From f9b42e8d23ad93f790be76d960e7dbe941de4c33 Mon Sep 17 00:00:00 2001
From: Agnik47 <140933190+Agnik47@users.noreply.github.com>
Date: Fri, 2 Oct 2026 10:35:31 +0530
Subject: [PATCH] fix(extract): parse only Svelte script blocks in the AST pass
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

extract_svelte fed the whole .svelte file to the JS grammar. The template
is not JS, so every component reported a syntax error at line 1 and no
symbol declared inside <script> was ever reached — only imports survived,
via the regex rescue pass.

Blank everything outside the <script> bodies (keeping newlines so
locations still match) and parse with the grammar the first block's lang
implies: js/jsx -> JS, ts or unset -> TS, a superset of JS and Svelte's
own default. <script module> and <script context="module"> are parsed
alongside the instance block. Scripts with a non-JS type
(application/ld+json, ...) are blanked too, and a script region is
terminated in place of the following `<` so two blocks on one line do not
run together. The existing regex import rescue is unchanged; an import
both passes resolve yields one edge after build.dedupe_edges, as on the
Astro path.

The masker mirrors _astro_mask_non_script and shares the script-tag and
non-JS-type patterns with it instead of duplicating them.

Fixes #3928. Covers the Svelte half of #3942 (markup is blanked, so a
template {#if} no longer mints a spurious if() node).

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
(cherry picked from commit b64806924e270183d7e9ffd828c0ab3f494ed058)
---
 graphify/extract.py             |  97 ++++++++++++++----
 tests/test_svelte_extraction.py | 173 ++++++++++++++++++++++++++++++++
 2 files changed, 250 insertions(+), 20 deletions(-)
 create mode 100644 tests/test_svelte_extraction.py

diff --git a/graphify/extract.py b/graphify/extract.py
index a11e4261d..654bef59e 100644
--- a/graphify/extract.py
+++ b/graphify/extract.py
@@ -154,6 +154,7 @@ from graphify.extractors.resolution import (  # noqa: E402,F401
     _ts_collect_type_refs,
     _ts_heritage_clause_entries,
     _ts_walk_class_members,
+    _VUE_SCRIPT_LANG_RE as _SCRIPT_LANG_RE,
     _vue_mask_non_script,
     _walk_js_tree,
     _walk_python_tree,
@@ -2380,17 +2381,79 @@ def _emit_rescued_import(
     existing_ids.add(node_id)
 
 
-def extract_svelte(path: Path) -> dict:
-    """Extract imports from .svelte files: script-block via JS AST + template regex fallback.
+_HTML_SCRIPT_TAG_RE = re.compile(
+    r"<script\b((?:\"[^\"]*\"|'[^']*'|[^>\"'])*)>([\s\S]*?)</script\s*>",
+    re.IGNORECASE,
+)
+_NON_JS_SCRIPT_TYPE_RE = re.compile(
+    r"""\btype\s*=\s*["']?(?!module\b|text/javascript\b|application/javascript\b)""",
+    re.IGNORECASE,
+)
 
