Pczajka/critical user journeys by sfc-gh-pczajka · Pull Request #709 · snowflakedb/universal-driver

sfc-gh-pczajka · 2026-03-24T13:16:22Z

Tests scenarios representing critical user journeys uncovered by existing e2e tests. Scenarios are effect of LLM analysis of 1st party clients: Snowflake CLI, Snowfort, Snowpark, SQLAlchemy and existing driver test cases

The idea is to test scenarios used by Python early adopters in other drivers as well

Add 12 critical user journey test scenarios covering table lifecycle, session management, stored procedures, UDFs/UDTFs, cursor reuse, MERGE INTO, COPY INTO, CREATE VIEW/CTAS, object discovery, metadata/catalog operations, transaction control, and error handling. Each journey includes a language-agnostic Gherkin feature definition under tests/definitions/shared/ and a Python E2E implementation under python/tests/e2e/. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Copilot

Pull request overview

Adds a new set of “critical user journey” scenarios (Gherkin) and corresponding Python e2e pytest implementations to cover common Snowflake workflows (DDL/DML lifecycle, object discovery/metadata, session management, COPY/PUT, MERGE, UDF/UDTF, stored procedures).

Changes:

Introduces new shared Gherkin feature specs under tests/definitions/shared/critical_user_journeys/.
Adds Python e2e test modules under python/tests/e2e/critical_user_journeys/ implementing those scenarios.
Establishes a Python package marker for the new test area via __init__.py.

Reviewed changes

Copilot reviewed 19 out of 19 changed files in this pull request and generated 8 comments.

Show a summary per file

File	Description
`tests/definitions/shared/critical_user_journeys/udf_udtf_lifecycle.feature`	Defines UDF/UDTF lifecycle scenarios for Python e2e coverage.
`tests/definitions/shared/critical_user_journeys/table_lifecycle.feature`	Defines table DDL + filtering scenarios.
`tests/definitions/shared/critical_user_journeys/stored_procedure_lifecycle.feature`	Defines stored procedure create/call/drop scenario.
`tests/definitions/shared/critical_user_journeys/session_management.feature`	Defines role switching and restoration scenario.
`tests/definitions/shared/critical_user_journeys/object_discovery.feature`	Defines SHOW/DESCRIBE/DROP discovery scenarios.
`tests/definitions/shared/critical_user_journeys/metadata_catalog.feature`	Defines SHOW TABLES / DESCRIBE TABLE metadata scenarios.
`tests/definitions/shared/critical_user_journeys/merge_into.feature`	Defines MERGE INTO upsert scenario.
`tests/definitions/shared/critical_user_journeys/create_view_ctas.feature`	Defines view creation and CTAS scenarios.
`tests/definitions/shared/critical_user_journeys/copy_into.feature`	Defines PUT + COPY INTO bulk load scenario.
`python/tests/e2e/critical_user_journeys/test_udf_udtf_lifecycle.py`	Implements UDF/UDTF lifecycle e2e tests.
`python/tests/e2e/critical_user_journeys/test_table_lifecycle.py`	Implements table create/describe and WHERE filtering e2e tests.
`python/tests/e2e/critical_user_journeys/test_stored_procedure_lifecycle.py`	Implements stored procedure lifecycle e2e test.
`python/tests/e2e/critical_user_journeys/test_session_management.py`	Implements role switch/restore e2e test.
`python/tests/e2e/critical_user_journeys/test_object_discovery.py`	Implements object discovery and drop verification e2e tests.
`python/tests/e2e/critical_user_journeys/test_metadata_catalog.py`	Implements catalog inspection e2e tests.
`python/tests/e2e/critical_user_journeys/test_merge_into.py`	Implements MERGE INTO upsert e2e test.
`python/tests/e2e/critical_user_journeys/test_create_view_ctas.py`	Implements CREATE VIEW and CTAS e2e tests.
`python/tests/e2e/critical_user_journeys/test_copy_into.py`	Implements PUT + COPY INTO bulk load e2e test.
`python/tests/e2e/critical_user_journeys/__init__.py`	Marks the new critical journey e2e tests package.

Copilot · 2026-03-24T13:22:25Z

+        cursor.execute(
+            f"CREATE OR REPLACE TABLE {table_name} ("
+            f"id INT NOT NULL, "
+            f"name VARCHAR(100), "
+            f"active BOOLEAN, "
+            f"score FLOAT, "
+            f"amount NUMBER(10,2))"
+        )


The scenario/step text calls this a temporary table, but the test creates a non-temporary table. In e2e, tmp_schema is shared across the session and the repo guidance recommends CREATE OR REPLACE TEMPORARY TABLE to avoid name conflicts and cleanup overhead (see python/tests/e2e/conftest.py).

