immediately query for information_schema tables

Author: tswast

bigframes/core/array_value.py

@@ -141,6 +141,11 @@ def from_bq_data_source(
         )
         return cls(node)
 
+    @staticmethod
+    def is_table_type_supported(table_type: Optional[str]):
+        # Some external tables like those in GCS support all the features we want, such as time travel.
+        return table_type in ("TABLE", "MATERIALIZED_VIEW", "EXTERNAL")
+
     @property
     def column_ids(self) -> typing.Sequence[str]:
         """Returns column ids as strings."""

bigframes/session/_io/bigquery/read_gbq_table.py

@@ -42,7 +42,6 @@ def _convert_information_schema_table_id_to_table_reference(
     default_project: Optional[str],
 ) -> bigquery.TableReference:
     """Squeeze an INFORMATION_SCHEMA reference into a TableReference.
-
     This is kind-of a hack. INFORMATION_SCHEMA is a view that isn't available
     via the tables.get REST API.
     """
@@ -152,13 +151,7 @@ def get_table_metadata(
 
         return cached_table
 
-    table_id_casefold = table_id.casefold()
-    if (
-        # Ensure we don't have false positives for some user defined dataset
-        # like MY_INFORMATION_SCHEMA or tables called INFORMATION_SCHEMA.
-        ".INFORMATION_SCHEMA.".casefold() in table_id_casefold
-        or table_id_casefold.startswith("INFORMATION_SCHEMA.".casefold())
-    ):
+    if is_information_schema(table_id):
         table = get_information_schema_metadata(
             bqclient=bqclient, table_id=table_id, default_project=default_project
         )
@@ -179,6 +172,17 @@ def get_table_metadata(
     return cached_table
 
 
+def is_information_schema(table_id: str):
+    table_id_casefold = table_id.casefold()
+    # Include the "."s to ensure we don't have false positives for some user
+    # defined dataset like MY_INFORMATION_SCHEMA or tables called
+    # INFORMATION_SCHEMA.
+    return (
+        ".INFORMATION_SCHEMA.".casefold() in table_id_casefold
+        or table_id_casefold.startswith("INFORMATION_SCHEMA.".casefold())
+    )
+
+
 def is_time_travel_eligible(
     bqclient: bigquery.Client,
     table: google.cloud.bigquery.table.Table,
@@ -245,6 +249,8 @@ def is_time_travel_eligible(
                     msg, category=bfe.TimeTravelDisabledWarning, stacklevel=stacklevel
                 )
             return False
+        elif table.table_type == "VIEW":
+            return False
 
     # table might support time travel, lets do a dry-run query with time travel
     if should_dry_run:

bigframes/session/loader.py

@@ -47,6 +47,8 @@
 import pandas
 import pyarrow as pa
 
+import bigframes._tools
+import bigframes._tools.strings
 from bigframes.core import guid, identifiers, local_data, nodes, ordering, utils
 import bigframes.core as core
 import bigframes.core.blocks as blocks
@@ -701,18 +703,23 @@ def read_gbq_table(
         # Optionally, execute the query
         # -----------------------------
 
-        # max_results introduces non-determinism and limits the cost on
-        # clustered tables, so fallback to a query. We do this here so that
-        # the index is consistent with tables that have primary keys, even
-        # when max_results is set.
-        if max_results is not None:
+        if (
+            # max_results introduces non-determinism and limits the cost on
+            # clustered tables, so fallback to a query. We do this here so that
+            # the index is consistent with tables that have primary keys, even
+            # when max_results is set.
+            max_results is not None
+            # Views such as INFORMATION_SCHEMA also introduce non-determinism.
+            # They can update frequently and don't support time travel.
+            or not core.ArrayValue.is_table_type_supported(table.table_type)
+        ):
             # TODO(b/338111344): If we are running a query anyway, we might as
             # well generate ROW_NUMBER() at the same time.
             all_columns: Iterable[str] = (
                 itertools.chain(index_cols, columns) if columns else ()
             )
             query = bf_io_bigquery.to_query(
-                table_id,
+                f"{table.project}.{table.dataset_id}.{table.table_id}",
                 columns=all_columns,
                 sql_predicate=bf_io_bigquery.compile_filters(filters)
                 if filters

tests/system/small/pandas/test_read_gbq_information_schema.py

@@ -20,7 +20,7 @@
     "view_id",
     [
         # https://cloud.google.com/bigquery/docs/information-schema-intro
-        "region-US.INFORMATION_SCHEMA.JOBS_BY_USER",
+        "region-US.INFORMATION_SCHEMA.SESSIONS_BY_USER",
         "region-US.INFORMATION_SCHEMA.SCHEMATA",
     ],
 )
@@ -32,7 +32,7 @@ def test_read_gbq_jobs_by_user_returns_schema(
     else:
         table_id = view_id
 
-    df = unordered_session.read_gbq(table_id)
+    df = unordered_session.read_gbq(table_id, max_results=10)
     assert df.dtypes is not None