-    Tree-sitter only sees the <script> block. Svelte template syntax like
-    {#await import('./X.svelte')} lives in the markup layer and is invisible
-    to the JS parser, so a regex pass covers those dynamic imports.
+
+def _svelte_mask_non_script(src: str) -> tuple[str, str | None]:
+    """Blank everything in a ``.svelte`` file except JS ``<script>`` bodies.
+
+    Every character outside those bodies 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. Returns
+    ``(masked_source, lang)``; ``lang`` is the first JS block's declared
+    ``lang``. Mirrors :func:`_astro_mask_non_script`.
     """
-    result = _extract_generic(path, _JS_CONFIG)
+    chars = [c if c in "\r\n" else " " for c in src]
+    lang: str | None = None
+    for m in _HTML_SCRIPT_TAG_RE.finditer(src):
+        if _NON_JS_SCRIPT_TYPE_RE.search(m.group(1)):
+            continue
+        start, end = m.start(2), m.end(2)
+        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] = ";"
+        if lang is None:
+            lang_m = _SCRIPT_LANG_RE.search(m.group(1))
+            if lang_m:
+                lang = lang_m.group(1).lower()
+    return "".join(chars), lang
+
+
+def extract_svelte(path: Path) -> dict:
+    """Extract imports and symbols from .svelte files: script-block AST + template regex fallback.
+
+    The AST pass parses only the ``<script>`` bodies, blanking the markup and
+    style regions so line numbers still line up — the same masking
+    :func:`extract_vue` and :func:`extract_astro` use. Feeding the whole file to
+    the JS grammar made the template a top-level ERROR at line 1, so every
+    ``.svelte`` file was flagged as a syntax error and no symbol inside
+    ``<script>`` (functions, classes, runes) was ever reached; only imports
+    survived, via the regex rescue below (#3928).
+
+    The grammar follows the first block's declared ``lang`` (``js``/``jsx``→JS,
+    ``ts`` or unset→TS, a superset of JS), matching Svelte's own default. Both
+    ``<script>`` and ``<script module>`` / ``<script context="module">`` blocks
+    are parsed. A ``<script>`` carrying a non-JS ``type`` (``application/ld+json``
+    and the like) is masked out rather than parsed as code.
+
+    Svelte template syntax like {#await import('./X.svelte')} lives in the markup
+    layer and is invisible to the JS parser, so a regex pass covers those
+    dynamic imports.
+    """
+    try:
+        src = path.read_text(encoding="utf-8", errors="replace")
+    except OSError:
+        return {"nodes": [], "edges": []}
+
+    masked, lang = _svelte_mask_non_script(src)
+    config = _JS_CONFIG if lang in ("js", "jsx") else _TS_CONFIG
+    masked_bytes = masked.encode("utf-8")
+    if config is _TS_CONFIG:
+        masked_bytes = _normalize_ts_import_types(masked_bytes) or masked_bytes
+
+    result = _extract_generic(path, config, source_override=masked_bytes)
     try:
         import re as _re
-        src = path.read_text(encoding="utf-8", errors="replace")
         existing_ids = {n["id"] for n in result.get("nodes", [])}
         # Source file node ID must match the one _extract_generic creates:
         # _make_id(str(path)) - single arg, no stem prefix. Otherwise the source
@@ -2410,11 +2473,11 @@ def extract_svelte(path: Path) -> dict:
                 result, existing_ids, file_node_id, path, raw,
                 "dynamic_import", aliases, base_url,
             )
-        # Static imports inside <script> blocks. 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). Regex over each script
-        # body recovers them.
+        # Static imports inside <script> blocks (#713). The masked AST pass now
+        # reaches these import_statement nodes itself, so this is a fallback for
+        # a script body the grammar still cannot parse. An import both passes
+        # resolve yields the same (source, target, relation) edge twice, which
+        # build.dedupe_edges collapses — same as the Astro path (#3902).
         script_re = _re.compile(
             r"<script\b[^>]*>([\s\S]*?)</script\s*>", _re.IGNORECASE
         )
@@ -2437,14 +2500,8 @@ def extract_svelte(path: Path) -> dict:
 
 
 _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,
-)
+_ASTRO_SCRIPT_RE = _HTML_SCRIPT_TAG_RE
+_ASTRO_NON_JS_TYPE_RE = _NON_JS_SCRIPT_TYPE_RE
 
