sql-code-graph 1.1.0__py3-none-any.whl → 1.1.3__py3-none-any.whl

{sql_code_graph-1.1.0.dist-info → sql_code_graph-1.1.3.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sql-code-graph
-Version: 1.1.0
+Version: 1.1.3
 Summary: SQL code graph analyzer and lineage tracer
 Project-URL: Homepage, https://github.com/Warhorze/sql-code-graph
 Project-URL: Repository, https://github.com/Warhorze/sql-code-graph

{sql_code_graph-1.1.0.dist-info → sql_code_graph-1.1.3.dist-info}/RECORD

@@ -1,12 +1,12 @@
-sqlcg/__init__.py,sha256=CWoJX8Awg5Tf6p2E5lT66EFE8kd-Aru8aujKizglgdo,115
+sqlcg/__init__.py,sha256=YGDRrWVIrONmQholAKWh6hSKxlPd2dLcM1AdHHdBhEA,115
 sqlcg/__main__.py,sha256=1YoFLcqEgTwYq1J3TbUwpkdG0zeeLIf2fJvwWI-CLFU,109
 sqlcg/cli/__init__.py,sha256=W8fD0LpMq2xm_5WKGNMvJh2WBL1ho5E8hUeAqXQYT1g,28
 sqlcg/cli/main.py,sha256=WmdTjsOlz1ozi2Y3Aq4ezR_FCRl-Lc1YOKw3_d48dlY,1650
 sqlcg/cli/commands/__init__.py,sha256=oSHtr6VD-jNubOjuCQyZj2tBppjMEpQDh-IGQ8of9eA,30
-sqlcg/cli/commands/analyze.py,sha256=qtvM_TeqYzaLClZksM_o5hAdksZ9sqLM9HGDtLDrXwY,12646
-sqlcg/cli/commands/db.py,sha256=Q3VEdNJzhrs26KtskI5j9B3C0vBTZe4VN2sZXZG_6BY,7434
-sqlcg/cli/commands/find.py,sha256=5MbGavA-QS75zwm35dYK-0H1bJ1Zd_gJHgQ_lXnpMDU,3126
-sqlcg/cli/commands/gain.py,sha256=bOvia7CVla_fESrDEdftYze8Mm0xDio3SpCzIyoXg7A,8925
+sqlcg/cli/commands/analyze.py,sha256=hiKj0R1m5i4ZmwrXBlVT14xGy6rs9jmv_ZDCLVZj4Tg,14282
+sqlcg/cli/commands/db.py,sha256=5VpknLqYaimK6YA516w6iQVX6JmHcD52o6MuW5d088c,7462
+sqlcg/cli/commands/find.py,sha256=SsK6q4YRPknrz_lIQ4Gun6HRoAdoVRGClwAYdm_s2OU,3168
+sqlcg/cli/commands/gain.py,sha256=hz36QmuaXJXutI4vyNMDfcNsBeLTXa6EOw2bWe2AhTQ,8939
 sqlcg/cli/commands/git.py,sha256=yMgWOuoTCTBr2P1QgmghRi5ikmUYHuxDUVyBDYerErw,5728
 sqlcg/cli/commands/index.py,sha256=xMnxKDiUt5LH_3lKAotoRctL4VSOvcw7Gq--idLPtm0,11091
 sqlcg/cli/commands/install.py,sha256=KNABvrLbamPyYnmnVdCaM_MNezbDc-pr6IkignCWI8k,9186
@@ -16,21 +16,21 @@ sqlcg/cli/commands/report.py,sha256=JU0qjyMxwOukE7bN3XvvIzOI7zMg_Gsnvk_8F6pKNpA,
 sqlcg/cli/commands/uninstall.py,sha256=IYwQaqnMmmzW0Nlls40wD-L3tVkMgKIMRXUkcXPMUc4,9398
 sqlcg/cli/commands/watch.py,sha256=7N6c-QuvxAEGHzDZ0C3CU2BkHSraZW9YtgoFnz7SaQo,2373
 sqlcg/core/__init__.py,sha256=uNsJCrCMVWVT80sHPtI_f39BYqIf5N0i6LSq8x8HsyI,283
-sqlcg/core/config.py,sha256=8QtFNRnrzLK1Zw93AKX37h6bSASDLv-42FzDQ7zxTtI,13079
+sqlcg/core/config.py,sha256=LuB8HWPsIt1OsjOshTT1bJdXWXN01w76ABl9M-VB9DM,14777
 sqlcg/core/freshness.py,sha256=gRb8pRPw5SdIUxAYkMXIJ00DTdQ6CegRZPAvWnv0rU0,4575
 sqlcg/core/graph_db.py,sha256=Aa85wPFg26H-Ud9SrZyxCHH-99iitAI5S3X9T_62Yyw,7957
 sqlcg/core/jobs.py,sha256=Je-fCdSKRgiSsv1W8SgNAlp36a7t7-pJZ-qKPbka9OE,3298
-sqlcg/core/kuzu_backend.py,sha256=ziHt-AB9sEZY7qB8whseWFicbTfOZaNOxcNVKhjii5Y,16587
+sqlcg/core/kuzu_backend.py,sha256=3kL8bGEQm70fuxYUdt1p7fsY12lCLQ07x01NYg6FOGA,16821
 sqlcg/core/neo4j_backend.py,sha256=AM1TncP9GBGph-rSHwalZPmGUV2kFILzaJP-PSB0UYw,8437
-sqlcg/core/queries.cypher,sha256=91Pb10-ekSi0812wuHJTdXcMY4sT53_5o-oHhfSP_DQ,4967
