feat(extract): give .svelte and .astro a real AST pass via script masking

[RepairYourTech]

feat(extract): give .svelte and .astro a real AST pass via script masking

The problem

extract_svelte and extract_astro feed the raw file to the JS grammar:

result = _extract_generic(path, _JS_CONFIG)   # path is the whole .svelte/.astro file

HTML markup is not valid JS, so tree-sitter produces a top-level ERROR node, the
AST pass is abandoned, and both extractors fall back to regex-scraping imports.
Their own docstrings acknowledge this:

The JS tree-sitter parser fed the full .svelte file produces a top-level
ERROR node (HTML markup is not valid JS), so import_statement nodes are never
reached and static imports are silently dropped (#713).

Imports are rescued by regex, so the gap is invisible in import-shaped tests —
but every function, const, interface and type in those files is missing from
the graph. On a Svelte/Astro codebase that is most of the application.

The fix

graphify already solves exactly this for .vue. _vue_mask_non_script blanks
the markup (preserving \r/\n so line numbers stay accurate) and hands the
result to _extract_generic via source_override — a parameter that exists for
this purpose:

source_override parses the given bytes instead of reading path […] Lets
container formats (e.g. Vue SFCs) mask the wrapper and parse just the embedded
<script>.

Svelte and Astro were simply never wired to it.

• .svelte reuses _vue_mask_non_script verbatim — the <script> shape
  matches Vue's, including lang= grammar selection (tsx / js / ts).
• .astro gains _astro_mask, covering the --- fenced frontmatter plus
  client <script> bodies. _astro_best_mask falls back to frontmatter-only for
  the few files whose client script does not concatenate onto the frontmatter as
  valid TS, so those parse clean rather than degrading.
• _parse_js_tree masks both formats too, so import resolution sees the
  script rather than the markup.

The existing regex rescues are left in place — they still cover template-layer
dynamic imports such as {#await import('./X.svelte')}, which the AST cannot
see, and they dedupe against the AST nodes via existing_ids.

Results

Measured on a Svelte 5 + Astro monorepo (RepairYour.Tech, 1,367 files):

	Files	Nodes before → after	Edges before → after
.svelte	260	931 → 4,307 (4.6×)	1,047 → 6,225
.astro	1,107	2,946 → 6,607 (2.2×)	9,963 → 15,337

Parse cleanliness after masking: 258/258 Svelte, 1,104/1,107 Astro (from zero
clean before — every file errored at line 1).

Regression safety

• Node/edge digests over ts, tsx, rust, python, sql and json samples
  are byte-identical before and after. Only the .svelte/.astro dispatch
  paths and two new _parse_js_tree branches are touched; the .vue path is
  untouched.
• test_vue_extraction.py, test_astro_extraction.py and
  test_astro_import_ids.py all pass.
• Broader extraction sweep: 723 → 731 passing, with the same 231 pre-existing
  failures in my environment (missing optional tree-sitter grammars — identical
  set on a clean checkout).

Tests added

tests/test_svelte_astro_masking.py (8 tests), mirroring
test_vue_extraction.py:

• masks preserve line counts and blank markup/style
• Astro frontmatter-only vs. frontmatter+<script> variants
• a .astro file with no frontmatter masks safely to blank
• extract_svelte / extract_astro recover declarations and keep imports
• the masked pass strictly adds nodes over the unmasked baseline

Note for maintainers

This has no issue number yet — happy to open one first if you'd prefer, and to
add the (#NNNN, thanks @…) reference to the CHANGELOG entry to match the
existing convention.

[graphify-labs]

Graphify reviewed this change.

Looks safe to merge — no coupling regressions and no blocking issues, checked against the code graph (not a self-assessment).

Formal verification. No changes could be formally verified in this run.

---

Graphify review — findings

This pull request modifies the Svelte and Astro extractors so that both mask out the non-code regions of the file (markup/template/style) and parse the remaining script with the TypeScript grammar, rather than feeding the raw file to the JS grammar. It adds new _astro_mask and _astro_best_mask helpers in resolution.py, wires the masking into extract_svelte, extract_astro, and _parse_js_tree, and keeps the existing regex-based import fallback. A new test file exercises the masking helpers and end-to-end extraction for both file types, and a changelog entry describes the change.

No blocking issues surfaced. 2 lower-confidence candidates did not survive cross-model review.

Analysis details — impact, health, verification

Impact & health

Graphify review

Impact — 1839 functions depend on the 529 functions this change touches.

Health — this change adds coupling hotspots:

• worse: _extract_generic() — 20 callers, 23 callees
• worse: extract_astro() — 8 callers, 6 callees
• worse: extract_svelte() — 5 callers, 6 callees

Verification — 1839 functions in the blast radius were not formally verified this run (proofs are advisory here).

Gate & verification

graphify gate

PASS — objectively clean (no health regressions, tests not run — proofs not run this pass (advisory)). Grounded, not self-assessed.

Advisory (not blocking):

• verification_scope: 1699 function(s) in the blast radius were not formally verified this run

Formal verification

Could not verify: Could not verify extract\_astro.

The verifier did not have enough to check extract\_astro, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `path` is annotated `Path` — outside the synthesizable primitive/collection set

Could not verify: Could not verify extract\_svelte.

The verifier did not have enough to check extract\_svelte, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `path` is annotated `Path` — outside the synthesizable primitive/collection set

Could not verify: Could not verify \_parse\_js\_tree.

The verifier did not have enough to check \_parse\_js\_tree, so it is saying so rather than guessing. No false assurance is the whole point.

Guarantee: No guarantee either way, this is an honest abstention, not a pass.

Note: Reason: parameter `path` is annotated `Path` — outside the synthesizable primitive/collection set

· 3 more finding(s) on lines outside this diff (see the check run).