feat(plugin-chart-echarts): add Butterfly Chart (#43346)

Co-authored-by: Superset Dev <dev@superset.apache.org>

.github/workflows/testcontainers.yml

@@ -1,184 +0,0 @@
-# db_engine_specs tests against real databases (testcontainers)
-name: Testcontainers
-
-# Spins up real Docker containers (see tests/testcontainers/ for the current
-# dialect list) via testcontainers-python, which catches real dialect/driver
-# regressions -- the kind mocked db_engine_specs unit tests structurally
-# cannot, e.g. apache/superset#42899 (Trino emitting OFFSET before LIMIT).
-# Runs on a nightly cron (catches drift from a driver's own releases, not
-# just from Superset's changes) and on pull_request, scoped via `paths` to
-# only PRs that actually touch this test suite or the workflow itself, so
-# unrelated PRs across the repo are never affected.
-#
-# A matrix entry can set `nightly_only: true` to run only on the cron (or a
-# manual workflow_dispatch), never on pull_request -- for a dialect whose
-# image is too heavy (a multi-service cluster, a many-GB image, a slow
-# licensed installer) to justify adding its wall-clock/resource cost to
-# every PR that merely touches this suite. Omit the field entirely for a
-# normal dialect; it isn't nightly-only by default.
-permissions:
-  contents: read
-
-on:
-  schedule:
-    - cron: "0 5 * * *"
-  workflow_dispatch: {}
-  pull_request:
-    paths:
-      - ".github/workflows/testcontainers.yml"
-      - "tests/testcontainers/**"
-      - "superset/db_engine_specs/**"
-      - "pyproject.toml"
-      - "requirements/development.in"
-      - "requirements/development.txt"
-
-concurrency:
-  # Scoped by ref, not just workflow name -- otherwise every PR run and the
-  # nightly cron share one group, and starting the workflow on another PR
-  # (or the nightly firing mid-PR-run) cancels an unrelated in-progress run.
-  group: ${{ github.workflow }}-${{ github.ref }}
-  cancel-in-progress: true
-
-jobs:
-  compute-matrix:
-    # Filters out `nightly_only` dialects for a pull_request run *before* any
-    # testcontainers job is created for them, so a heavy dialect costs a PR
-    # nothing at all -- not even its checkout/setup/driver-install steps --
-    # rather than being merely skipped at the test step. A job-level `if:`
-    # can't reference `matrix` (only github/inputs/needs/vars are available
-    # there), so the filtering has to happen here, before the matrix exists,
-    # instead of on the testcontainers job itself.
-    runs-on: ubuntu-26.04
-    outputs:
-      include: ${{ steps.filter.outputs.include }}
-    steps:
-      - name: Filter nightly-only dialects
-        id: filter
-        env:
-          # One entry per dialect rather than one job for the whole suite: a
-          # single slow container would otherwise inflate the wall-clock
-          # time for every dialect, not just its own. Running in parallel
-          # means the suite's total time is bounded by the slowest dialect,
-          # not the sum of all of them. Db2's first-boot init is documented
-          # upstream as notably slow (a real instance bring-up, not just a
-          # process start) and untested locally here (no arm64 image), so
-          # it gets a wider timeout margin than the rest until real CI data
-          # says otherwise.
-          FULL_MATRIX: |
-            [
-              {"dialect": "cockroachdb", "timeout": 10},
-              {"dialect": "crate", "timeout": 10},
-              {"dialect": "trino", "timeout": 10},
-              {"dialect": "mssql", "timeout": 10},
-              {"dialect": "elasticsearch", "timeout": 10},
-              {"dialect": "oracle", "timeout": 15},
-              {"dialect": "db2", "timeout": 25},
-              {"dialect": "mariadb", "timeout": 10},
-              {"dialect": "timescaledb", "timeout": 10},
-              {"dialect": "yugabytedb", "timeout": 10},
-              {"dialect": "monetdb", "timeout": 10},
-              {"dialect": "mongodb", "timeout": 10},
-              {"dialect": "postgres", "timeout": 10},
-              {"dialect": "mysql", "timeout": 10},
-              {"dialect": "clickhouse", "timeout": 10},
-              {"dialect": "starrocks", "timeout": 15},
-              {"dialect": "databend", "timeout": 10},
-              {"dialect": "risingwave", "timeout": 10},
-              {"dialect": "firebird", "timeout": 10},
-              {"dialect": "ydb", "timeout": 10},
-              {"dialect": "oceanbase", "timeout": 20, "nightly_only": true}
-            ]
-        run: |
-          if [[ "${{ github.event_name }}" == "schedule" || "${{ github.event_name }}" == "workflow_dispatch" ]]; then
-            include="$(jq -c '.' <<<"$FULL_MATRIX")"
-          else
-            include="$(jq -c '[.[] | select(.nightly_only != true)]' <<<"$FULL_MATRIX")"
-          fi
-          echo "include=${include}" >> "$GITHUB_OUTPUT"
-
-  testcontainers:
-    needs: [compute-matrix]
-    runs-on: ubuntu-26.04
-    strategy:
-      fail-fast: false
-      matrix:
-        include: ${{ fromJson(needs.compute-matrix.outputs.include) }}
-    timeout-minutes: ${{ matrix.timeout }}
-    env:
-      PYTHONPATH: ${{ github.workspace }}
-      SUPERSET_TESTENV: true
-      SUPERSET_SECRET_KEY: not-a-secret
-      # This job's matrix installs exactly one dialect's testcontainers
-      # driver for exactly this job, so treat that driver as required: a
-      # broken/missing import should fail the job, not silently skip to a
-      # misleadingly green, zero-tests-run result. See _driver.py.
-      SUPERSET_TESTCONTAINERS_STRICT: true
-    steps:
-      - name: Checkout
-        uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
-        with:
-          persist-credentials: false
-      - name: Setup Python
-        uses: ./.github/actions/setup-backend/
-        with:
-          python-version: current
-      - name: Install db2 driver (ibm-db-sa)
-        # ibm-db (the db2 DBAPI) ships no Linux arm64 wheel, so it's kept out
-        # of the baseline dev install (requirements/development.in) to avoid
-        # breaking the multi-platform dev Docker image build. Install it here
-        # instead, only for this leg of the matrix.
-        if: matrix.dialect == 'db2'
-        run: uv pip install --system -e .[db2]
-      - name: Install oceanbase driver (oceanbase_py)
-        # oceanbase_py pins sqlalchemy-utils>=0.38.3,<0.39, which conflicts
-        # outright with Superset's own sqlalchemy-utils==0.42.1 pin -- kept
-        # out of the baseline dev install for the same reason as db2 above.
-        # Installed as its own standalone package (not via `-e .[oceanbase]`)
-        # so --no-deps only skips *oceanbase_py's* dependencies -- applied
-        # to `-e .[oceanbase]` instead, --no-deps blocks pip from installing
-        # anything the extras marker pulls in, including oceanbase_py
-        # itself, which "succeeds" without actually installing it
-        # (confirmed on real CI: the install step reported success, but the
-        # module was still missing). This job only needs oceanbase_py's
-        # dialect module importable, not its sqlalchemy-utils dependency
-        # satisfied, since nothing here calls into it.
-        if: >-
-          matrix.dialect == 'oceanbase' &&
-          (matrix.nightly_only != true ||
-           github.event_name == 'schedule' ||
-           github.event_name == 'workflow_dispatch')
-        run: uv pip install --system --no-deps "oceanbase_py>=0.0.1.2"
-      - name: Install Firebird client library (libfbclient2)
-        # sqlalchemy-firebird's driver (firebird-driver) is a pure-Python
-        # ctypes wrapper (its wheel is py3-none-any) that dynamically loads
-        # the native Firebird client library from the host at import time
-        # -- it doesn't bundle that library itself, so it has to come from
-        # the system package manager, only for this leg of the matrix.
-        if: matrix.dialect == 'firebird'
-        run: |
-          sudo apt-get update
-          sudo apt-get install -y libfbclient2
-      - name: Run testcontainers db_engine_specs tests (${{ matrix.dialect }})
-        # nightly_only dialects are already excluded from the matrix itself
-        # on pull_request runs (see the compute-matrix job above), so this
-        # step needs no additional gating.
-        run: |
-          pytest --durations-min=2 -v -m testcontainers \
-            ./tests/testcontainers/db_engine_specs/test_${{ matrix.dialect }}.py \
-            --junit-xml=test-results/junit-testcontainers-${{ matrix.dialect }}.xml
-      - name: Upload JUnit test results
-        if: always()
-        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
-        with:
-          name: junit-results-testcontainers-${{ matrix.dialect }}
-          path: test-results/
-          retention-days: 7
-
-  actions-timeline:
-    needs: [testcontainers]
-    if: always()
-    runs-on: ubuntu-26.04
-    permissions:
-      actions: read
-    steps:
-      - uses: Kesin11/actions-timeline@57fc93f20c6da7fbc14063c6d24a2a5627c799ad # v3.2.0

UPDATING.md

@@ -25,7 +25,7 @@ assists people when migrating to a new version.
 ## Next
 
 - `SAMPLES_ROW_LIMIT` is now the default for `/datasource/samples` requests without a valid explicit `per_page`, rather than a hard per-request ceiling; explicit limits are honored up to the existing global row-limit ceiling, matching `/chart/data` SAMPLES requests.
-- The `cockroachdb` extra (`pip install apache-superset[cockroachdb]`) now installs `sqlalchemy-cockroachdb` instead of the abandoned `cockroachdb` package, whose SQLAlchemy dialect could not be imported under SQLAlchemy 2.0. Existing environments with the old package installed must `pip uninstall cockroachdb` before reinstalling the extra -- both packages register the same `cockroachdb` SQLAlchemy dialect entry point, so leaving the old one in place can still load the abandoned implementation.
+- The `cockroachdb` extra (`pip install apache-superset[cockroachdb]`) now installs `sqlalchemy-cockroachdb` instead of the abandoned `cockroachdb` package, whose SQLAlchemy dialect could not be imported under SQLAlchemy 2.0. Existing environments with the old package installed should `pip uninstall cockroachdb && pip install sqlalchemy-cockroachdb` (or simply reinstall the extra) to restore CockroachDB connectivity.
 