-sqlcg/core/queries.py,sha256=JLgV4MIgP7KVIQ0xpGj3_-MBhBfY_9XPoCdcI2mO-TM,2148
+sqlcg/core/queries.cypher,sha256=cvPOVe5GUOzJN4bxUvDxNI--xIIP8gm42TR-gUnea4U,4685
+sqlcg/core/queries.py,sha256=gkl4bhkZM8FsvbSA-IaK17sRFcO3hB5YlVCemkCXgWM,2064
 sqlcg/core/schema.cypher,sha256=rK5QMhSrzZhuj73NeNXGX6oM-rPPPvxFjex0fEyUvkQ,2859
 sqlcg/core/schema.py,sha256=JO5rkspYKjL9AEl5mt0VIJKn-IPOH3kJV_fVmAMuFCI,1467
 sqlcg/indexer/__init__.py,sha256=Wh20Unz2OHs1oIyWLrpurPAasF0BET2g4iXtNk7mh2U,56
 sqlcg/indexer/dbt_adapter.py,sha256=EB5x1WU5Z9d-I97ADDj88S_hG1C4z4nbrv8JUCzXfy8,686
 sqlcg/indexer/error_classify.py,sha256=MYjPVprwT-ARPjBCyCzu2F9DSrZfnTVtVIoBgm8s4H8,5329
 sqlcg/indexer/git_delta.py,sha256=P-QM4vnVURT2KLiE6u3cQynRUF-mTH13cbB4I20YHPQ,4468
-sqlcg/indexer/indexer.py,sha256=DYdUr59hRKCjJTRiQUWOC72JUQ9TgBrH0W4UOYNwqx8,60913
+sqlcg/indexer/indexer.py,sha256=KyyowxiSNU3Gm4JE-mj8gVm6D80XERJPd-he59I2sIk,62018
 sqlcg/indexer/pool.py,sha256=BTYx-pBe6zwUG89MHh0X7nzGNVlsHN-GjovYKanVI1s,18553
 sqlcg/indexer/walker.py,sha256=umNaqDbuerr75VYG1TEOv0ATsbI40O3SIw35f7XJcDE,1931
 sqlcg/indexer/watcher.py,sha256=mJQq1LASRLKKwhz0WhCUWPLLqyPR2_-FD_8efYU6gE8,8442
@@ -41,7 +41,7 @@ sqlcg/metrics/__init__.py,sha256=hLJ6wm4St8qqYwKh3o9QG7lcEt1BEYM31ccqO9tGpIg,133
 sqlcg/metrics/store.py,sha256=BaMf7QYTmYMlX_Jzi1GNU8R2sMVkWdn07f-ZSndtcNk,8879
 sqlcg/parsers/__init__.py,sha256=AamA8wBbDZV9_zEtZCI4Hyen5UAVKHmBwjTghTt2PZE,785
 sqlcg/parsers/ansi_parser.py,sha256=tu1MWWaSYmpefKjgk2PPyGStIFjV47Z_1WjyBh5Zi2c,17180
-sqlcg/parsers/base.py,sha256=uL0W22zpbIz_9eq-i-4LSlonxy2J1yChuISMLSYgvRU,49345
+sqlcg/parsers/base.py,sha256=IiOkVsm6jz9-48RqDCXiW-UXAraNxQ4pKXvSA7aolnA,49907
 sqlcg/parsers/bigquery_parser.py,sha256=mOnWTfXB_Dp4JwFE1PVYOB6CDPf5nYE0Dea8kJCl9uQ,2827
 sqlcg/parsers/postgres_parser.py,sha256=lYfUpQY6j4Qm7ndXBtXbgPoGzYqYddWt5YeFnWKdA6I,946
 sqlcg/parsers/registry.py,sha256=LXy1F6rqQI6VdxpRvZg_tNpoEucW3mXZHYBMlMONbX4,1496
@@ -59,7 +59,7 @@ sqlcg/utils/__init__.py,sha256=--iqt5ThTXmT8Wz7da8hs3n0zDfYPl8P-z5OgRJ_77E,154
 sqlcg/utils/hashing.py,sha256=H25-sYfxHKb3_IERFnHyAIYNiXN470Oqo5sJT_D3YOA,438
 sqlcg/utils/ignore.py,sha256=wJjwa0mjnQ_xJExOUxk25y00g065XmmzJapqV3ifD5o,1151
 sqlcg/utils/logging.py,sha256=u0fCmYsLj9o81vawm3xZTHaw68GQYVm7JxG-gP81u8A,840
-sql_code_graph-1.1.0.dist-info/METADATA,sha256=blW1eYNjfy6P61747uUtc22qm5MDETMcVYImXPa762g,13615
-sql_code_graph-1.1.0.dist-info/WHEEL,sha256=QccIxa26bgl1E6uMy58deGWi-0aeIkkangHcxk2kWfw,87
-sql_code_graph-1.1.0.dist-info/entry_points.txt,sha256=Wfe49sVzV9p4eVFGo5RxcV-frr3HOP0yzzst8JBxQLQ,46
-sql_code_graph-1.1.0.dist-info/RECORD,,
+sql_code_graph-1.1.3.dist-info/METADATA,sha256=Z_aRnsDOgZ_ngAHkIr3x2XpEjF-x6UMUQwcIkAMlGjo,13615
+sql_code_graph-1.1.3.dist-info/WHEEL,sha256=QccIxa26bgl1E6uMy58deGWi-0aeIkkangHcxk2kWfw,87
+sql_code_graph-1.1.3.dist-info/entry_points.txt,sha256=Wfe49sVzV9p4eVFGo5RxcV-frr3HOP0yzzst8JBxQLQ,46
+sql_code_graph-1.1.3.dist-info/RECORD,,

