Recover CREATE TABLE statements dropped into ERROR nodes by unsupported column defaults (fixes #4613)

(cherry picked from commit d496e9113e)

graphify/extractors/sql.py

@@ -59,6 +59,22 @@ _TRIGGER_RECOVERY_RX = re.compile(
     re.IGNORECASE | re.DOTALL,
 )
 
+# CREATE TABLE recovery, sibling to the routine/trigger patterns above. A
+# column default the grammar cannot model — MySQL/MariaDB
+# `DEFAULT current_timestamp()` / `ON UPDATE current_timestamp()`, whose
+# parentheses tree-sitter-sql rejects (upstream DerekStride/tree-sitter-sql)
+# — puts the whole statement in an ERROR node, so walk() never dispatches on
+# it and the table vanished, leaving only the sourceless stub that other
+# tables' REFERENCES minted for it (#4613). The name atom is _TRIGGER_IDENT
+# (the same delimited-identifier shape); the label is the bare name, with no
+# `()` suffix — a table is not callable either.
+_TABLE_RECOVERY_RX = re.compile(
+    r"\bCREATE\s+(?:OR\s+(?:REPLACE|ALTER)\s+)?TABLE\s+"
+    r"(?:IF\s+NOT\s+EXISTS\s+)?"
+    r"(" + _TRIGGER_IDENT + r")",
+    re.IGNORECASE,
+)
+
 # _mask_sql_comments is a linear character scanner, not a regex: the four
 # span kinds interact in ways a single pattern cannot express safely —
 # comment-opener parity inside strings, nested block comments, and
@@ -813,6 +829,8 @@ def extract_sql(path: Path, content: str | bytes | None = None) -> dict:
         for child in node.children:
             _walk_from_refs(child, caller_nid, line, cte_names)
 
+    src_text = source.decode("utf-8", errors="replace")
+
     # Pre-pass: register every table/view DEFINED in this file before walking,
     # so forward references (a FK to a table created later in the same file)
     # still resolve to the real sourced node instead of falling back to a stub.
@@ -826,6 +844,21 @@ def extract_sql(path: Path, content: str | bytes | None = None) -> dict:
 
     _collect_defined_names(root)
 
+    # A CREATE TABLE the grammar failed on (#4613) never reaches the pre-pass
+    # — its statement sits inside an ERROR node — so references to it would
+    # mint sourceless stubs and the REFERENCES fallback below would skip it.
+    # When the parse failed, pre-register the recovered table names from the
+    # masked source so resolution sees the real id; the node itself is
+    # emitted in the has_error fallback, alongside the recovered routines
+    # and triggers.
+    if root.has_error:
+        _masked, _ident_spans = _scan_sql(src_text)
+        for m in _TABLE_RECOVERY_RX.finditer(_masked):
+            if any(s <= m.start() < e for s, e in _ident_spans):
+                continue
+            table_nids.setdefault(
+                _norm_ident(m.group(1)), _make_id(stem, m.group(1)))
+
     # Secondary bare-name aliases: a reference written without a schema
     # (`REFERENCES users`) should resolve to a schema-qualified definition
     # (`public.users`) when that is unambiguous. Never shadow an explicit
@@ -861,7 +894,6 @@ def extract_sql(path: Path, content: str | bytes | None = None) -> dict:
     # (e.g. Firebird COMPUTED BY columns push constraints out of the tree entirely).
     # Snapshot after tree walk so we don't re-emit edges already captured above.
     emitted = {(e["source"], e["target"]) for e in edges if e["relation"] == "references"}
-    src_text = source.decode("utf-8", errors="replace")
     for m in re.finditer(r"CREATE\s+TABLE\s+([\w$]+)\s*\(", src_text, re.IGNORECASE):
         tbl_name = m.group(1)
         tbl_nid = table_nids.get(_norm_ident(tbl_name))
@@ -946,4 +978,14 @@ def extract_sql(path: Path, content: str | bytes | None = None) -> dict:
             tbl_nid = table_nids.get(_norm_ident(tbl_name)) or _ref_stub(tbl_name)
             _add_edge(trig_nid, tbl_nid, "triggers", trig_line)
 