 ### MCP tool results preserve stored string values
 

pyproject.toml

@@ -148,9 +148,10 @@ clickhouse = ["clickhouse-connect>=1.7.2, <2.0"]
 # 2.0). sqlalchemy-cockroachdb is the actively maintained replacement,
 # already linked from CockroachDbEngineSpec.metadata's docs_url, and
 # registers the same `cockroachdb` SQLAlchemy dialect entry point.
-# sqlalchemy-cockroachdb itself declares no DBAPI dependency (its own docs
-# require picking one), so pull in the same psycopg2-binary pin as the
-# `postgres` extra -- CockroachDB speaks the Postgres wire protocol.
+# sqlalchemy-cockroachdb depends only on SQLAlchemy itself, not on a DBAPI
+# driver, so psycopg2-binary is pinned alongside it (matching the `postgres`
+# extra) to keep this extra self-contained -- CockroachDB speaks the
+# PostgreSQL wire protocol, so psycopg2 is what actually opens connections.
 cockroachdb = ["sqlalchemy-cockroachdb>=2.0.0, <3", "psycopg2-binary==2.9.12"]
 crate = ["sqlalchemy-cratedb>=0.43.1, <1"]
 # sqlalchemy-d1's only release (0.1.0, Nov 2025) pins sqlalchemy<2,>=1.4,
@@ -222,11 +223,6 @@ impala = ["impyla>=0.24.0, <0.25"]
 # superset/db_engine_specs/kusto.py's known_incompatibilities metadata.
 kusto = ["sqlalchemy-kusto>=3.1.2, <4"]
 kylin = ["kylinpy>=2.8.4, <2.9"]
-# MariaDB is a MySQL fork implementing the same wire protocol - connects via
-# the plain mysql dialect, same driver as mysql.
-mariadb = ["apache-superset[mysql]"]
-monetdb = ["sqlalchemy-monetdb>=2.1.0, <3", "pymonetdb>=1.9.1, <2"]
-mongodb = ["pymongosql>=0.7.3, <1"]
 mssql = ["pymssql>=2.3.13, <3"]
 # motherduck is an alias for duckdb - MotherDuck works via the duckdb driver
 motherduck = ["apache-superset[duckdb]"]
@@ -275,9 +271,6 @@ tdengine = [
     "taos-ws-py>=0.7.0"
 ]
 teradata = ["teradatasql>=20.0.0.66"]
-# TimescaleDB is a genuine Postgres extension, not a fork - connects via the
-# plain postgresql dialect, same driver as postgres.
-timescaledb = ["apache-superset[postgres]"]
 thumbnails = [] # deprecated, will be removed in 7.0
 vertica = ["sqlalchemy-vertica-python>= 0.6.3, < 0.7"]
 netezza = ["nzalchemy>= 11.1.2, < 11.2"]
@@ -285,18 +278,11 @@ starrocks = ["starrocks>=1.3.4, <2"]
 doris = ["pydoris>=1.2.0, <2.0.0"]
 oceanbase = ["oceanbase_py>=0.0.1.2"]
 ydb = ["ydb-sqlalchemy>=0.1.22", "ydb-sqlglot-plugin>=0.2.8"]
-# YugabyteDB's YSQL layer is fully Postgres-wire compatible - connects via
-# the plain postgresql dialect, same driver as postgres.
-yugabytedb = ["apache-superset[postgres]"]
 development = [
     # no bounds for apache-superset-extensions-cli until a stable version
     "apache-superset-extensions-cli",
     "boto3",
-    # 7.0.0 raises `docker.errors.DockerException: ... Not supported URL
-    # scheme http+docker` against the requests/urllib3 versions pinned
-    # elsewhere in this file -- breaks testcontainers (tests/testcontainers/)
-    # before any container even starts. 7.2.0 is confirmed working.
-    "docker>=7.2.0",
+    "docker",
     "flask-testing",
     "freezegun",
     "grpcio>=1.82.1",

pytest.ini

@@ -19,13 +19,7 @@ testpaths =
     tests
 python_files = *_test.py test_*.py *_tests.py *viz/utils.py
 # `-p no:warnings` temporarily disabled in favor of more finely tuned `filterwarnings`.
-# `not testcontainers` excludes tests/testcontainers/ by default: those spin up
-# real Docker containers, and `testpaths = tests` would otherwise pull them into
-# every plain `pytest` run. The dedicated CI job (testcontainers.yml) overrides
-# this with an explicit `-m testcontainers` to run them.
-addopts = -m "not testcontainers"
-markers =
-    testcontainers: exercises a real database via testcontainers-python (needs Docker); excluded by default, see .github/workflows/testcontainers.yml
+#addopts = -p no:warnings
 asyncio_mode = auto
 
 # `ignore` is effectively equivalent to `-p no:warnings`.

requirements/development.in

@@ -16,32 +16,5 @@
 # specific language governing permissions and limitations
 # under the License.
 #
--e .[development,bigquery,clickhouse,cockroachdb,crate,databend,druid,duckdb,elasticsearch,fastmcp,firebird,gevent,gsheets,monetdb,mongodb,mssql,mysql,oracle,postgres,presto,prophet,risingwave,starrocks,trino,thumbnails,ydb]
+-e .[development,bigquery,cockroachdb,druid,duckdb,fastmcp,gevent,gsheets,mysql,postgres,presto,prophet,trino,thumbnails]
 -e ./superset-extensions-cli[test]
-# testcontainers-backed db_engine_specs tests (tests/testcontainers/) --
-# see .github/workflows/testcontainers.yml
-#
-# `db2` (the `ibm-db-sa`/`ibm-db` driver) and `oceanbase` (the `oceanbase_py`
-# driver) are both deliberately left out of the baseline dev install above:
-# `ibm-db` ships no Linux arm64 wheel, breaking the multi-platform
-# (amd64+arm64) dev Docker image build; `oceanbase_py` pins
-# `sqlalchemy-utils>=0.38.3,<0.39`, which conflicts outright with Superset's
-# own `sqlalchemy-utils==0.42.1` pin -- there's no version of both that can
-# coexist in one resolved environment. Both testcontainers CI jobs install
-# their driver on demand instead, only for their own matrix leg -- see
-# .github/workflows/testcontainers.yml.
-#
-# mariadb/timescaledb/yugabytedb need no testcontainers extra of their own:
-# they reuse the postgres/mysql container classes pointed at a different
-# image, and psycopg2-binary/mysqlclient are already pulled in above via
-# the postgres/mysql extras. Plain postgres/mysql obviously need nothing
-# extra either. clickhouse and starrocks also need no testcontainers extra:
-# ClickHouseContainer has no driver import of its own (clickhouse-connect,
-# pulled in above via the clickhouse extra, is all the test needs), and
-# StarRocks has no dedicated testcontainers module at all -- its test uses
-# a generic DockerContainer plus the same mysqlclient the mysql extra
-# already provides. databend/risingwave/firebird/ydb are the same story:
-# none has a dedicated testcontainers module, so each test uses a generic
-# DockerContainer plus whatever driver its own extra above already
-# provides.
-testcontainers[cockroachdb,cratedb,mongodb,mssql,mysql,oracle,postgres,trino]>=4.15.0,<5

