mirror of
https://github.com/heygen-com/hyperframes.git
synced 2026-09-03 12:54:29 +00:00
## What
Adds `ensureHfIds(html: string): string` in `packages/core/src/parsers/hfIds.ts`. Single DOM pass (via linkedom) that mints a `data-hf-id` attribute on every eligible element before the caller sees the markup.
**Id derivation:** FNV-1a 32-bit hash of `tagName | sorted-attrs(\x00/\x01 separated) | ownText`, last 4 chars of base-36, `hf-` prefix. Collision resolution appends a sibling counter and re-hashes. Preserves existing ids (elements with `data-hf-id` already set are skipped). Excludes non-visual tags: `script`, `style`, `template`, `meta`, `link`, `noscript`, `base`.
**Fragment handling:** detects bare HTML fragments (no `<!doctype` / `<html`) and wraps in a full document shell before parsing, then returns `body.innerHTML` — matching the pattern used by `parseSourceDocument` in `sourceMutation`.
## Why
Counter-based ids (`element-0`, `element-1`, …) are positional. Inserting a new layer at position 0 shifts every id below it. The R1 milestone requires content-based, stable ids so that targeting operations (split, patch, probe) stay valid across re-parses and element insertions. T2 spec (`stableIds.test.ts`) defines the contract: same content → same id, adding a sibling doesn't change other ids, format matches `/^hf-[a-z0-9]{4}$/`.
## How
- `toHfId(hash)` — `slice(-4)` of `hash.toString(36)` for better distribution across the suffix space
- `data-hf-id` is excluded from the hash input (prevents circular dependency on the attribute being set)
- Already-assigned ids tracked in a `Set`; duplicates get a counter suffix before re-hashing
- Fragment detection: `/<!doctype|<html[\s>]/i.test(html)` — if bare, wrap→parse→`body.innerHTML`
## Test plan
- [x] T2 spec: `packages/core/src/parsers/stableIds.test.ts` — 7 tests pass (3 were `.fails` stubs targeting R1; 4 were pre-existing baselines that must not regress)
- [x] No changes to existing htmlParser tests needed at this layer (wiring is PR #1270)