Copilot · 2026-03-24T13:22:26Z

+        cursor.execute(
+            f"CREATE OR REPLACE TABLE {table_name} ("
+            f"id INT NOT NULL, "
+            f"name VARCHAR(100), "
+            f"active BOOLEAN, "
+            f"score FLOAT, "
+            f"amount NUMBER(10,2))"
+        )


Same issue as above: the step says a temporary table exists, but this creates a non-temporary table in the session-scoped tmp_schema. Prefer CREATE OR REPLACE TEMPORARY TABLE here to match the feature and to avoid persistent objects across the e2e session (per python/tests/e2e/conftest.py).

Copilot · 2026-03-24T13:22:26Z

+        cursor.execute(f"CREATE STAGE {stage_name}")
+
+        # And A temporary table "copy_test" with columns (id INT, name VARCHAR, val FLOAT) exists
+        cursor.execute(f"CREATE TABLE {table_name} (id INT, name VARCHAR, val FLOAT)")


This creates a non-temporary stage/table with fixed names and without OR REPLACE / IF NOT EXISTS, which can make reruns flaky within the session-scoped tmp_schema. Prefer a temporary stage/table (or unique names) and use CREATE OR REPLACE / IF NOT EXISTS; there are existing helpers in python/tests/e2e/put_get/put_get_helper.py (e.g., create_temporary_stage).

Suggested change

cursor.execute(f"CREATE STAGE {stage_name}")

# And A temporary table "copy_test" with columns (id INT, name VARCHAR, val FLOAT) exists

cursor.execute(f"CREATE TABLE {table_name} (id INT, name VARCHAR, val FLOAT)")

cursor.execute(f"CREATE OR REPLACE TEMPORARY STAGE {stage_name}")

# And A temporary table "copy_test" with columns (id INT, name VARCHAR, val FLOAT) exists

cursor.execute(

f"CREATE OR REPLACE TEMPORARY TABLE {table_name} (id INT, name VARCHAR, val FLOAT)"

)

Copilot · 2026-03-24T13:22:26Z

+        csv_path.write_text("1,Alice,3.14\n2,Bob,2.72\n3,Charlie,1.41\n")
+
+        # When The CSV file is PUT to the stage
+        cursor.execute(f"PUT 'file://{csv_path}' @{stage_name} AUTO_COMPRESS=FALSE OVERWRITE=TRUE")


Building the PUT 'file://...' URI from a raw Path may break on Windows due to backslashes. Use the existing as_file_uri(Path) helper (or csv_path.as_posix()) when formatting the PUT command to keep the test cross-platform.

Suggested change

cursor.execute(f"PUT 'file://{csv_path}' @{stage_name} AUTO_COMPRESS=FALSE OVERWRITE=TRUE")

cursor.execute(

f"PUT 'file://{csv_path.as_posix()}' @{stage_name} AUTO_COMPRESS=FALSE OVERWRITE=TRUE"

)

Copilot · 2026-03-24T13:22:26Z

+        cursor.execute(f"CREATE TABLE {target_table} (id INT, name VARCHAR, amount INT)")
+        cursor.execute(f"INSERT INTO {target_table} VALUES (1, 'original_1', 100), (2, 'original_2', 200)")
+
+        # And A source table with rows (2, 'updated_2', 250) and (3, 'new_3', 300) exists
+        cursor.execute(f"CREATE TABLE {source_table} (id INT, name VARCHAR, amount INT)")
+        cursor.execute(f"INSERT INTO {source_table} VALUES (2, 'updated_2', 250), (3, 'new_3', 300)")


These tables are created without OR REPLACE / IF NOT EXISTS and are non-temporary in the session-scoped tmp_schema, which can cause flakiness on reruns. Prefer CREATE OR REPLACE TEMPORARY TABLE (or at least CREATE OR REPLACE TABLE) for both target/source to match the e2e fixture guidance (python/tests/e2e/conftest.py).