requirements/development.txt

@@ -12,17 +12,10 @@
     #   apache-superset
 aiofile==3.9.0
     # via py-key-value-aio
-aiohappyeyeballs==2.7.1
-    # via aiohttp
-aiohttp==3.14.3
-    # via ydb
-aiosignal==1.4.0
-    # via aiohttp
 alembic==1.15.2
     # via
     #   -c requirements/base-constraint.txt
     #   flask-migrate
-    #   starrocks
 amqp==5.3.1
     # via
     #   -c requirements/base-constraint.txt
@@ -31,8 +24,6 @@ annotated-types==0.7.0
     # via
     #   -c requirements/base-constraint.txt
     #   pydantic
-antlr4-python3-runtime==4.13.2
-    # via pymongosql
 anyio==4.11.0
     # via
     #   httpx
@@ -51,12 +42,9 @@ apsw==3.50.1.0
     #   shillelagh
 astroid==3.3.10
     # via pylint
-asyncmy2==0.2.21
-    # via starrocks
 attrs==25.3.0
     # via
     #   -c requirements/base-constraint.txt
-    #   aiohttp
     #   cattrs
     #   cyclopts
     #   jsonschema
@@ -77,7 +65,6 @@ backports-tarfile==1.2.0
 backports-zstd==1.6.0
     # via
     #   -c requirements/base-constraint.txt
-    #   clickhouse-connect
     #   flask-compress
 bcrypt==4.3.0
     # via
@@ -130,11 +117,8 @@ celery==5.6.3
 certifi==2026.5.20
     # via
     #   -c requirements/base-constraint.txt
-    #   clickhouse-connect
-    #   elasticsearch
     #   httpcore
     #   httpx
-    #   opensearch-py
     #   requests
 cffi==2.0.0
     # via
@@ -176,8 +160,6 @@ click-repl==0.3.0
     # via
     #   -c requirements/base-constraint.txt
     #   celery
-clickhouse-connect==1.7.2
-    # via apache-superset
 cmdstanpy==1.1.0
     # via prophet
 colorama==0.4.6
@@ -189,8 +171,6 @@ contourpy==1.0.7
     # via matplotlib
 coverage==7.6.8
     # via pytest-cov
-crate==2.2.1
-    # via sqlalchemy-cratedb
 cron-descriptor==1.4.5
     # via
     #   -c requirements/base-constraint.txt
@@ -206,20 +186,14 @@ cryptography==50.0.0
     #   authlib
     #   google-auth
     #   joserfc
-    #   oracledb
     #   paramiko
     #   pyjwt
-    #   pymysql
     #   pyopenssl
     #   secretstorage
 cycler==0.12.1
     # via matplotlib
 cyclopts==4.2.4
     # via fastmcp-slim
-databend-driver==0.34.2
-    # via databend-sqlalchemy
-databend-sqlalchemy==0.5.5
-    # via apache-superset
 db-dtypes==1.3.1
     # via pandas-gbq
 defusedxml==0.7.1
@@ -242,11 +216,8 @@ dnspython==2.7.0
     # via
     #   -c requirements/base-constraint.txt
     #   email-validator
-    #   pymongo
-docker==7.2.0
-    # via
-    #   apache-superset
-    #   testcontainers
+docker==7.0.0
+    # via apache-superset
 docstring-parser==0.17.0
     # via cyclopts
 docutils==0.22.2
@@ -257,10 +228,6 @@ duckdb==1.5.5
     #   duckdb-engine
 duckdb-engine==0.17.0
     # via apache-superset
-elasticsearch==7.17.13
-    # via elasticsearch-dbapi
-elasticsearch-dbapi==0.2.13
-    # via apache-superset
 email-validator==2.2.0
     # via
     #   -c requirements/base-constraint.txt
@@ -270,8 +237,6 @@ et-xmlfile==2.0.0
     # via
     #   -c requirements/base-constraint.txt
     #   openpyxl
-events==0.5
-    # via opensearch-py
 exceptiongroup==1.3.0
     # via fastmcp-slim
 fastmcp==3.4.7
@@ -282,10 +247,6 @@ filelock==3.20.3
     # via
     #   -c requirements/base-constraint.txt
     #   virtualenv
-firebird-base==2.0.3
-    # via firebird-driver
-firebird-driver==2.0.3
-    # via sqlalchemy-firebird
 flask==2.3.3
     # via
     #   -c requirements/base-constraint.txt
@@ -366,18 +327,12 @@ fonttools==4.60.2
     # via matplotlib
 freezegun==1.5.1
     # via apache-superset
-frozenlist==1.8.0
-    # via
-    #   aiohttp
-    #   aiosignal
 future==1.0.0
     # via pyhive
 geographiclib==2.0
     # via
     #   -c requirements/base-constraint.txt
     #   geopy
-geojson==3.3.0
-    # via sqlalchemy-cratedb
 geopy==2.4.1
     # via
     #   -c requirements/base-constraint.txt
@@ -434,7 +389,6 @@ grpcio==1.83.0
     #   apache-superset
     #   google-api-core
     #   grpcio-status
-    #   ydb
 grpcio-status==1.60.1
     # via google-api-core
 gunicorn==26.2.0
@@ -460,7 +414,6 @@ httpx==0.28.1
     # via
     #   fastmcp-slim
     #   mcp
-    #   testcontainers
 httpx-sse==0.4.1
     # via mcp
 humanize==4.12.3
@@ -477,7 +430,6 @@ idna==3.15
     #   httpx
     #   requests
     #   url-normalize
-    #   yarl
 importlib-metadata==8.7.0
     # via
     #   keyring
@@ -516,7 +468,6 @@ jmespath==1.1.0
     # via
     #   boto3
     #   botocore
-    #   pymongosql
 joserfc==1.7.2
     # via fastmcp-slim
 jsonpath-ng==1.8.0
@@ -549,8 +500,6 @@ kombu==5.6.2
     # via
     #   -c requirements/base-constraint.txt
     #   celery
-lark==1.3.1
-    # via starrocks
 lazy-object-proxy==1.10.0
     # via openapi-spec-validator
 limits==5.1.0
@@ -558,9 +507,7 @@ limits==5.1.0
     #   -c requirements/base-constraint.txt
     #   flask-limiter
 lz4==4.4.5
-    # via
-    #   clickhouse-connect
-    #   trino
+    # via trino
 mako==1.4.1
     # via
     #   -c requirements/base-constraint.txt
@@ -620,10 +567,6 @@ msgspec==0.19.0
     # via
     #   -c requirements/base-constraint.txt
     #   flask-session
-multidict==6.7.1
-    # via
-    #   aiohttp
-    #   yarl
 mysqlclient==2.2.8
     # via apache-superset
 nh3==0.3.6
@@ -662,22 +605,14 @@ openpyxl==3.1.5
     # via
     #   -c requirements/base-constraint.txt
     #   pandas
-opensearch-py==2.8.0
-    # via elasticsearch-dbapi
 opentelemetry-api==1.39.1
     # via fastmcp-slim
-oracledb==4.0.2
-    # via
-    #   apache-superset
-    #   testcontainers
 ordered-set==4.1.0
     # via
     #   -c requirements/base-constraint.txt
     #   flask-limiter
 orjson==3.11.9
-    # via
-    #   crate
-    #   trino
+    # via trino
 packaging==25.0
     # via
     #   -c requirements/base-constraint.txt
@@ -685,8 +620,8 @@ packaging==25.0
     #   apispec
     #   db-dtypes
     #   deprecation
+    #   docker
     #   duckdb-engine
-    #   elasticsearch-dbapi
     #   fastmcp-slim
     #   google-cloud-bigquery
     #   kombu
@@ -696,8 +631,6 @@ packaging==25.0
     #   pytest
     #   shillelagh
     #   sqlalchemy-bigquery
-    #   sqlalchemy-firebird
-    #   ydb
 pandas==2.3.3
     # via
     #   -c requirements/base-constraint.txt
@@ -759,22 +692,16 @@ prompt-toolkit==3.0.51
     # via
     #   -c requirements/base-constraint.txt
     #   click-repl
-propcache==0.5.2
-    # via
-    #   aiohttp
-    #   yarl
 prophet==1.4.0
     # via apache-superset
 proto-plus==1.25.0
     # via google-api-core
 protobuf==5.29.6
     # via
-    #   firebird-base
     #   google-api-core
     #   googleapis-common-protos
     #   grpcio-status
     #   proto-plus
-    #   ydb
 psutil==6.1.0
     # via
     #   apache-superset
@@ -848,24 +775,6 @@ pyjwt==2.13.0
     #   mcp
 pylint==3.3.7
     # via apache-superset
-pymonetdb==1.9.1
-    # via
-    #   apache-superset
-    #   sqlalchemy-monetdb
-pymongo==4.17.0
-    # via
-    #   pymongosql
-    #   testcontainers
-pymongosql==0.7.3
-    # via apache-superset
-pymssql==2.3.13
-    # via
-    #   apache-superset
-    #   testcontainers
-pymysql==1.2.0
-    # via
-    #   starrocks
-    #   testcontainers
 pynacl==1.6.2
     # via
     #   -c requirements/base-constraint.txt
@@ -910,13 +819,11 @@ python-dateutil==2.9.0.post0
     #   botocore
     #   celery
     #   croniter
-    #   firebird-driver
     #   flask-appbuilder
     #   freezegun
     #   google-cloud-bigquery
     #   holidays
     #   matplotlib
-    #   opensearch-py
     #   pandas
     #   pyhive
     #   shillelagh
@@ -927,7 +834,6 @@ python-dotenv==1.2.2
     #   apache-superset
     #   fastmcp-slim
     #   pydantic-settings
-    #   testcontainers
 python-ldap==3.4.7
     # via apache-superset
 python-multipart==0.0.29
@@ -971,7 +877,6 @@ requests==2.33.0
     #   google-api-core
     #   google-cloud-bigquery
     #   jsonschema-path
-    #   opensearch-py
     #   pydruid
     #   pyhive
     #   requests-cache
@@ -1050,9 +955,7 @@ sqlalchemy==2.0.52
     #   alembic
     #   apache-superset
     #   apache-superset-core
-    #   databend-sqlalchemy
     #   duckdb-engine
-    #   elasticsearch-dbapi
     #   flask-appbuilder
     #   flask-sqlalchemy
     #   marshmallow-sqlalchemy
@@ -1060,14 +963,7 @@ sqlalchemy==2.0.52
     #   sqlalchemy-bigquery
     #   sqlalchemy-cockroachdb
     #   sqlalchemy-continuum
-    #   sqlalchemy-cratedb
-    #   sqlalchemy-firebird
-    #   sqlalchemy-monetdb
-    #   sqlalchemy-risingwave
     #   sqlalchemy-utils
-    #   starrocks
-    #   testcontainers
-    #   ydb-sqlalchemy
 sqlalchemy-bigquery==1.17.2
     # via apache-superset
 sqlalchemy-cockroachdb==2.0.4
@@ -1076,16 +972,6 @@ sqlalchemy-continuum==1.7.0
     # via
     #   -c requirements/base-constraint.txt
     #   apache-superset
-sqlalchemy-cratedb==0.43.1
-    # via
-    #   apache-superset
-    #   testcontainers
-sqlalchemy-firebird==2.2.0
-    # via apache-superset
-sqlalchemy-monetdb==2.1.0
-    # via apache-superset
-sqlalchemy-risingwave==2.1.0
-    # via apache-superset
 sqlalchemy-utils==0.42.1
     # via
     #   -c requirements/base-constraint.txt
@@ -1097,7 +983,6 @@ sqlglot==30.17.0
     #   -c requirements/base-constraint.txt
     #   apache-superset
     #   apache-superset-core
-    #   ydb-sqlglot-plugin
 sqloxide==0.1.51
     # via apache-superset
 sse-starlette==3.0.2
@@ -1110,8 +995,6 @@ starlette==1.3.1
     # via
     #   fastmcp-slim
     #   mcp
-starrocks==1.3.4
-    # via apache-superset
 statsd==4.0.1
     # via apache-superset
 syntaqlite==0.9.0
@@ -1120,8 +1003,6 @@ tabulate==0.10.0
     # via
     #   -c requirements/base-constraint.txt
     #   apache-superset
-testcontainers==4.15.0
-    # via -r requirements/development.in
 tiktoken==0.14.0
     # via apache-superset
 tomli-w==1.2.0
@@ -1133,14 +1014,10 @@ tqdm==4.67.1
     #   cmdstanpy
     #   prophet
 trino==0.339.0
-    # via
-    #   apache-superset
-    #   testcontainers
+    # via apache-superset
 typing-extensions==4.16.0
     # via
     #   -c requirements/base-constraint.txt
-    #   aiohttp
-    #   aiosignal
     #   alembic
     #   anyio
     #   apache-superset
@@ -1152,7 +1029,6 @@ typing-extensions==4.16.0
     #   limits
     #   mcp
     #   opentelemetry-api
-    #   oracledb
     #   py-key-value-aio
     #   pydantic
     #   pydantic-core
@@ -1161,7 +1037,6 @@ typing-extensions==4.16.0
     #   shillelagh
     #   sqlalchemy
     #   starlette
-    #   testcontainers
     #   typing-inspection
 typing-inspection==0.4.2
     # via
@@ -1189,22 +1064,13 @@ urllib3==2.7.0
     # via
     #   -c requirements/base-constraint.txt
     #   botocore
-    #   clickhouse-connect
-    #   crate
     #   docker
-    #   elasticsearch
-    #   opensearch-py
     #   requests
     #   requests-cache
-    #   testcontainers
 uvicorn==0.37.0
     # via
     #   fastmcp-slim
     #   mcp
-verlib2==0.3.2
-    # via
-    #   crate
-    #   sqlalchemy-cratedb
 vine==5.1.0
     # via
     #   -c requirements/base-constraint.txt
@@ -1238,7 +1104,6 @@ wrapt==1.17.2
     # via
     #   -c requirements/base-constraint.txt
     #   deprecated
-    #   testcontainers
 wtforms==3.2.2
     # via
     #   -c requirements/base-constraint.txt
@@ -1259,18 +1124,6 @@ xlsxwriter==3.2.9
     #   -c requirements/base-constraint.txt
     #   apache-superset
     #   pandas
-yarl==1.24.5
-    # via aiohttp
-ydb==3.31.4
-    # via
-    #   ydb-dbapi
-    #   ydb-sqlalchemy
-ydb-dbapi==0.1.23
-    # via ydb-sqlalchemy
-ydb-sqlalchemy==0.1.22
-    # via apache-superset
-ydb-sqlglot-plugin==0.2.8
-    # via apache-superset
 zipp==3.23.0
     # via importlib-metadata
 zope-event==5.0

superset/db_engine_specs/__init__.py

@@ -41,7 +41,6 @@ import sqlalchemy.dialects
 from flask import current_app as app
 from sqlalchemy.engine.default import DefaultDialect
 from sqlalchemy.exc import NoSuchModuleError
-from sqlalchemy.sql import compiler as sqla_compiler
 
 from superset import feature_flag_manager
 from superset.db_engine_specs.base import BaseEngineSpec
@@ -160,16 +159,6 @@ def get_available_engine_specs() -> dict[type[BaseEngineSpec], set[str]]:  # noq
             continue
 
