { import {
type FileTarget, type FileTarget,
@@ -118,10 +118,12 @@ const GENERATED_HEADINGS = new Set([
"ask an agent for it", "ask an agent for it",
"make the texture move", "make the texture move",
"every texture", "every texture",
// headings earlier revisions emitted — dropped on purpose, never carried // headings earlier revisions emitted — dropped on purpose, never carried.
// `usage` is deliberately NOT listed: the current template never emits it, and
// it is a heading a human might reasonably write, so ownership stays explicit
// (anything not in this set is hand-written) rather than sniffing the body.
"details", "details",
"files", "files",
"usage",
"source prompt", "source prompt",
"agent usage", "agent usage",
"animated texture", "animated texture",
@@ -130,19 +132,6 @@ const GENERATED_HEADINGS = new Set([
"related topics", "related topics",
]); ]);
/**
* Openers of every "## Usage" body this generator has written. A Usage section
* that starts with none of them was rewritten by hand, so it is kept and the
* generated usage prose steps aside for it.
*/
const GENERATED_USAGE_OPENERS = [
/^After installing, add the block/,
/^Open `[^`]+` and paste its contents/,
/^Open `[^`]+` and copy what is inside/,
/^After `npx hyperframes add/,
/^It runs for /,
];
/** /**
* Marks the start of the generated provenance footer (tags, credit, prompt). * Marks the start of the generated provenance footer (tags, credit, prompt).
* That footer carries no heading of its own, so without this marker the section * That footer carries no heading of its own, so without this marker the section
@@ -170,8 +159,10 @@ const RELATED_TOPICS: readonly string[] = [
* Pull the hand-written `## sections` out of an already-generated page. * Pull the hand-written `## sections` out of an already-generated page.
* Returns the raw lines, heading included, in their original order. * Returns the raw lines, heading included, in their original order.
*/ */
// Exported for the preservation fixture in
// packages/core/src/registry/catalogGeneratorInstructions.test.ts.
// fallow-ignore-next-line complexity // fallow-ignore-next-line complexity
function carriedSectionsFrom(pagePath: string): CarriedContent { export function carriedSectionsFrom(pagePath: string): CarriedContent {
const empty: CarriedContent = { sections: [], hasCustomUsage: false }; const empty: CarriedContent = { sections: [], hasCustomUsage: false };
if (!existsSync(pagePath)) return empty; if (!existsSync(pagePath)) return empty;
let text: string; let text: string;
@@ -192,12 +183,12 @@ function carriedSectionsFrom(pagePath: string): CarriedContent {
while (buffer.length && buffer[0]!.trim() === "") buffer.shift(); while (buffer.length && buffer[0]!.trim() === "") buffer.shift();
while (buffer.length && buffer.at(-1)!.trim() === "") buffer.pop(); while (buffer.length && buffer.at(-1)!.trim() === "") buffer.pop();
// Ownership is explicit: a section is generated iff its heading is one the
// template emits (GENERATED_HEADINGS). Everything else is hand-written and
// carried verbatim — no content heuristic that could misread custom prose as
// generated and silently delete it on the next regeneration.
const key = heading.toLowerCase(); const key = heading.toLowerCase();
const handWritten = if (!GENERATED_HEADINGS.has(key) && buffer.length) {
!GENERATED_HEADINGS.has(key) ||
(key === "usage" && !GENERATED_USAGE_OPENERS.some((re) => re.test(buffer[0] ?? "")));
if (handWritten && buffer.length) {
if (key === "usage") hasCustomUsage = true; if (key === "usage") hasCustomUsage = true;
sections.push(`## ${heading}`, "", ...buffer, ""); sections.push(`## ${heading}`, "", ...buffer, "");
} }
@@ -844,4 +835,7 @@ function main(): void {
console.log("\nDone."); console.log("\nDone.");
} }
// Only regenerate when run directly, so the module can be imported by tests.
if (import.meta.url === pathToFileURL(process.argv[1] ?? "").href) {
main(); main();
}