Suggested change

cursor.execute(f"CREATE TABLE {target_table} (id INT, name VARCHAR, amount INT)")

cursor.execute(f"INSERT INTO {target_table} VALUES (1, 'original_1', 100), (2, 'original_2', 200)")

# And A source table with rows (2, 'updated_2', 250) and (3, 'new_3', 300) exists

cursor.execute(f"CREATE TABLE {source_table} (id INT, name VARCHAR, amount INT)")

cursor.execute(f"INSERT INTO {source_table} VALUES (2, 'updated_2', 250), (3, 'new_3', 300)")

cursor.execute(

f"CREATE OR REPLACE TEMPORARY TABLE {target_table} (id INT, name VARCHAR, amount INT)"

)

cursor.execute(

f"INSERT INTO {target_table} VALUES (1, 'original_1', 100), (2, 'original_2', 200)"

)

# And A source table with rows (2, 'updated_2', 250) and (3, 'new_3', 300) exists

cursor.execute(

f"CREATE OR REPLACE TEMPORARY TABLE {source_table} (id INT, name VARCHAR, amount INT)"

)

cursor.execute(

f"INSERT INTO {source_table} VALUES (2, 'updated_2', 250), (3, 'new_3', 300)"

)

Copilot · 2026-03-24T13:22:27Z

+        cursor.execute(f"INSERT INTO {source_table} VALUES (1, 'first', 1.5), (2, 'second', 2.5), (3, 'third', 3.5)")
+
+        # When CREATE TABLE AS SELECT is executed filtering val > 2.0
+        cursor.execute(f"CREATE TABLE {new_table} AS SELECT * FROM {source_table} WHERE val > 2.0")


CREATE TABLE ... AS SELECT will fail on reruns if the table already exists in the session-scoped tmp_schema. Prefer CREATE OR REPLACE TABLE (or explicitly DROP TABLE IF EXISTS first) to keep the e2e test idempotent.

Suggested change

cursor.execute(f"CREATE TABLE {new_table} AS SELECT * FROM {source_table} WHERE val > 2.0")

cursor.execute(f"CREATE OR REPLACE TABLE {new_table} AS SELECT * FROM {source_table} WHERE val > 2.0")

Copilot · 2026-03-24T13:22:27Z

+    And A table with columns (id INT NOT NULL, name VARCHAR(100), val NUMBER(10,2)) exists
+    When SHOW TABLES LIKE 'disc_show_describe' is executed
+    Then 1 row should be returned
+    When DESCRIBE TABLE e2e_discovery_test is executed


The scenario creates a table named disc_show_describe, but the step text says DESCRIBE TABLE e2e_discovery_test which implies a different object. Update the step wording (and matching test comment) to reference the table that is actually created so the feature remains an accurate executable spec.

Suggested change

When DESCRIBE TABLE e2e_discovery_test is executed

When DESCRIBE TABLE disc_show_describe is executed

Copilot · 2026-03-24T13:22:27Z

+    Then SHOW TABLES should return 0 rows
+    And SHOW VIEWS should return 0 rows


SHOW TABLES / SHOW VIEWS returning 0 rows is not deterministic in a shared schema (other tests may create objects). Consider scoping these steps to a specific object name (e.g., SHOW TABLES LIKE ... IN SCHEMA ...) to make the scenario reliably true and aligned with the Python implementation.

Suggested change

Then SHOW TABLES should return 0 rows

And SHOW VIEWS should return 0 rows

Then SHOW TABLES LIKE 'disc_show_describe' should return 0 rows

And SHOW VIEWS LIKE 'disc_view_show' should return 0 rows

sfc-gh-turbaszek

Imo we need to properly define the aim of those tests. Currently majority of them is testing Snowflake behavior not driver. In my opinion we should focus on scenarios where driver does anything with the result. For example the MERGE statements where rowcount / rows affected are crucial for processing. We should look for similar scenarios where numDmlDuplicates is used.

sfc-gh-turbaszek · 2026-03-25T10:55:00Z

+        csv_path = tmp_path / "data.csv"
+        csv_path.write_text("1,Alice,3.14\n2,Bob,2.72\n3,Charlie,1.41\n")


Should we create test file? I think this is a pattern we use for other tests?

sfc-gh-turbaszek · 2026-03-25T10:55:24Z

+        # Then LS @copy_test_stage should show 1 file
+        cursor.execute(f"LS @{stage_name}")
+        ls_results = cursor.fetchall()
+        assert len(ls_results) == 1


Let's assert this is the expected file

sfc-gh-turbaszek · 2026-03-25T10:56:26Z

+        assert rows[0][0] == 1
+        assert rows[0][1] == "Alice"
+        assert rows[0][2] == pytest.approx(3.14)
+        assert rows[1][0] == 2
+        assert rows[1][1] == "Bob"
+        assert rows[1][2] == pytest.approx(2.72)
+        assert rows[2][0] == 3
+        assert rows[2][1] == "Charlie"
+        assert rows[2][2] == pytest.approx(1.41)