     # installed 3rd-party dialects
-    #
-    # `ep.load()` runs arbitrary module-level code in the third-party package.
-    # Some dialects (e.g. sqlalchemy-monetdb) mutate SQLAlchemy's shared,
-    # process-global `compiler.OPERATORS` mapping in place on import instead
-    # of subclassing it, which would otherwise silently change SQL rendering
-    # (e.g. `!=` -> `<>`) for every dialect for the rest of the process, not
-    # just the misbehaving one. Snapshot/restore around each load so a
-    # buggy connector can't leak global compiler state into unrelated
-    # dialects just because it was enumerated here.
-    operators_snapshot = dict(sqla_compiler.OPERATORS)
     for ep in entry_points(group="sqlalchemy.dialects"):
         try:
             dialect = ep.load()
@@ -204,10 +193,6 @@ def get_available_engine_specs() -> dict[type[BaseEngineSpec], set[str]]:  # noq
             if isinstance(driver, bytes):
                 driver = driver.decode()
             drivers[backend].add(driver)
-        finally:
-            if sqla_compiler.OPERATORS != operators_snapshot:
-                sqla_compiler.OPERATORS.clear()
-                sqla_compiler.OPERATORS.update(operators_snapshot)
 
     dbs_denylist = app.config["DBS_AVAILABLE_DENYLIST"]
     if not feature_flag_manager.is_feature_enabled("ENABLE_SUPERSET_META_DB"):

superset/db_engine_specs/cockroachdb.py

@@ -44,10 +44,7 @@ class CockroachDbEngineSpec(PostgresEngineSpec):
             DatabaseCategory.TRADITIONAL_RDBMS,
             DatabaseCategory.OPEN_SOURCE,
         ],
-        # sqlalchemy-cockroachdb declares no DBAPI dependency of its own (see
-        # the comment on the `cockroachdb` extra in pyproject.toml), so a
-        # plain `cockroachdb://` URL also needs psycopg2 installed to connect.
-        "pypi_packages": ["sqlalchemy-cockroachdb", "psycopg2-binary"],
+        "pypi_packages": ["sqlalchemy-cockroachdb", "psycopg2"],
         "connection_string": "cockroachdb://root@{hostname}:{port}/{database}?sslmode=disable",
         "default_port": 26257,
         "docs_url": "https://github.com/cockroachdb/sqlalchemy-cockroachdb",

tests/integration_tests/databases/api_tests.py

@@ -2487,15 +2487,8 @@ class TestDatabaseApi(SupersetTestCase):
         }
         assert response == expected_response
 
-        # Uses a `dialect+driver://` URI (rather than the bare `broken://`
-        # above) to also cover engine-name extraction stripping the driver
-        # suffix. The dialect itself ("broken") must stay one that no
-        # installed extra ever registers a real SQLAlchemy plugin for --
-        # this PR's own testcontainers extras (mssql, oracle, db2, ...)
-        # install real drivers for those dialects, which would make this
-        # URI actually attempt a connection instead of failing to load.
         data = {
-            "sqlalchemy_uri": "broken+driver://url",
+            "sqlalchemy_uri": "mssql+pymssql://url",
             "database_name": "examples",
             "impersonate_user": False,
             "server_cert": None,
@@ -2507,7 +2500,7 @@ class TestDatabaseApi(SupersetTestCase):
         expected_response = {
             "errors": [
                 {
-                    "message": "Could not load database driver for: broken",
+                    "message": "Could not load database driver for: mssql",
                     "error_type": "GENERIC_COMMAND_ERROR",
                     "level": "warning",
                     "extra": {

tests/testcontainers/__init__.py

@@ -1,16 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.

tests/testcontainers/db_engine_specs/__init__.py

@@ -1,16 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.

tests/testcontainers/db_engine_specs/_driver.py

@@ -1,47 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Shared import guard for the per-dialect testcontainers modules
-(tests/testcontainers/db_engine_specs/test_*.py), each of which needs its
-own optional `testcontainers[...]` driver submodule to even import.
-"""
-
-import importlib
-import os
-
-
-def require_driver(module_name: str) -> None:
-    """
-    Import `module_name`, a dialect's `testcontainers` driver submodule.
-
-    Most environments treat that driver as optional: a bare local `pytest`
-    run, or another CI job that never installed the `testcontainers` extras,
-    should skip the module rather than fail collection outright.
-
-    The dedicated per-dialect CI job (.github/workflows/testcontainers.yml)
-    sets SUPERSET_TESTCONTAINERS_STRICT, because there the driver is not
-    optional -- that job's matrix installs exactly this one driver for
-    exactly this one module. A broken or missing import there means the job
-    is misconfigured, and should fail loudly instead of silently reporting
-    a misleadingly green, zero-tests-run result.
-    """
-    if os.environ.get("SUPERSET_TESTCONTAINERS_STRICT"):
-        importlib.import_module(module_name)
-    else:
-        import pytest
-
-        pytest.importorskip(module_name)

tests/testcontainers/db_engine_specs/_pagination.py

@@ -1,68 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Shared body for the "paginated query returns correct rows in order" test
-that db_engine_specs.{cockroachdb,crate,db2,mssql,oracle,trino}'s
-testcontainers suites each run against their own real instance: a plain
-SQLAlchemy Core LIMIT/OFFSET query, compiled and executed for real. Mocked
-tests cannot catch a dialect compiling this incorrectly (see
-apache/superset#42899, where Trino emitted OFFSET before LIMIT) -- only
-real execution can.
-
-Each call site keeps its own test function (and dialect-specific docstring)
-so failures still report against the right module; this only factors out
-the identical table setup/assert body, via an optional post-insert hook for
-dialects (CrateDB) that need one, and an optional extra-table-args hook for
-dialects (ClickHouse) whose CREATE TABLE requires a schema item a plain
-Column/primary key can't express.
-"""
-
-from collections.abc import Callable
-from typing import Any
-
-from sqlalchemy import Column, insert, Integer, MetaData, select, Table as SATable
-from sqlalchemy.engine import Connection, Engine
-
-
-def assert_paginated_query_returns_correct_rows_in_order(
-    engine: Engine,
-    after_insert: Callable[[Connection], None] | None = None,
-    extra_table_args: tuple[Any, ...] = (),
-) -> None:
-    metadata = MetaData()
-    t = SATable(
-        "pilot_pagination",
-        metadata,
-        # autoincrement=False: a single-column integer primary key otherwise
-        # implicitly becomes AUTO_INCREMENT on MySQL/MariaDB. That column
-        # type treats an explicit 0 as NULL by default (NO_AUTO_VALUE_ON_ZERO
-        # is off), so the id=0 row below would silently get auto-assigned 1,
-        # colliding with the explicit id=1 row in the same batch insert.
-        Column("id", Integer, primary_key=True, autoincrement=False),
-        *extra_table_args,
-    )
-    metadata.create_all(engine)
-    with engine.begin() as conn:
-        conn.execute(insert(t), [{"id": i} for i in range(10)])
-        if after_insert is not None:
-            after_insert(conn)
-
-    with engine.connect() as conn:
-        stmt = select(t.c.id).order_by(t.c.id).limit(3).offset(4)
-        rows = conn.execute(stmt).fetchall()
-
-    assert [row.id for row in rows] == [4, 5, 6]

tests/testcontainers/db_engine_specs/test_clickhouse.py

@@ -1,130 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.clickhouse against a real ClickHouse instance, spun
-up on demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-Superset's recommended ClickHouse connector is `clickhouse-connect`
-(`ClickHouseConnectEngineSpec`, engine "clickhousedb"), which talks HTTP,
-not `ClickHouseContainer`'s own documented `clickhouse_driver` (a different,
-native-TCP-protocol package Superset doesn't use at all). The container
-exposes both the native TCP port (9000) and the HTTP port (8123); this test
-connects over the HTTP port to match Superset's actual driver.
-
-Unlike every other dialect in this suite, ClickHouse tables have no real
-primary key/constraint concept -- CREATE TABLE requires an explicit engine
-(e.g. MergeTree), or clickhouse-connect's DDL compiler raises a CompileError
-rather than defaulting to one.
-
-`superset.db_engine_specs.clickhouse` runs module-level setup code (default
-type-formatting overrides) that dereferences `current_app.config` whenever
-clickhouse-connect is installed, so importing it outside a Flask app context
-raises RuntimeError the first time it's imported in a process.
-`tests/unit_tests/db_engine_specs/test_clickhouse.py` gets an app context
-for free from that suite's autouse fixture; this suite has no such fixture,
-so this test pushes one explicitly around just that one-time import, reusing
-the real app instance `tests/conftest.py` already builds for the rest of the
-test run rather than constructing a second one.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.clickhouse")
-require_driver("clickhouse_connect")
-
-from clickhouse_connect.cc_sqlalchemy.ddl.tableengine import MergeTree  # noqa: E402
-from testcontainers.community.clickhouse import ClickHouseContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-HTTP_PORT = 8123
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with ClickHouseContainer("clickhouse/clickhouse-server:latest") as container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(HTTP_PORT)
-        yield create_engine(
-            f"clickhousedb://{container.username}:{container.password}"
-            f"@{host}:{port}/{container.dbname}"
-        )
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(
-        engine, extra_table_args=(MergeTree(order_by="id"),)
-    )
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    ClickHouseConnectEngineSpec.get_columns wraps a real SQLAlchemy
-    Inspector; this exercises that against actual server-reported column
-    metadata rather than a mocked Inspector.
-    """
-    from tests.integration_tests.test_app import app
-
-    with app.app_context():
-        from superset.db_engine_specs.clickhouse import ClickHouseConnectEngineSpec
-
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-        MergeTree(order_by="id"),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = ClickHouseConnectEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = ClickHouseConnectEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_cockroachdb.py