+        # Tables the grammar dropped into an ERROR node (#4613), pre-registered
+        # above so references already resolved to this id. _add_node dedupes,
+        # so a table that parsed cleanly keeps its walk-emitted node.
+        for m in _TABLE_RECOVERY_RX.finditer(masked_src):
+            if any(s <= m.start() < e for s, e in ident_spans):
+                continue
+            tbl_name = m.group(1)
+            tbl_line = src_text[: m.start()].count("\n") + 1
+            _add_node(_make_id(stem, tbl_name), tbl_name, tbl_line)
+
     return {"nodes": nodes, "edges": edges}

tests/test_multilang.py

@@ -1046,6 +1046,61 @@ def test_sql_procedural_body_trigger_is_recovered(tmp_path):
     assert all(e["source"] in node_ids and e["target"] in node_ids for e in r["edges"])
 
 
+def test_sql_mysql_current_timestamp_parens_table_is_recovered(tmp_path):
+    """MySQL/MariaDB `DEFAULT current_timestamp()` / `ON UPDATE
+    current_timestamp()`: tree-sitter-sql rejects the parentheses, so the
+    whole CREATE TABLE lands in an ERROR node and the table was dropped —
+    the only trace left was the sourceless stub minted by other tables'
+    REFERENCES to it (#4613).
+    """
+    pytest.importorskip("tree_sitter_sql")
+    p = tmp_path / "repro.sql"
+    p.write_text(
+        "CREATE TABLE foo (\n"
+        "  created_at datetime NOT NULL DEFAULT current_timestamp(),\n"
+        "  updated_at datetime DEFAULT NULL ON UPDATE current_timestamp(),\n"
+        "  KEY idx_x (created_by)\n"
+        ");\n"
+        "\n"
+        "CREATE TABLE after_table (\n"
+        "  id int NOT NULL,\n"
+        "  foo_id int REFERENCES foo (id)\n"
+        ");\n",
+        encoding="utf-8",
+    )
+    r = extract_sql(p)
+    by_label = {n["label"]: n for n in r["nodes"]}
+    assert sorted(by_label) == ["after_table", "foo", "repro.sql"], sorted(by_label)
+    # The recovered table is a real sourced node, not the sourceless stub a
+    # REFERENCES would otherwise mint for it.
+    assert by_label["foo"]["source_file"] == str(p)
+    # ... so the FK from after_table resolves onto it.
+    edges = {(e["source"], e["relation"], e["target"]) for e in r["edges"]}
+    assert (by_label["after_table"]["id"], "references", by_label["foo"]["id"]) in edges
+    node_ids = {n["id"] for n in r["nodes"]}
+    assert all(e["source"] in node_ids and e["target"] in node_ids for e in r["edges"])
+
+
+def test_sql_commented_out_create_table_is_not_recovered(tmp_path):
+    """The table recovery runs over the comment-masked source: a commented-out
+    CREATE TABLE in an error-bearing file must not fabricate a node.
+    """
+    pytest.importorskip("tree_sitter_sql")
+    p = tmp_path / "schema.sql"
+    p.write_text(
+        "-- CREATE TABLE ghost (\n"
+        "--   id int\n"
+        "-- );\n"
+        "CREATE TABLE foo (\n"
+        "  created_at datetime NOT NULL DEFAULT current_timestamp()\n"
+        ");\n",
+        encoding="utf-8",
+    )
+    r = extract_sql(p)
+    labels = sorted(n["label"] for n in r["nodes"])
+    assert labels == ["foo", "schema.sql"], labels
+
+
 def test_sql_tsql_bracketed_procedure_is_recovered(tmp_path):
     """T-SQL CREATE PROCEDURE [Schema].[Name] ... AS BEGIN...END.