frostlake-dbt

The campaign to run dbt against Frostlake, in three steps:

1. frostlake_connector (done) — a snowflake.connector-shaped facade over the frostlake Python driver.
2. dbt-frostlake (done) — a thin adapter package that subclasses dbt-snowflake, overriding only the credentials shape and connection opening, so all of dbt-snowflake's macros and materializations run as-is. It lives in dbt-frostlake/.
3. Grind dbt's standard acceptance suite (dbt-tests-adapter) green; every failure is either a facade gap or a Frostlake SQL gap worth fixing.

Engine version

Requires a Frostlake engine 0.0.7 or newer. Ask a running server which one it is with SELECT CURRENT_VERSION() — every release answers it, so the check works against any engine.

The adapter versions independently of the engine: it speaks the HTTP protocol, not the jar, so this is a floor rather than a lockstep pin.

The facade

import frostlake_connector as snowflake_connector

conn = snowflake_connector.connect(host="localhost", port=18082,
                                   database="MY_DB", schema="PUBLIC",
                                   warehouse="COMPUTE_WH",
                                   session_parameters={"QUERY_TAG": "ci"})
cur = conn.cursor()
cur.execute("SELECT id, name FROM people WHERE id = %s", (1,))
print(cur.fetchall())          # [(1, 'Ada')]

Covered surface:

• connect(**kwargs) with the connector's keyword vocabulary — host/port select the server; role, warehouse, database, schema become USE statements (in that order) and session_parameters/timezone become ALTER SESSION SET; everything else (account, user, password, authenticator, client_session_keep_alive, …) is accepted and ignored.
• Cursors: execute (returns the cursor), executemany, fetchone/fetchmany/ fetchall, connector-style ResultMetadata descriptions with the connector's numeric type_code scheme (constants.FIELD_ID_TO_NAME), rowcount from DML, sfqid, iteration, context-manager use. DictCursor returns dicts.
• Binding: pyformat by default (%s, %(name)s, %% — skipped inside string literals, quoted identifiers and comments), paramstyle="qmark" for ?.
• execute_string() splits a script on top-level semicolons and returns the per-statement cursors.
• autocommit(mode), commit(), rollback(), is_closed(), session_id.
• errors module: when the real snowflake-connector-python is installed (it always is under dbt-snowflake), its error classes are re-exported and raised, so existing except snowflake.connector.errors.ProgrammingError clauses catch facade errors; standalone, connector-shaped stand-ins with the same msg/errno/ sqlstate attributes are used. Engine compile errors surface as ProgrammingError(errno=1003, sqlstate="42000") (an approximation of the connector's numbering — Frostlake's message text is authoritative).

The transport is the frostlake PEP 249 driver, which installs automatically as a dependency:

pip install frostlake-connector

Running the tests

export JAVA_HOME=/path/to/jdk17
export FROSTLAKE_CLASSPATH="/path/to/frostlake/engine/target/classes:<engine deps>"
python3 test/test_facade.py

The suite boots a real DatabaseHttpServer and covers connect-kwargs context (CURRENT_DATABASE/CURRENT_SCHEMA/CURRENT_WAREHOUSE), binding in both paramstyles, descriptions and type codes, DictCursor, executemany, execute_string, the error shape, and dbt-style transaction sequences. Without FROSTLAKE_CLASSPATH the integration tests skip and only the binding/split unit tests run.