@@ -1,97 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.cockroachdb against a real CockroachDB instance,
-spun up on demand via testcontainers. Run via
-.github/workflows/testcontainers.yml -- these exercise real SQL execution
-and dialect introspection, which mocked unit tests structurally cannot.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.cockroachdb import CockroachDbEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.cockroachdb")
-
-from testcontainers.community.cockroachdb import CockroachDBContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    # sqlalchemy-cockroachdb registers its dialect under the plain
-    # "cockroachdb" name; the container's own default ("cockroachdb+psycopg2")
-    # matches the abandoned `cockroachdb` package instead (see #43501).
-    with CockroachDBContainer(dialect="cockroachdb") as container:
-        yield create_engine(container.get_connection_url())
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    CockroachDbEngineSpec.get_columns wraps a real SQLAlchemy Inspector;
-    this exercises that against actual server-reported column metadata
-    rather than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = CockroachDbEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = CockroachDbEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_crate.py

@@ -1,103 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.crate against a real CrateDB instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-crate/crate only publishes an amd64 image (no arm64 build), and requires a
-host CPU supporting the x86-64-v3 instruction set -- QEMU emulation on
-Apple Silicon cannot satisfy that, so this file cannot run locally on an
-Apple Silicon machine even with `docker pull --platform linux/amd64`. It
-runs natively on GitHub Actions' x86_64 runners.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Connection, Engine
-
-from superset.db_engine_specs.crate import CrateEngineSpec
-from superset.sql.parse import Table
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.cratedb")
-
-from testcontainers.community.cratedb import CrateDBContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with CrateDBContainer() as container:
-        yield create_engine(container.get_connection_url())
-
-
-def _refresh_pilot_pagination(conn: Connection) -> None:
-    # CrateDB is eventually consistent: a row is not guaranteed visible to
-    # subsequent selects immediately after insert.
-    conn.exec_driver_sql("REFRESH TABLE pilot_pagination")
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(
-        engine, after_insert=_refresh_pilot_pagination
-    )
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    CrateEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = CrateEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = CrateEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None

tests/testcontainers/db_engine_specs/test_databend.py

@@ -1,114 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.databend against a real Databend instance, spun up
-on demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-Databend has no dedicated testcontainers module, so this uses a generic
-DockerContainer against the official `datafuselabs/databend` standalone
-image. Superset's DatabendEngineSpec defaults to `sslmode=require`
-(`encryption_parameters`), but the local standalone image has no TLS
-listener, so this connects with `sslmode=disable` explicitly.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.databend import DatabendEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("databend_sqlalchemy")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import LogMessageWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-HTTP_PORT = 8000
-DBNAME = "default"
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("datafuselabs/databend")
-    container.with_exposed_ports(HTTP_PORT)
-    # The image's own startup banner documents this exact line as proof its
-    # HTTP query endpoint is bound and ready.
-    container.waiting_for(LogMessageWaitStrategy(f"listened at 0.0.0.0:{HTTP_PORT}"))
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(HTTP_PORT)
-        # "root" with no password is the image's builtin user -- confirmed
-        # directly against a running container, not from the image's own
-        # doc text, which only shows ${USER}/${PASSWORD} placeholders.
-        yield create_engine(f"databend://root:@{host}:{port}/{DBNAME}?sslmode=disable")
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    DatabendEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = DatabendEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = DatabendEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_db2.py

@@ -1,95 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.db2 against a real IBM Db2 instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-icr.io/db2_community/db2 only publishes amd64/ppc64le/s390x images (no
-arm64 build), so this cannot run locally on an Apple Silicon machine. It
-runs natively on GitHub Actions' x86_64 runners. Db2 is also a notably slow
-starter (a full instance bring-up, not just a process start) -- expect this
-module alone to take several minutes.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.db2 import Db2EngineSpec
-from superset.sql.parse import Table
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.db2")
-
-from testcontainers.community.db2 import Db2Container  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with Db2Container() as container:
-        yield create_engine(container.get_connection_url())
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    Db2EngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = Db2EngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = Db2EngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None

tests/testcontainers/db_engine_specs/test_elasticsearch.py

@@ -1,107 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.elasticsearch against a real Elasticsearch instance,
-spun up on demand via testcontainers. Run via
-.github/workflows/testcontainers.yml.
-
-Unlike the SQL-native dialects in this directory, Elasticsearch has no
-CREATE TABLE / INSERT: indices and documents get created via its REST API
-(elasticsearch-dbapi's SQLAlchemy dialect is read-focused, translating SQL
-to the _sql endpoint), matching how Superset actually encounters
-Elasticsearch in practice -- data arrives via ingestion tooling, not
-through Superset itself.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-import requests
-from sqlalchemy import create_engine, inspect, text
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.elasticsearch import ElasticSearchEngineSpec
-from superset.sql.parse import Table
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.elasticsearch")
-
-from testcontainers.community.elasticsearch import ElasticSearchContainer  # noqa: E402
-
-INDEX = "pilot_pagination"
-
-
-def _index_document(
-    base_url: str, index: str, doc_id: int, body: dict[str, int]
-) -> None:
-    response = requests.put(f"{base_url}/{index}/_doc/{doc_id}", json=body, timeout=10)
-    response.raise_for_status()
-
-
-def _refresh(base_url: str, index: str) -> None:
-    response = requests.post(f"{base_url}/{index}/_refresh", timeout=10)
-    response.raise_for_status()
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with ElasticSearchContainer("elasticsearch:8.11.0") as container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(container.port)
-        base_url = f"http://{host}:{port}"
-
-        for i in range(10):
-            _index_document(base_url, INDEX, i, {"id": i})
-        _refresh(base_url, INDEX)
-
-        yield create_engine(f"elasticsearch+http://{host}:{port}/")
-
-
-def test_ordered_limited_query_returns_correct_rows(engine: Engine) -> None:
-    """
-    A plain LIMIT query, compiled and executed against a real instance.
-    Mocked tests cannot catch a dialect compiling this incorrectly (see
-    apache/superset#42899, where Trino emitted OFFSET before LIMIT) -- only
-    real execution can. No OFFSET here: Elasticsearch's SQL layer genuinely
-    doesn't support it (a protocol limitation, not a bug -- confirmed
-    against a real instance, which raises a parsing_exception on OFFSET).
-    ElasticSearchEngineSpec.supports_offset = False documents this already.
-    """
-    with engine.connect() as conn:
-        rows = conn.execute(
-            text(f"SELECT id FROM {INDEX} ORDER BY id LIMIT 3")  # noqa: S608
-        ).fetchall()
-
-    assert [row.id for row in rows] == [0, 1, 2]
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    ElasticSearchEngineSpec.get_columns wraps a real SQLAlchemy Inspector;
-    this exercises that against actual server-reported field mappings
-    rather than a mocked Inspector.
-    """
-    inspector = inspect(engine)
-    columns = ElasticSearchEngineSpec.get_columns(inspector, Table(INDEX))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert "id" in by_name
-    spec = ElasticSearchEngineSpec.get_column_spec(str(by_name["id"]["type"]))
-    assert spec is not None

tests/testcontainers/db_engine_specs/test_firebird.py

@@ -1,146 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.firebird against a real Firebird instance, spun up
-on demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-Firebird references a database *file* rather than a server-managed named
-database -- the connection URI is `firebird://user:pass@host:port/<path>`,
-where <path> is the path to a .fdb file on the server. This uses the
-well-known `jacobalberty/firebird` image, which creates that file (per
-FIREBIRD_DATABASE) under /firebird/data on first boot.
-
-FirebirdEngineSpec sets `limit_method = LimitMethod.FETCH_MANY` with a
-comment claiming Firebird "uses FIRST to limit" -- stale relative to the
-modern sqlalchemy-firebird driver, which compiles real ROWS-based
-pagination (confirmed via an offline dialect compile: `SELECT ... ROWS
-4 + 1 TO 4 + 3`, correctly ordered, not a Trino-style bug). That staleness
-affects what Superset's own query layer emits, not what this suite's
-direct dialect-compilation check exercises.
-
-Could not be verified against a real running instance in this
-environment: `firebird-driver` is a pure-Python ctypes wrapper (its wheel
-is `py3-none-any`, confirmed by downloading it directly) that dynamically
-loads the native Firebird client library (`libfbclient`) from the host at
-import time -- it doesn't bundle that library itself. This machine has no
-Homebrew formula or straightforward install path for it. The container
-itself was confirmed to start and pass its own healthcheck; CI installs
-the `libfbclient2` system package separately (see
-.github/workflows/testcontainers.yml) for the actual client-library
-dependency this driver needs.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.firebird import FirebirdEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("firebird.driver")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import HealthcheckWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-PORT = 3050
-PASSWORD = "masterkey"  # noqa: S105 -- fixed test-fixture password, not a secret
-DB_FILE = "test.fdb"
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("jacobalberty/firebird")
-    container.with_exposed_ports(PORT)
-    container.with_env("ISC_PASSWORD", PASSWORD)
-    container.with_env("FIREBIRD_DATABASE", DB_FILE)
-    # The image logs nothing beyond a single startup banner line and never
-    # prints a distinct "ready" message -- it ships its own Docker
-    # HEALTHCHECK instead, confirmed via `docker ps` reporting (healthy).
-    container.waiting_for(HealthcheckWaitStrategy())
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(PORT)
-        eng = create_engine(
-            f"firebird://sysdba:{PASSWORD}@{host}:{port}//firebird/data/{DB_FILE}"
-        )
-        yield eng
-        # firebird-driver registers its own atexit handler that talks to
-        # the Firebird subsystem to shut it down cleanly. Without disposing
-        # here first, that handler fires at interpreter exit against a
-        # server the container has *already* torn down -- confirmed on
-        # real CI as a segfault (exit code 139) after both tests had
-        # already passed. Disposing while the server is still up lets the
-        # driver close out normally, so the later atexit call has nothing
-        # left to talk to.
-        eng.dispose()
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    FirebirdEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = FirebirdEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = FirebirdEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_mariadb.py