I would opt for assert rows == [...] to see what changes in whole output.

sfc-gh-turbaszek · 2026-03-25T10:56:56Z

+    def test_should_create_view_and_query_filtered_data(self, execute_query, cursor, tmp_schema):
+        """Test CREATE VIEW and query filtered data."""
+        # Given Snowflake client is logged in
+        assert_connection_is_open(execute_query)


Is it necccessary? I think @sfc-gh-fpawlowski was getting rid of it

sfc-gh-turbaszek · 2026-03-25T10:57:37Z

+        cursor.execute(f"INSERT INTO {source_table} VALUES (1, 'first', 1.0), (2, 'second', 2.0), (3, 'third', 3.0)")
+
+        # When A view is created that filters rows where id > 1
+        cursor.execute(f"CREATE OR REPLACE VIEW {view_name} AS SELECT * FROM {source_table} WHERE id > 1")


What is special about view? What CUJ we check here?

sfc-gh-turbaszek · 2026-03-25T11:02:01Z

+        )
+
+        # When SHOW TABLES LIKE 'meta_show' is executed
+        cursor.execute(f"SHOW TABLES LIKE 'meta_show' IN SCHEMA {tmp_schema}")


Seems to be copy of https://github.com/snowflakedb/universal-driver/blob/main/python/tests/e2e/query/test_cuj.py let's unify the CUJ testing

sfc-gh-turbaszek · 2026-03-25T11:02:35Z

+        assert len(results) == 1, f"Expected 1 table, got {len(results)}"
+        # The table name appears in the result (typically in column index 1)
+        row = results[0]
+        assert "META_SHOW" in str(row).upper(), f"Table name not found in result: {row}"


Should we assert proper schema of the result?

sfc-gh-turbaszek · 2026-03-25T11:03:11Z

+class TestObjectDiscovery:
+    """Tests for object discovery operations."""
+
+    def test_should_discover_table_via_show_and_describe(self, execute_query, cursor, tmp_schema):


how this differs from TestMetadataCatalog?

sfc-gh-turbaszek · 2026-03-25T11:03:25Z

+        # Then 1 row should be returned
+        assert len(show_results) == 1, f"Expected 1 view, got {len(show_results)}"
+
+    def test_should_verify_objects_are_gone_after_drop(self, execute_query, cursor, tmp_schema):


This is testing Snowflake not driver.

sfc-gh-turbaszek · 2026-03-25T11:04:19Z

+        # Then SELECT CURRENT_ROLE() should return "PUBLIC"
+        cursor.execute("SELECT CURRENT_ROLE()")
+        result = cursor.fetchone()
+        assert result[0] == "PUBLIC"


How do we know the role has really changed?

sfc-gh-pczajka · 2026-04-15T10:25:04Z

Closing as outdated

sfc-gh-pczajka and others added 2 commits March 24, 2026 10:23

test: remove duplicate CUJ tests, fix parallel safety and table naming

1175e0a

Copilot AI review requested due to automatic review settings March 24, 2026 13:16

sfc-gh-pczajka requested a review from a team as a code owner March 24, 2026 13:16

Copilot started reviewing on behalf of sfc-gh-pczajka March 24, 2026 13:16 View session

Update docstring for critical user journey tests

2f7cb75

Copilot AI reviewed Mar 24, 2026

View reviewed changes

sfc-gh-turbaszek requested changes Mar 25, 2026

View reviewed changes

sfc-gh-pczajka closed this Apr 15, 2026

	cursor.execute(f"CREATE TABLE {new_table} AS SELECT * FROM {source_table} WHERE val > 2.0")
	cursor.execute(f"CREATE OR REPLACE TABLE {new_table} AS SELECT * FROM {source_table} WHERE val > 2.0")

	When DESCRIBE TABLE e2e_discovery_test is executed
	When DESCRIBE TABLE disc_show_describe is executed

		Then SHOW TABLES should return 0 rows
		And SHOW VIEWS should return 0 rows

		csv_path = tmp_path / "data.csv"
		csv_path.write_text("1,Alice,3.14\n2,Bob,2.72\n3,Charlie,1.41\n")

Conversation

sfc-gh-pczajka commented Mar 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 24, 2026

Choose a reason for hiding this comment

Uh oh!

sfc-gh-turbaszek left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

sfc-gh-pczajka commented Apr 15, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

sfc-gh-pczajka commented Mar 24, 2026 •

edited

Loading