From fe07657fe10103c2750c1c2f08804b4bc70f5644 Mon Sep 17 00:00:00 2001
From: Kevin <bosteels.kevin@gmail.com>
Date: Mon, 28 Sep 2026 21:00:53 +0200
Subject: [PATCH] fix(extract): parse only Astro frontmatter and scripts in the
 AST pass

extract_astro fed the whole .astro file to the JS grammar. The template is
not JS, so every page reported a syntax error (usually at line 1-2) and
symbols tree-sitter could not recover past the template were dropped.

Blank everything outside the frontmatter and JS <script> bodies (keeping
newlines so locations still match, as _vue_mask_non_script does) and parse
with the TS grammar, Astro's default. Scripts with a non-JS type
(application/ld+json, ...) are blanked too. The existing regex import
rescue is unchanged.

Fixes #2788. Covers the unmasked-script half of #3337.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 graphify/extract.py            | 49 ++++++++++++++++++++++++++++-
 tests/test_astro_extraction.py | 56 ++++++++++++++++++++++++++++++++++
 2 files changed, 104 insertions(+), 1 deletion(-)

diff --git a/graphify/extract.py b/graphify/extract.py
index 4b9efa732..a830d44d7 100644
--- a/graphify/extract.py
+++ b/graphify/extract.py
@@ -2405,6 +2405,42 @@ def extract_svelte(path: Path) -> dict:
     return result
 
 
+_ASTRO_FRONTMATTER_RE = re.compile(r"\A\s*---[^\S\r\n]*\r?\n([\s\S]*?)\r?\n---")
+_ASTRO_SCRIPT_RE = re.compile(
+    r"<script\b((?:\"[^\"]*\"|'[^']*'|[^>\"'])*)>([\s\S]*?)</script\s*>",
+    re.IGNORECASE,
+)
+_ASTRO_NON_JS_TYPE_RE = re.compile(
+    r"""\btype\s*=\s*["']?(?!module\b|text/javascript\b|application/javascript\b)""",
+    re.IGNORECASE,
+)
+
+
+def _astro_mask_non_script(src: str) -> str:
+    """Blank everything in a ``.astro`` file except frontmatter and JS ``<script>`` bodies.
+
+    Every character outside those regions becomes a space (``\\r``/``\\n`` are kept),
+    so AST locations match the original file. Scripts with a non-JS ``type``
+    (``application/ld+json`` and the like) are blanked too: their bodies are not
+    statements and would only add parse errors.
+    """
+    keep: list[tuple[int, int]] = []
+    fm = _ASTRO_FRONTMATTER_RE.match(src)
+    if fm:
+        keep.append((fm.start(1), fm.end(1)))
+    for m in _ASTRO_SCRIPT_RE.finditer(src, fm.end() if fm else 0):
+        if not _ASTRO_NON_JS_TYPE_RE.search(m.group(1)):
+            keep.append((m.start(2), m.end(2)))
+    chars = [c if c in "\r\n" else " " for c in src]
+    for start, end in keep:
+        chars[start:end] = src[start:end]
+        # Terminate the region in place of the following `<`, so two scripts on
+        # one line don't run together into a single statement.
+        if end < len(chars) and chars[end] == " ":
+            chars[end] = ";"
+    return "".join(chars)
+
+
 def extract_astro(path: Path) -> dict:
     """Extract imports from .astro files: frontmatter (TS) + template regex fallback.
 
@@ -2417,8 +2453,19 @@ def extract_astro(path: Path) -> dict:
     silently dropped (#850). Mirrors :func:`extract_svelte` — same regex-rescue
     approach, scanning the frontmatter block and any client-side ``<script>`` blocks
     for static and dynamic imports.
+
+    The AST pass parses only the frontmatter and JS ``<script>`` bodies with the TS
+    grammar (Astro's default), blanking everything else so line numbers still
+    line up — the same masking as :func:`_vue_mask_non_script`. Parsing the whole
+    file flagged every template as a syntax error and dropped frontmatter symbols.
     """
-    result = _extract_generic(path, _JS_CONFIG)
+    try:
+        src = path.read_text(encoding="utf-8", errors="replace")
+    except OSError:
+        return {"nodes": [], "edges": []}
+    masked = _astro_mask_non_script(src).encode("utf-8")
+    masked = _normalize_ts_import_types(masked) or masked
+    result = _extract_generic(path, _TS_CONFIG, source_override=masked)
     try:
         import re as _re
         src = path.read_text(encoding="utf-8", errors="replace")
diff --git a/tests/test_astro_extraction.py b/tests/test_astro_extraction.py
index c21e66c24..2b4fbb344 100644
--- a/tests/test_astro_extraction.py
+++ b/tests/test_astro_extraction.py
@@ -141,3 +141,59 @@ import Hero from '@components/Hero.astro';
     result = extract_astro(page)
     targets = _import_targets(result, relation="imports_from")
     assert _make_id(str(hero)) in targets
+
+
+def _labels(result: dict) -> dict[str, str]:
+    """Label -> source_location for every node except the file node."""
+    return {
+        str(n.get("label")): str(n.get("source_location"))
+        for n in result.get("nodes", [])
+        if n.get("id") != result.get("nodes", [{}])[0].get("id")
+    }
+
+
+def test_extract_astro_template_is_not_a_parse_error(tmp_path):
+    """The template is not TS; only frontmatter and <script> bodies are parsed (#2788).
+
+    Parsing the whole file as JS reported every page as a syntax error, and
+    symbols the parser could not recover past the template were dropped.
+    """
+    page = _write(
+        tmp_path / "src/pages/results.astro",
+        """---
+import Layout from '../layouts/Layout.astro';
+interface Props { year: number }
+const { year } = Astro.props;
+function rank(scores: number[]): number[] {
+  return [...scores].sort((a, b) => b - a);
+}
+---
+
+<Layout title={`Results ${year}`}>
+  <ol>{rank([3, 1, 2]).map((s) => <li class="score">{s}</li>)}</ol>
+</Layout>
+<script type="application/ld+json">{ "@type": "Event", "name": "x" }</script>
+<script>
+  function toggle(el: HTMLElement) { el.classList.toggle('open'); }
+  document.querySelectorAll('ol').forEach((el) => toggle(el));
+</script>
+""",
+    )
+    result = extract_astro(page)
+    assert result.get("parse_errors") is None
+    labels = _labels(result)
+    assert "rank()" in labels and "toggle()" in labels
+    # Masking keeps offsets, so locations still point at the original lines.
+    assert labels["rank()"] == "L5"
+    assert labels["toggle()"] == "L15"
+
+
+def test_extract_astro_scripts_on_one_line_do_not_merge(tmp_path):
+    """Two script bodies on one line must not parse as a single statement."""
+    page = _write(
+        tmp_path / "src/pages/inline.astro",
+        "<script>const a = 1</script><script>function b() {}</script>\n",
+    )
+    result = extract_astro(page)
+    assert result.get("parse_errors") is None
+    assert "b()" in _labels(result)