 
 def _astro_mask_non_script(src: str) -> str:
diff --git a/tests/test_svelte_extraction.py b/tests/test_svelte_extraction.py
new file mode 100644
index 000000000..bd6595edf
--- /dev/null
+++ b/tests/test_svelte_extraction.py
@@ -0,0 +1,173 @@
+"""Tests for `.svelte` extraction (#3928).
+
+A `.svelte` file is markup with one or two `<script>` blocks. Tree-sitter fed
+the whole file produces a top-level ERROR node at line 1 because the template is
+not JS, so the AST pass never reached the `function_declaration` nodes inside
+`<script>` — every component was reported as a syntax error and only its
+imports survived, via the regex rescue pass. :func:`extract_svelte` masks the
+markup and style regions and parses just the script bodies, the same strategy
+:func:`extract_vue` and :func:`extract_astro` use.
+"""
+from __future__ import annotations
+
+from pathlib import Path
+
+from graphify.extract import _make_id, extract_svelte
+
+
+def _write(path: Path, body: str) -> Path:
+    path.parent.mkdir(parents=True, exist_ok=True)
+    path.write_text(body, encoding="utf-8")
+    return path
+
+
+def _labels(result: dict) -> dict[str, str]:
+    """Label -> source_location for every node except the file node."""
+    file_id = result.get("nodes", [{}])[0].get("id")
+    return {
+        str(n.get("label")): str(n.get("source_location"))
+        for n in result.get("nodes", [])
+        if n.get("id") != file_id
+    }
+
+
+def _import_targets(result: dict, *, relation: str | None = None) -> set[str]:
+    return {
+        str(e.get("target") or "")
+        for e in result.get("edges", [])
+        if relation is None or e.get("relation") == relation
+    }
+
+
+def test_extract_svelte_template_is_not_a_parse_error(tmp_path):
+    """The template is not JS; only `<script>` bodies are parsed (#3928).
+
+    Parsing the whole file reported every component as a syntax error at line 1
+    and dropped every symbol declared inside `<script>`.
+    """
+    component = _write(
+        tmp_path / "src/A.svelte",
+        """<script>
+  import { foo } from "./x";
+  let n = 1;
+  function bump() { n++; }
+</script>
+<button onclick={bump}>{n}</button>
+""",
+    )
+    result = extract_svelte(component)
+    assert result.get("parse_errors") is None
+    labels = _labels(result)
+    assert "bump()" in labels
+    # Masking keeps offsets, so locations still point at the original lines.
+    assert labels["bump()"] == "L4"
+
+
+def test_extract_svelte_parses_module_and_instance_scripts(tmp_path):
+    """Svelte 5 `<script module>` and the instance block are both code."""
+    component = _write(
+        tmp_path / "src/Both.svelte",
+        """<script module>
+  export function shared() { return 1; }
+</script>
+<script>
+  function local() { return 2; }
+</script>
+<p>hi</p>
+""",
+    )
+    result = extract_svelte(component)
+    assert result.get("parse_errors") is None
+    labels = _labels(result)
+    assert "shared()" in labels and "local()" in labels
+    assert labels["shared()"] == "L2"
+    assert labels["local()"] == "L5"
+
+
+def test_extract_svelte_lang_ts_parses_type_syntax(tmp_path):
+    """`<script lang="ts">` must be parsed with the TS grammar, not JS."""
+    component = _write(
+        tmp_path / "src/Typed.svelte",
+        """<script lang="ts">
+  interface Props { year: number }
+  let n = $state(1);
+  function bump(): void { n++; }
+</script>
+<button onclick={bump}>{n}</button>
+""",
+    )
+    result = extract_svelte(component)
+    assert result.get("parse_errors") is None
+    labels = _labels(result)
+    assert "Props" in labels
+    assert "bump()" in labels
+
+
+def test_extract_svelte_non_js_script_type_is_not_parsed_as_code(tmp_path):
+    """`<script type="application/ld+json">` is data, not statements."""
+    component = _write(
+        tmp_path / "src/Ld.svelte",
+        """<script type="application/ld+json">{ "@type": "Event", "name": "x" }</script>
+<script>
+  function go() { return 1; }
+</script>
+<p>hi</p>
+""",
+    )
+    result = extract_svelte(component)
+    assert result.get("parse_errors") is None
+    assert "go()" in _labels(result)
+
+
+def test_extract_svelte_static_import_still_resolves(tmp_path):
+    """Masking must not cost the imports the regex rescue already recovered."""
+    component = _write(
+        tmp_path / "src/Imports.svelte",
+        """<script>
+  import { helper } from "./helper";
+</script>
+<p>hi</p>
+""",
+    )
+    helper = _write(tmp_path / "src/helper.ts", "export function helper(){}\n")
+
+    result = extract_svelte(component)
+    assert _make_id(str(helper)) in _import_targets(result, relation="imports_from")
+
+
+def test_extract_svelte_dynamic_import_in_template(tmp_path):
+    """`{#await import('./X.svelte')}` lives in markup, so the regex pass owns it."""
+    component = _write(
+        tmp_path / "src/Lazy.svelte",
+        """<script>
+  let show = true;
+</script>
+{#await import('./Other.svelte') then Mod}
+  <Mod.default />
+{/await}
+""",
+    )
+    other = _write(tmp_path / "src/Other.svelte", "<p>o</p>\n")
+
+    result = extract_svelte(component)
+    assert _make_id(str(other)) in _import_targets(result, relation="dynamic_import")
+
+
+def test_extract_svelte_scripts_on_one_line_do_not_merge(tmp_path):
+    """Two script bodies on one line must not parse as a single statement."""
+    component = _write(
+        tmp_path / "src/Inline.svelte",
+        "<script>const a = 1</script><script>function b() {}</script>\n",
+    )
+    result = extract_svelte(component)
+    assert result.get("parse_errors") is None
+    assert "b()" in _labels(result)
+
+
+def test_extract_svelte_markup_only_component_does_not_crash(tmp_path):
+    """A `.svelte` file need not have a `<script>` block at all."""
+    component = _write(tmp_path / "src/Plain.svelte", "<h1>no script here</h1>\n")
+    result = extract_svelte(component)
+    assert isinstance(result, dict)
+    assert result.get("parse_errors") is None
+    assert _import_targets(result, relation="imports_from") == set()