sqlcg/__init__.py

@@ -1,5 +1,5 @@
 """SQL Code Graph - SQL lineage and dependency analysis tool."""
 
-__version__ = "1.1.0"
+__version__ = "1.1.3"
 
 __all__ = ["__version__"]

sqlcg/cli/commands/analyze.py

@@ -19,6 +19,29 @@ app = typer.Typer(help="Lineage analysis")
 console = Console()
 
 
+def _kind_filter(source_alias: str, include_intermediate: bool) -> str:
+    """Build the Half-B (#38) kind filter for the lineage traversal query.
+
+    When ``include_intermediate`` is False, the filter uses ``OPTIONAL MATCH`` plus
+    ``t.kind IS NULL OR t.kind IN ['table', 'external']`` so a source whose SqlTable
+    node is ABSENT (a CTE-body source on a graph indexed before the #39 fix, or not yet
+    re-indexed) is KEPT rather than silently dropped.  Reverting this to an inner
+    ``MATCH (t:SqlTable {...}) ... WHERE t.kind IN [...]`` is the #38 regression:
+    node-less physical sources vanish from results.
+
+    ``source_alias`` is ``src`` for upstream and ``dst`` for downstream — it names both
+    the node whose table is looked up and the variable carried through the WITH clause.
+    This is the single production source of the filter string; the #40 recall guard
+    imports it so reverting Half B here turns the guard red.
+    """
+    if include_intermediate:
+        return ""
+    return (
+        f"OPTIONAL MATCH (t:SqlTable {{qualified: {source_alias}.table_qualified}}) "
+        f"WITH c, {source_alias}, t WHERE t.kind IS NULL OR t.kind IN ['table', 'external'] "
+    )
+
+
 @app.command("upstream")
 def upstream(  # noqa: B008
     ref: str = typer.Argument(..., help="Column reference"),  # noqa: B008
@@ -34,15 +57,14 @@ def upstream(  # noqa: B008
         console.print("[red]Error: --depth must be between 1 and 100[/red]")
         raise typer.Exit(1)
 