@@ -1,117 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.mariadb against a real MariaDB instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-MariaDB is a MySQL fork implementing the same wire protocol: connects via
-the plain "mysql" dialect with mysqlclient, same as vanilla MySQL, just
-pointed at the mariadb image instead of mysql:latest.
-
-Could not be verified locally in this environment: mysqlclient (MySQLdb)
-has a pre-existing, unrelated native-library linking issue against this
-machine's Homebrew-installed libmysqlclient. CI installs it via apt on
-Linux, where this does not occur.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.mariadb import MariaDBEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.mysql")
-
-from testcontainers.community.mysql import MySqlContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with MySqlContainer("mariadb:11") as container:
-        # get_connection_url() has no host override and defaults to
-        # get_container_host_ip(), which is the literal string "localhost"
-        # on native Linux Docker (e.g. GitHub Actions runners). MySQLdb
-        # (mysqlclient) treats a "localhost" host specially and attempts a
-        # Unix socket connection instead of TCP, which fails since there's
-        # no local MySQL socket -- the container is reached over the
-        # network. Only rewrite that specific local case to 127.0.0.1; a
-        # remote Docker daemon reports its own real host/IP here, which
-        # must be preserved so the suite can still reach it.
-        host = container.get_container_host_ip()
-        if host == "localhost":
-            host = "127.0.0.1"
-        port = container.get_exposed_port(container.port)
-        yield create_engine(
-            f"mysql://{container.username}:{container.password}"
-            f"@{host}:{port}/{container.dbname}"
-        )
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    MariaDBEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = MariaDBEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = MariaDBEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_monetdb.py

@@ -1,127 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.monetdb against a real MonetDB instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-monetdb/monetdb publishes an amd64-only image, so this needs Rosetta/QEMU
-emulation on Apple Silicon -- unlike CrateDB's x86-64-v3 CPU requirement,
-this one actually runs fine under emulation (verified locally). No native
-testcontainers module exists for MonetDB, so this uses a generic
-DockerContainer with the documented MDB_* environment variables and waits
-for the daemon's own startup log line.
-"""
-
-import re
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.monetdb import MonetDbEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("sqlalchemy_monetdb")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import (  # noqa: E402
-    CompositeWaitStrategy,
-    LogMessageWaitStrategy,
-    PortWaitStrategy,
-)
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-PORT = 50000
-PASSWORD = "monetdb"  # noqa: S105 -- fixed test-fixture password, not a secret
-DBNAME = "test"
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("monetdb/monetdb:latest")
-    container.with_exposed_ports(PORT)
-    container.with_env("MDB_DB_ADMIN_PASS", PASSWORD)
-    container.with_env("MDB_CREATE_DBS", DBNAME)
-    # The "Starting MonetDB daemon" log line is emitted before the image
-    # actually runs `monetdbd start -n`, so it alone isn't proof the server
-    # is accepting connections yet. Follow it with a port-connect check,
-    # which only succeeds once monetdbd is really listening.
-    container.waiting_for(
-        CompositeWaitStrategy(
-            LogMessageWaitStrategy(re.compile("Starting MonetDB daemon")),
-            PortWaitStrategy(PORT),
-        )
-    )
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(PORT)
-        yield create_engine(f"monetdb://monetdb:{PASSWORD}@{host}:{port}/{DBNAME}")
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    MonetDbEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = MonetDbEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = MonetDbEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_mongodb.py

@@ -1,141 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied. See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.mongodb against a real MongoDB instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-MongoDB is schemaless, and Superset talks to it via `pymongosql`, a
-SQL-to-MongoDB translation layer (dialect requires a `?mode=superset` query
-param -- not part of testcontainers' own MongoDbContainer.get_connection_url()).
-Documents get inserted via the native pymongo driver, not SQL INSERT,
-matching how Superset actually encounters MongoDB in practice and avoiding
-any assumption about pymongosql's own INSERT/DDL support.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    column,
-    create_engine,
-    inspect,
-    Integer,
-    select,
-    table,
-    text,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.mongodb import MongoDBEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.mongodb")
-require_driver("pymongosql")
-
-from testcontainers.community.mongodb import MongoDbContainer  # noqa: E402
-
-COLLECTION = "pilot_pagination"
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with MongoDbContainer("mongo:7.0.7") as container:
-        client = container.get_connection_client()
-        client[container.dbname][COLLECTION].insert_many([{"id": i} for i in range(10)])
-        # MongoDbContainer.get_connection_url() has no database path segment
-        # or query string at all (it only builds user:pass@host:port), so
-        # naively appending "&mode=superset" glues it straight onto the port
-        # number instead of starting a query string. Build the full URL
-        # ourselves instead of relying on string concatenation.
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(container.port)
-        # authSource=admin is required: MongoDbContainer creates its root
-        # user via MONGO_INITDB_ROOT_USERNAME, which lives in the `admin`
-        # database, not in `dbname` -- without it, auth fails against
-        # whatever database is in the URL path.
-        yield create_engine(
-            f"mongodb://{container.username}:{container.password}@{host}:{port}"
-            f"/{container.dbname}?mode=superset&authSource=admin"
-        )
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed
-    against a real instance. Mocked tests cannot catch a dialect compiling
-    this incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can. Unlike Elasticsearch's SQL
-    layer (which has no OFFSET support at all), pymongosql maps OFFSET to
-    MongoDB's native `skip`, so this dialect supports it.
-
-    Compiled with `literal_binds=True`, matching how Superset actually
-    issues chart/SQL Lab queries (see `models/helpers.py`'s
-    `get_query_str_extended`): pymongosql's SQL-to-Mongo AST parser reads
-    LIMIT/OFFSET straight off the compiled SQL text ahead of parameter
-    substitution, so a bound `LIMIT ?`/`OFFSET ?` placeholder is rejected
-    ("invalid literal for int() with base 10: '?'") and the clause is
-    silently dropped -- unlike its WHERE-clause parameter handling, which
-    does substitute correctly. Literal binds sidestep that and exercise
-    the dialect's actual LIMIT/OFFSET compilation, per this test's intent.
-
-    Uses a bare `column("id")`/`table(...)` pair rather than a full
-    `Table`-bound column: SQLAlchemy always qualifies a Table-bound column
-    reference as `pilot_pagination.id` once there's a FROM clause, and
-    pymongosql's projection builder takes that qualified text completely
-    literally as a MongoDB field path -- `{"pilot_pagination.id": 1}` reads
-    a *nested* field under a top-level `pilot_pagination` key, which
-    doesn't exist on these flat documents, silently projecting None instead
-    of raising. An unbound column compiles unqualified ("id"), which
-    resolves correctly, while still exercising the dialect's own
-    LIMIT/OFFSET compilation via a real Core `select()`.
-    """
-    id_col = column("id")
-    stmt = (
-        select(id_col)
-        .select_from(table(COLLECTION))
-        .order_by(id_col)
-        .limit(3)
-        .offset(4)
-    )
-    compiled = stmt.compile(engine, compile_kwargs={"literal_binds": True})
-    with engine.connect() as conn:
-        rows = conn.execute(text(str(compiled))).fetchall()
-
-    assert [row.id for row in rows] == [4, 5, 6]
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    MongoDBEngineSpec.get_columns wraps a real SQLAlchemy Inspector, which
-    pymongosql implements by sampling real documents to infer column types
-    -- this exercises that against an actual running instance rather than
-    a mocked Inspector.
-    """
-    inspector = inspect(engine)
-    columns = MongoDBEngineSpec.get_columns(inspector, Table(COLLECTION))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert "id" in by_name
-    spec = MongoDBEngineSpec.get_column_spec(str(by_name["id"]["type"]))
-    assert spec is not None
-    assert spec.generic_type == GenericDataType.NUMERIC
-    assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_mssql.py

@@ -1,93 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.mssql against a real SQL Server instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-mcr.microsoft.com/mssql/server only publishes an amd64 image (SQL Server on
-Linux has no ARM build), so this cannot run locally on an Apple Silicon
-machine. It runs natively on GitHub Actions' x86_64 runners.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.mssql import MssqlEngineSpec
-from superset.sql.parse import Table
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.mssql")
-
-from testcontainers.community.mssql import SqlServerContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with SqlServerContainer() as container:
-        yield create_engine(container.get_connection_url())
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    MssqlEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = MssqlEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = MssqlEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None

tests/testcontainers/db_engine_specs/test_mysql.py

@@ -1,117 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.mysql against a real MySQL instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-Plain MySQL itself was never covered by this suite: MariaDB and StarRocks
-both reuse `MySQLEngineSpec`'s plain "mysql" dialect via mysqlclient, but
-neither stands in for vanilla MySQL server's own dialect quirks.
-
-Could not be verified locally in this environment: mysqlclient (MySQLdb)
-has a pre-existing, unrelated native-library linking issue against this
-machine's Homebrew-installed libmysqlclient. CI installs it via apt on
-Linux, where this does not occur.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.mysql import MySQLEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.mysql")
-
-from testcontainers.community.mysql import MySqlContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with MySqlContainer("mysql:8.0") as container:
-        # get_connection_url() has no host override and defaults to
-        # get_container_host_ip(), which is the literal string "localhost"
-        # on native Linux Docker (e.g. GitHub Actions runners). MySQLdb
-        # (mysqlclient) treats a "localhost" host specially and attempts a
-        # Unix socket connection instead of TCP, which fails since there's
-        # no local MySQL socket -- the container is reached over the
-        # network. Only rewrite that specific local case to 127.0.0.1; a
-        # remote Docker daemon reports its own real host/IP here, which
-        # must be preserved so the suite can still reach it.
-        host = container.get_container_host_ip()
-        if host == "localhost":
-            host = "127.0.0.1"
-        port = container.get_exposed_port(container.port)
-        yield create_engine(
-            f"mysql://{container.username}:{container.password}"
-            f"@{host}:{port}/{container.dbname}"
-        )
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    MySQLEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = MySQLEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = MySQLEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_oceanbase.py