-    # By default, filter out CTE/derived intermediate nodes; --include-intermediate restores them
-    kind_filter = (
-        ""
-        if include_intermediate
-        else "MATCH (t:SqlTable {qualified: src.table_qualified}) "
-        "WHERE t.kind IN ['table', 'external'] "
-    )
+    # By default, filter out CTE/derived intermediate nodes; --include-intermediate restores them.
+    # Half B (#38): use OPTIONAL MATCH so a missing SqlTable node (e.g. CTE-body source not yet
+    # re-indexed after #39 fix) is KEPT rather than silently dropped.  WHERE t.kind IS NULL OR
+    # t.kind IN [...] means: keep when node absent (NULL) OR when kind is a physical source.
+    # CTE aliases (kind='cte') and derived tables (kind='derived') are filtered out.
+    kind_filter = _kind_filter("src", include_intermediate)
 
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (c:{NodeLabel.COLUMN} {{id: $ref}})"
             f"<-[:{RelType.COLUMN_LINEAGE}*1..{depth}]-(src:{NodeLabel.COLUMN}) "
@@ -95,15 +117,13 @@ def downstream(  # noqa: B008
         console.print("[red]Error: --depth must be between 1 and 100[/red]")
         raise typer.Exit(1)
 
-    # By default, filter out CTE/derived intermediate nodes; --include-intermediate restores them
-    kind_filter = (
-        ""
-        if include_intermediate
-        else "MATCH (t:SqlTable {qualified: dst.table_qualified}) "
-        "WHERE t.kind IN ['table', 'external'] "
-    )
+    # By default, filter out CTE/derived intermediate nodes; --include-intermediate restores them.
+    # Half B (#38): OPTIONAL MATCH keeps sources whose SqlTable node is absent (NULL) or is a
+    # physical kind.  WITH c, dst, t carries the three variables in scope at this interpolation
+    # point; direct and q are bound later in the query.
+    kind_filter = _kind_filter("dst", include_intermediate)
 
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (c:{NodeLabel.COLUMN} {{id: $ref}})"
             f"-[:{RelType.COLUMN_LINEAGE}*1..{depth}]->(dst:{NodeLabel.COLUMN}) "
@@ -171,7 +191,7 @@ def impact(  # noqa: B008
     raw: bool = typer.Option(False, "--raw", help="Disable noise filtering on results"),  # noqa: B008
 ) -> None:
     """Show all queries impacted by a table."""
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (t:{NodeLabel.TABLE} {{qualified: $t}})"
             f"<-[:{RelType.SELECTS_FROM}]-(q:{NodeLabel.QUERY}) "
@@ -207,7 +227,7 @@ def failures(
     with 'sqlcg db reset && sqlcg index <path>' if the graph was built with
     an earlier version.
     """
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         cypher = (
             f"MATCH (f:{NodeLabel.FILE}) WHERE f.parse_failed = true "
             "AND ($cause IS NULL OR f.parse_cause = $cause) "
@@ -224,7 +244,7 @@ def unused(
     raw: bool = typer.Option(False, "--raw", help="Disable noise filtering on results"),  # noqa: B008
 ) -> None:
     """Find tables with no query references."""
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (t:{NodeLabel.TABLE}) WHERE NOT (t)<-[:{RelType.SELECTS_FROM}]-() "
             "RETURN DISTINCT t.qualified AS qualified LIMIT 100",

sqlcg/cli/commands/db.py

@@ -75,7 +75,7 @@ def db_reset(  # noqa: B008
 @app.command("info")
 def db_info() -> None:
     """Show database stats."""
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         version = backend.get_schema_version() or "unknown"
         console.print(f"Schema version: {version}")
 
@@ -167,7 +167,7 @@ def db_info() -> None:
 @app.command("list-repos")
 def list_repos() -> None:
     """List all indexed repositories."""
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         result = backend.run_read("MATCH (r:Repo) RETURN r.path AS path, r.name AS name", {})
 
         if not result:

sqlcg/cli/commands/find.py

@@ -18,7 +18,7 @@ def find_table(  # noqa: B008
 ) -> None:
     """Find a table by name."""
     name = name.lower()  # graph keys are lowercased at index time (C2 normalization)
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (t:{NodeLabel.TABLE}) WHERE t.qualified CONTAINS $name "
             "RETURN t.qualified AS qualified, t.kind AS kind LIMIT 50",
@@ -42,7 +42,7 @@ def find_column(  # noqa: B008
 ) -> None:
     """Find a column by table.column reference."""
     ref = ref.lower()  # graph keys are lowercased at index time (C2 normalization)
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (c:{NodeLabel.COLUMN}) WHERE c.id CONTAINS $ref RETURN c.id AS id LIMIT 50",
             {"ref": ref},
@@ -61,7 +61,7 @@ def find_pattern(  # noqa: B008
     pattern: str = typer.Argument(..., help="SQL pattern to search for"),  # noqa: B008
 ) -> None:
     """Find queries containing a SQL pattern."""
-    with get_backend() as backend:
+    with get_backend(read_only=True) as backend:
         results = backend.run_read(
             f"MATCH (q:{NodeLabel.QUERY}) WHERE q.sql CONTAINS $pattern "
             "RETURN q.id AS id, q.kind AS kind LIMIT 50",

sqlcg/cli/commands/gain.py

@@ -123,7 +123,7 @@ def gain_cmd(
         # Section F: parse quality from graph
         parse_quality: dict[str, int] | None = None
         try:
-            with get_backend() as backend:
+            with get_backend(read_only=True) as backend:
                 mode_rows = backend.run_read(
                     "MATCH (q:SqlQuery) RETURN q.parsing_mode AS mode,"
                     " COUNT(q) AS cnt ORDER BY cnt DESC",

sqlcg/core/config.py

@@ -346,9 +346,25 @@ def get_external_consumers(path: Path) -> list[ExternalConsumerSpec]:
     return []
 
 
-def get_backend() -> "GraphBackend":
+def get_backend(read_only: bool = False) -> "GraphBackend":
     """Get a graph backend instance respecting the SQLCG_BACKEND env var.
 
+    Args:
+        read_only: Open in read-only mode.  When ``True``, the KùzuDB open
+            does not take an exclusive write lock, enabling *multiple concurrent
+            read-only opens* (reader/reader concurrency).  CLI read commands
+            pass ``True`` so they do not hold the exclusive write lock and
+            therefore do not block other concurrent readers or a pending reindex.
+            Note: this does NOT allow reads while a read-write writer already
+            holds the exclusive lock — KùzuDB's exclusive write lock is
+            process-level; a ``read_only=True`` open still fails with
+            "Database is locked" when a writer is active.  Reads during an
+            active writer remain a known limitation (future work: route reads
+            through the live MCP server).
+            Neo4j has no single-writer lock; this flag is a no-op there.
+            All writer call sites (index, reindex, db init/reset, server
+            init_backend) use the default ``False``.
+
     Returns:
         A GraphBackend instance (KuzuBackend by default, or Neo4jBackend)
 
@@ -361,14 +377,26 @@ def get_backend() -> "GraphBackend":
         from sqlcg.core.kuzu_backend import KuzuBackend
 
         kuzu_cfg = KuzuConfig.from_env()
-        return KuzuBackend(
-            str(kuzu_cfg.db_path),
-            buffer_pool_size_mb=kuzu_cfg.buffer_pool_size_mb,
-        )
+        try:
+            return KuzuBackend(
+                str(kuzu_cfg.db_path),
+                buffer_pool_size_mb=kuzu_cfg.buffer_pool_size_mb,
+                read_only=read_only,
+            )
+        except RuntimeError as exc:
+            if read_only and "READ ONLY" in str(exc):
+                # KùzuDB refuses to open a non-existent or empty DB in read-only
+                # mode ("Cannot create an empty database under READ ONLY mode").
+                # Surface the same empty-DB guidance the user sees from `db info`.
+                raise RuntimeError(
+                    "Database not initialised — run 'sqlcg db init' and 'sqlcg index <path>' first."
+                ) from exc
+            raise
     elif backend_type == "neo4j":
         from sqlcg.core.neo4j_backend import Neo4jBackend
 
         neo4j_cfg = Neo4jConfig.from_env()
+        # Neo4j has no single-writer lock; read_only is a no-op here.
         return Neo4jBackend(neo4j_cfg.uri, neo4j_cfg.user, neo4j_cfg.password)
     else:
         raise ValueError(f"Unknown backend type: {backend_type}")

sqlcg/core/kuzu_backend.py

@@ -58,7 +58,10 @@ class KuzuBackend(GraphBackend):
         Args:
             db_path: Path to the KùzuDB database file (or ':memory:' for in-memory)
             buffer_pool_size_mb: Buffer pool size in MB (0 = use KuzuDB default)
-            read_only: Open in read-only mode (allows concurrent indexing)
+            read_only: Open in read-only mode.  Enables concurrent read-only
+                opens (reader/reader concurrency) by not taking the exclusive
+                write lock.  Does NOT allow reads while a read-write writer
+                holds the lock — KùzuDB's exclusive lock is process-level.
 