@@ -1,221 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.oceanbase against a real OceanBase instance, spun up
-on demand via testcontainers. Run via .github/workflows/testcontainers.yml,
-on the nightly cron / manual dispatch only (see `nightly_only: true` on
-this dialect's matrix entry) -- OceanBase bootstraps a distributed-style
-cluster even in single-node MODE=MINI, a substantially heavier first-boot
-than a single-process database, not a good fit for every PR's CI budget.
-
-OceanBaseEngineSpec extends MySQLEngineSpec and its dialect
-(oceanbase_py.sqlalchemy.dialect.OceanBaseDialect) extends
-MySQLDialect_mysqldb directly with no custom DDL or LIMIT/OFFSET compiler,
-so this follows the same mysqlclient-based pattern as MariaDB/MySQL/
-StarRocks in this suite -- including the same "localhost" -> "127.0.0.1"
-fix MySQLdb needs on native Linux Docker.
-
-Could not be verified locally in this environment: mysqlclient (MySQLdb)
-has a pre-existing, unrelated native-library linking issue against this
-machine's Homebrew-installed libmysqlclient, and this dialect wasn't
-pulled/run locally at all given its heavier resource footprint -- CI-only
-verification, matching the nightly_only gating.
-
-oceanbase_py.sqlalchemy.dialect.OceanBaseDialect has real bugs, all
-confirmed on real CI, in both has_table() (called by create_all()'s
-default checkfirst=True) and get_columns() (called by
-OceanBaseEngineSpec.get_columns(), which this suite's second test needs
-to actually exercise):
-1. Both pass a raw string straight to Connection.execute() (e.g.
-   `connection.execute(f"DESCRIBE {full_name}")`), which SQLAlchemy 2.0
-   rejects outright (ObjectNotExecutableError). Every *other* raw-SQL
-   method in the same dialect module correctly uses
-   `connection.exec_driver_sql(...)` instead.
-2. has_table() never catches the error DESCRIBE raises for a table that
-   doesn't exist (1146) -- so even with (1) fixed, it can only ever
-   return True, raising instead of returning False for exactly the case
-   checkfirst exists to handle.
-This test monkeypatches both methods to do what the rest of the dialect's
-raw-SQL methods already do, plus has_table()'s missing not-found
-handling, rather than working around any of this from the test side.
-"""
-
-from collections.abc import Iterator
-from typing import Any
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Connection, Engine, URL
-from sqlalchemy.exc import NoSuchTableError, ProgrammingError
-
-from superset.db_engine_specs.oceanbase import OceanBaseEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("oceanbase_py")
-
-from oceanbase_py.sqlalchemy import datatype  # noqa: E402
-from oceanbase_py.sqlalchemy.dialect import OceanBaseDialect  # noqa: E402
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import LogMessageWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-def _has_table(
-    self: OceanBaseDialect,
-    connection: Connection,
-    table_name: str,
-    schema: str | None = None,
-    **kw: object,
-) -> bool:
-    if schema is None:
-        schema = self.default_schema_name
-    quote = self.identifier_preparer.quote_identifier
-    full_name = quote(table_name)
-    if schema:
-        full_name = f"{quote(schema)}.{full_name}"
-    try:
-        res = connection.exec_driver_sql(f"DESCRIBE {full_name}")
-    except ProgrammingError:
-        # The original never catches this at all -- DESCRIBE on a
-        # nonexistent table raises 1146 ("table doesn't exist") rather
-        # than returning an empty result set, so the unpatched method can
-        # only ever return True, and raises instead of returning False for
-        # exactly the case create_all()'s checkfirst exists to handle
-        # (confirmed on real CI).
-        return False
-    return res.first() is not None
-
-
-def _get_columns(
-    self: OceanBaseDialect,
-    connection: Connection,
-    table_name: str,
-    schema: str | None = None,
-    **kw: object,
-) -> list[dict[str, Any]]:
-    # Same connection.execute(raw string) anti-pattern as has_table(),
-    # confirmed on real CI as the same ObjectNotExecutableError -- this is
-    # the actual column-introspection call OceanBaseEngineSpec.get_columns
-    # (and this test) needs, so it gets the same exec_driver_sql fix.
-    if not self.has_table(connection, table_name, schema):
-        raise NoSuchTableError(f"schema={schema}, table={table_name}")
-    schema = schema or self._get_default_schema_name(connection)
-
-    quote = self.identifier_preparer.quote_identifier
-    full_name = quote(table_name)
-    if schema:
-        full_name = f"{quote(schema)}.{full_name}"
-
-    res = connection.exec_driver_sql(f"SHOW COLUMNS FROM {full_name}")
-    return [
-        {
-            "name": record.Field,
-            "type": datatype.parse_sql_type(record.Type),
-            "nullable": record.Null == "YES",
-            "default": record.Default,
-        }
-        for record in res
-    ]
-
-
-OceanBaseDialect.has_table = _has_table
-OceanBaseDialect.get_columns = _get_columns
-
-PORT = 2881
-PASSWORD = "pilot"  # noqa: S105 -- fixed test-fixture password, not a secret
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("oceanbase/oceanbase-ce")
-    container.with_exposed_ports(PORT)
-    container.with_env("MODE", "MINI")
-    container.with_env("OB_TENANT_PASSWORD", PASSWORD)
-    container.waiting_for(LogMessageWaitStrategy("boot success!"))
-
-    with container:
-        host = container.get_container_host_ip()
-        if host == "localhost":
-            host = "127.0.0.1"
-        port = container.get_exposed_port(PORT)
-        # OceanBase usernames for a MySQL-mode tenant use "user@tenant"
-        # (e.g. "root@test"), a literal "@" that URL.create() percent-encodes
-        # correctly -- an f-string would produce a second "@" that breaks
-        # the URL's own host/user boundary parsing.
-        yield create_engine(
-            URL.create(
-                "oceanbase",
-                username="root@test",
-                password=PASSWORD,
-                host=host,
-                port=int(port),
-                database="test",
-            )
-        )
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    OceanBaseEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = OceanBaseEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = OceanBaseEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_oracle.py

@@ -1,95 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.oracle against a real Oracle instance (gvenzl/oracle-free),
-spun up on demand via testcontainers. Run via
-.github/workflows/testcontainers.yml.
-
-gvenzl/oracle-free ships with its datafiles pre-baked into the image, so
-once the (large-ish, ~1GB) image is pulled, container startup is fast --
-under 15s measured locally. Almost all the wall-clock cost here is the
-image pull itself, same as any other dialect's container.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.oracle import OracleEngineSpec
-from superset.sql.parse import Table
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.oracle")
-
-from testcontainers.community.oracle import OracleDbContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with OracleDbContainer() as container:
-        yield create_engine(container.get_connection_url())
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    OracleEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = OracleEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = OracleEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None

tests/testcontainers/db_engine_specs/test_postgres.py

@@ -1,99 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.postgres against a real PostgreSQL instance, spun up
-on demand via testcontainers. Run via .github/workflows/testcontainers.yml
--- these exercise real SQL execution and dialect introspection, which
-mocked unit tests structurally cannot.
-
-Plain Postgres itself was never covered by this suite: CockroachDB,
-TimescaleDB and YugabyteDB all speak the Postgres wire protocol and already
-exercise `PostgresContainer`/the "postgresql" dialect, but none of them
-stand in for vanilla PostgreSQL's own dialect quirks.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.postgres import PostgresEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.postgres")
-
-from testcontainers.community.postgres import PostgresContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with PostgresContainer("postgres:17-alpine") as container:
-        yield create_engine(container.get_connection_url())
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    PostgresEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = PostgresEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = PostgresEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_risingwave.py

@@ -1,127 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.risingwave against a real RisingWave instance, spun
-up on demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-RisingWave speaks the Postgres wire protocol, but doesn't run the real
-Postgres server binary or its POSTGRES_PASSWORD-style bootstrap env vars,
-so this can't reuse `PostgresContainer` the way TimescaleDB/YugabyteDB do
--- it needs a generic DockerContainer against the official
-`risingwavelabs/risingwave` single-binary playground image instead.
-`RisingWaveDbEngineSpec` extends `PostgresEngineSpec`, and
-`sqlalchemy-risingwave`'s dialect is a genuine subclass of SQLAlchemy's own
-Postgres dialect (via psycopg2), so DDL/pagination compile with standard
-Postgres semantics -- no ClickHouse-style mandatory table option needed.
-"""
-
-import re
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-    text,
-)
-from sqlalchemy.engine import Connection, Engine
-
-from superset.db_engine_specs.risingwave import RisingWaveDbEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("sqlalchemy_risingwave")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import LogMessageWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-PORT = 4566
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("risingwavelabs/risingwave")
-    container.with_exposed_ports(PORT)
-    container.with_command("playground")
-    # The actual startup banner reads "RisingWave standalone mode is
-    # ready." -- confirmed against a real container's logs.
-    container.waiting_for(
-        LogMessageWaitStrategy(re.compile("RisingWave standalone mode is ready"))
-    )
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(PORT)
-        yield create_engine(f"risingwave://root@{host}:{port}/dev")
-
-
-def _flush(conn: Connection) -> None:
-    # RisingWave's storage engine checkpoints asynchronously: without an
-    # explicit FLUSH, a SELECT immediately after INSERT can see zero rows
-    # -- confirmed against a real instance (a bare INSERT commits fine, but
-    # the data isn't visible to a subsequent query until flushed).
-    conn.execute(text("FLUSH"))
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine, after_insert=_flush)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    RisingWaveDbEngineSpec.get_columns wraps a real SQLAlchemy Inspector;
-    this exercises that against actual server-reported column metadata
-    rather than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = RisingWaveDbEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = RisingWaveDbEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_starrocks.py

@@ -1,156 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.starrocks against a real StarRocks instance, spun up
-on demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-StarRocks has no dedicated testcontainers module, so this uses a generic
-DockerContainer against the official `starrocks/allin1-ubuntu` image, which
-brings up both the FE (query frontend, MySQL wire protocol on port 9030)
-and BE (execution backend) in a single container -- a heavier bring-up than
-a single-process database. `root` has no password by default and no
-database exists yet, so the fixture creates one itself before yielding an
-engine pointed at it. The FE's query port accepts connections, and can even
-run metadata statements like CREATE DATABASE, before the BE has registered
-with it -- an actual CREATE TABLE/INSERT then fails with "Backend node not
-found" -- so the fixture retries a real create-table-and-insert probe
-against a throwaway table rather than trusting the open port or a bare
-CREATE DATABASE as a readiness signal.
-
-Not verified locally in this environment: the `allin1-ubuntu` image is
-multiple GB and was skipped here to keep local Docker resource usage low,
-per session guidance to lean on CI (which has no such constraint) for
-dialects with unusually heavy images.
-"""
-
-import time
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-    text,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.starrocks import StarRocksEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("starrocks")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import PortWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-QUERY_PORT = 9030
-DBNAME = "pilot"
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("starrocks/allin1-ubuntu")
-    container.with_exposed_ports(QUERY_PORT)
-    container.waiting_for(PortWaitStrategy(QUERY_PORT))
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(QUERY_PORT)
-        bootstrap_engine = create_engine(
-            f"starrocks://root:@{host}:{port}/default_catalog.information_schema"
-        )
-
-        # The FE's query port accepts connections, and can even run metadata
-        # statements like CREATE DATABASE, before any BE (execution backend)
-        # has registered with it -- an actual table create/insert then fails
-        # with "Backend node not found". Probe with the real operations the
-        # tests below need, in a throwaway table, so readiness is confirmed
-        # for what actually matters rather than just the FE's own port.
-        last_error: Exception | None = None
-        for _ in range(60):
-            try:
-                with bootstrap_engine.begin() as conn:
-                    conn.execute(text(f"CREATE DATABASE IF NOT EXISTS {DBNAME}"))
-                probe_engine = create_engine(
-                    f"starrocks://root:@{host}:{port}/default_catalog.{DBNAME}"
-                )
-                with probe_engine.begin() as conn:
-                    conn.execute(
-                        text("CREATE TABLE IF NOT EXISTS pilot_ready (id INT)")
-                    )
-                    conn.execute(text("INSERT INTO pilot_ready VALUES (1)"))
-                    conn.execute(text("DROP TABLE pilot_ready"))
-                break
-            except Exception as ex:  # noqa: BLE001 -- retry on any not-ready-yet error
-                last_error = ex
-                time.sleep(2)
-        else:
-            raise RuntimeError(
-                "StarRocks FE/BE never became ready to create and use a table"
-            ) from last_error
-
-        yield create_engine(f"starrocks://root:@{host}:{port}/default_catalog.{DBNAME}")
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    StarRocksEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = StarRocksEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = StarRocksEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_timescaledb.py