         Raises:
             RuntimeError: If the database is locked or cannot be opened.

sqlcg/core/queries.cypher

@@ -38,12 +38,6 @@ RETURN dst.id AS id, dst.col_name AS col_name, dst.table_qualified AS table_qual
 MATCH (dst:SqlColumn {id: $id})<-[:COLUMN_LINEAGE]-(src:SqlColumn)
 RETURN src.id AS id, src.col_name AS col_name, src.table_qualified AS table_qualified
 
--- GET_UPSTREAM_DEPENDENCIES_FILTERED
-MATCH (dst:SqlColumn {id: $id})<-[:COLUMN_LINEAGE]-(src:SqlColumn)
-MATCH (t:SqlTable {qualified: src.table_qualified})
-WHERE t.kind IN ['table', 'external']
-RETURN src.id AS id, src.col_name AS col_name, src.table_qualified AS table_qualified
-
 -- SEARCH_SQL_PATTERN
 MATCH (q:SqlQuery)-[:QUERY_DEFINED_IN]->(f:File)
 WHERE contains(q.sql, $query)

sqlcg/core/queries.py

@@ -28,7 +28,6 @@ TRACE_COLUMN_LINEAGE_QUERY = _Q["TRACE_COLUMN_LINEAGE"]
 FIND_TABLE_USAGES_QUERY = _Q["FIND_TABLE_USAGES"]
 GET_DOWNSTREAM_DEPENDENCIES_QUERY = _Q["GET_DOWNSTREAM_DEPENDENCIES"]
 GET_UPSTREAM_DEPENDENCIES_QUERY = _Q["GET_UPSTREAM_DEPENDENCIES"]
-GET_UPSTREAM_DEPENDENCIES_FILTERED_QUERY = _Q["GET_UPSTREAM_DEPENDENCIES_FILTERED"]
 SEARCH_SQL_PATTERN_QUERY = _Q["SEARCH_SQL_PATTERN"]
 LIST_DIALECTS_AND_REPOS_QUERY = _Q["LIST_DIALECTS_AND_REPOS"]
 EXPAND_STAR_SOURCES_QUERY = _Q["EXPAND_STAR_SOURCES"]

sqlcg/indexer/indexer.py

@@ -1126,6 +1126,24 @@ class Indexer:
                         "table_name": edge.src.table.name,
                     }
                 )
+                # Half A (#39): emit a SqlTable node for the source table.
+                # CTE-body-only sources are not in stmt.sources (which only covers
+                # tables reachable via the parser's top-level FROM list), so they were
+                # previously missing from the graph.  edge.src.table is a frozen
+                # TableRef with schema-aliasing already applied at parse time — the
+                # qualified value is guaranteed to match edge.src.table_qualified.
+                # key set is identical to other table_rows entries → upsert_nodes_bulk
+                # homogeneity preserved; MERGE on primary key deduplicates re-emits.
+                rows.table_rows.append(
+                    {
+                        "qualified": edge.src.table.full_id,
+                        "name": edge.src.table.name,
+                        "catalog": edge.src.table.catalog or "",
+                        "db": edge.src.table.db or "",
+                        "kind": edge.src.table.role,
+                        "defined_in_file": "",
+                    }
+                )
                 rows.column_rows.append(
                     {
                         "id": dst_id,

sqlcg/parsers/base.py

@@ -967,10 +967,16 @@ class SqlParser(ABC):
                             if not isinstance(cte_body, (exp.Select, exp.Union)):
                                 continue
 
-                            # For Union bodies, use the left branch's projections.
+                            # For Union bodies, use the deepest left-branch Select's projections.
                             # Union.expressions is always empty; projections are on Union.this.
+                            # For N=2: cte_body.this is a Select — the while loop is a no-op.
+                            # For N≥3: cte_body.this is a nested Union (A UNION ALL B UNION ALL C
+                            # parses as Union(Union(A,B),C)), so we walk down to the deepest
+                            # left-branch Select (whose star qualify() already expanded in place).
                             if isinstance(cte_body, exp.Union):
                                 projection_source = cte_body.this
+                                while isinstance(projection_source, exp.Union):
+                                    projection_source = projection_source.this
                             else:
                                 projection_source = cte_body