@@ -1,98 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.timescaledb against a real TimescaleDB instance,
-spun up on demand via testcontainers. Run via
-.github/workflows/testcontainers.yml -- these exercise real SQL execution
-and dialect introspection, which mocked unit tests structurally cannot.
-
-TimescaleDB is a genuine Postgres extension, not a fork: connects via the
-plain "postgresql" dialect with psycopg2, same as vanilla Postgres, just
-pointed at the timescale/timescaledb image instead of postgres:latest.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.timescaledb import TimescaleDBEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.postgres")
-
-from testcontainers.community.postgres import PostgresContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with PostgresContainer("timescale/timescaledb:2.29.2-pg16") as container:
-        yield create_engine(container.get_connection_url())
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    TimescaleDBEngineSpec.get_columns wraps a real SQLAlchemy Inspector;
-    this exercises that against actual server-reported column metadata
-    rather than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = TimescaleDBEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = TimescaleDBEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_trino.py

@@ -1,104 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.trino against a real Trino instance, spun up on
-demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-Only Presto is covered by existing docker-compose-based integration CI;
-Trino, despite sharing lineage with Presto, is not.
-"""
-
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.trino import TrinoEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.community.trino")
-
-from testcontainers.community.trino import TrinoContainer  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    with TrinoContainer() as container:
-        # TrinoContainer.get_connection_url() (testcontainers 4.15.0) returns
-        # the container-internal port (e.g. 8080) instead of the Docker-
-        # mapped host port, so the URL it builds cannot actually connect.
-        # Build it manually with get_exposed_port() instead. Filed upstream:
-        # https://github.com/testcontainers/testcontainers-python/issues
-        url = (
-            f"trino://{container.user}@{container.get_container_host_ip()}"
-            f":{container.get_exposed_port(container.port)}/memory/default"
-        )
-        yield create_engine(url)
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. This is the exact bug class in apache/superset#42899,
-    where Trino emitted OFFSET before LIMIT for paginated queries -- a
-    dialect-compiler bug invisible to mocked tests, only catchable by
-    actually executing the compiled SQL.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    TrinoEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = TrinoEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = TrinoEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_ydb.py

@@ -1,148 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.ydb against a real YDB instance, spun up on demand
-via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-YDB has no dedicated testcontainers module, so this uses a generic
-DockerContainer against the official `ydbplatform/local-ydb` image, which
-needs no auth for local/anonymous access -- YDBEngineSpec's own
-`sqlalchemy_uri_placeholder` ("ydb://{host}:{port}/{database_name}") has
-no username/password at all.
-"""
-
-import time
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.ydb import YDBEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-require_driver("ydb_sqlalchemy")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import PortWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-GRPC_PORT = 2136
-DATABASE = "/local"
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("ydbplatform/local-ydb")
-    container.with_exposed_ports(GRPC_PORT)
-    # YDB's gRPC client does endpoint discovery: it asks the server for its
-    # "real" endpoints and reconnects to whatever comes back, rather than
-    # just using the address it was originally given. By default that's
-    # the container's own internal Docker hostname (e.g. "6abbb4bb0ab7"),
-    # which isn't reachable from the host. Binding the same port number on
-    # the host as inside the container, plus advertising "localhost" as
-    # the container's own hostname, makes the discovered endpoint
-    # ("localhost:2136") actually resolve to something reachable.
-    container.with_bind_ports(GRPC_PORT, GRPC_PORT)
-    container.with_kwargs(hostname="localhost")
-    container.with_env("YDB_USE_IN_MEMORY_PDISKS", "true")
-    container.waiting_for(PortWaitStrategy(GRPC_PORT))
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(GRPC_PORT)
-        eng = create_engine(f"yql://{host}:{port}{DATABASE}")
-
-        # The gRPC port opens, and even a bare SELECT succeeds, before YDB's
-        # storage pools are fully initialized -- an actual CREATE TABLE can
-        # still fail with "database doesn't have storage pools at all to
-        # create tablet channels" (confirmed on a real instance). Probe with
-        # metadata.create_all()/drop_all() specifically, the same call the
-        # real tests below make: a raw `text("CREATE TABLE ...")` hits a
-        # separate, unrelated error ("Scheme operations cannot be executed
-        # inside transaction") that create_all()'s own DDL execution path
-        # doesn't, even with AUTOCOMMIT set on a manually-opened connection.
-        probe_metadata = MetaData()
-        SATable("pilot_ready", probe_metadata, Column("id", Integer, primary_key=True))
-        last_error: Exception | None = None
-        for _ in range(30):
-            try:
-                probe_metadata.create_all(eng)
-                probe_metadata.drop_all(eng)
-                break
-            except Exception as ex:  # noqa: BLE001 -- retry on any not-ready-yet error
-                last_error = ex
-                time.sleep(2)
-        else:
-            raise RuntimeError(
-                "YDB never became ready to create and use a table"
-            ) from last_error
-
-        yield eng
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    YDBEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = YDBEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = YDBEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/testcontainers/db_engine_specs/test_yugabytedb.py

@@ -1,119 +0,0 @@
-# Licensed to the Apache Software Foundation (ASF) under one
-# or more contributor license agreements.  See the NOTICE file
-# distributed with this work for additional information
-# regarding copyright ownership.  The ASF licenses this file
-# to you under the Apache License, Version 2.0 (the
-# "License"); you may not use this file except in compliance
-# with the License.  You may obtain a copy of the License at
-#
-#   http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing,
-# software distributed under the License is distributed on an
-# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
-# KIND, either express or implied.  See the License for the
-# specific language governing permissions and limitations
-# under the License.
-"""
-Tests db_engine_specs.yugabytedb against a real YugabyteDB instance, spun
-up on demand via testcontainers. Run via .github/workflows/testcontainers.yml.
-
-YugabyteDB's YSQL layer is fully Postgres-wire compatible (postgresql+
-psycopg2, port 5433), but the image doesn't ship a plain `psql` binary
-(only its own `ysqlsh`), so testcontainers' PostgresContainer can't be
-reused directly -- its built-in readiness check execs `psql`, which would
-fail here. This uses a generic DockerContainer, starting the node via
-`yugabyted start --background=false` and waiting for yugabyted's own final
-startup message instead.
-"""
-
-import re
-from collections.abc import Iterator
-
-import pytest
-from sqlalchemy import (
-    Column,
-    create_engine,
-    inspect,
-    Integer,
-    MetaData,
-    Table as SATable,
-)
-from sqlalchemy.engine import Engine
-
-from superset.db_engine_specs.yugabytedb import YugabyteDBEngineSpec
-from superset.sql.parse import Table
-from superset.utils.core import GenericDataType
-
-pytestmark = pytest.mark.testcontainers
-
-from ._driver import require_driver  # noqa: E402
-
-require_driver("testcontainers.core.container")
-
-from testcontainers.core.container import DockerContainer  # noqa: E402
-from testcontainers.core.wait_strategies import LogMessageWaitStrategy  # noqa: E402
-
-from ._pagination import (  # noqa: E402
-    assert_paginated_query_returns_correct_rows_in_order,
-)
-
-YSQL_PORT = 5433
-
-
-@pytest.fixture(scope="module")
-def engine() -> Iterator[Engine]:
-    container = DockerContainer("yugabytedb/yugabyte:latest")
-    container.with_exposed_ports(YSQL_PORT)
-    container.with_command("bin/yugabyted start --background=false")
-    container.waiting_for(
-        LogMessageWaitStrategy(
-            re.compile("Data placement constraint successfully verified")
-        )
-    )
-
-    with container:
-        host = container.get_container_host_ip()
-        port = container.get_exposed_port(YSQL_PORT)
-        # Default single-node credentials/database, per yugabyted's own
-        # documented quickstart defaults -- no env vars needed to set them.
-        yield create_engine(
-            f"postgresql+psycopg2://yugabyte:yugabyte@{host}:{port}/yugabyte"
-        )
-
-
-def test_paginated_query_returns_correct_rows_in_order(engine: Engine) -> None:
-    """
-    A plain SQLAlchemy Core LIMIT/OFFSET query, compiled and executed against
-    a real instance. Mocked tests cannot catch a dialect compiling this
-    incorrectly (see apache/superset#42899, where Trino emitted OFFSET
-    before LIMIT) -- only real execution can.
-    """
-    assert_paginated_query_returns_correct_rows_in_order(engine)
-
-
-def test_get_columns_maps_native_types(engine: Engine) -> None:
-    """
-    YugabyteDBEngineSpec.get_columns wraps a real SQLAlchemy Inspector; this
-    exercises that against actual server-reported column metadata rather
-    than a mocked Inspector.
-    """
-    metadata = MetaData()
-    SATable(
-        "pilot_types",
-        metadata,
-        Column("id", Integer, primary_key=True),
-        Column("amount", Integer),
-    )
-    metadata.create_all(engine)
-
-    inspector = inspect(engine)
-    columns = YugabyteDBEngineSpec.get_columns(inspector, Table("pilot_types"))
-
-    by_name = {col["column_name"]: col for col in columns}
-    assert set(by_name) == {"id", "amount"}
-    for col in by_name.values():
-        spec = YugabyteDBEngineSpec.get_column_spec(str(col["type"]))
-        assert spec is not None
-        assert spec.generic_type == GenericDataType.NUMERIC
-        assert isinstance(spec.sqla_type, Integer)

tests/unit_tests/db_engine_specs/test_init.py

@@ -137,56 +137,6 @@ def test_get_available_engine_specs_keeps_valid_third_party_dialect(
     assert available[SqliteEngineSpec] == {"valid_driver"}
 
 
-def test_get_available_engine_specs_restores_compiler_operators(
-    mocker: MockerFixture,
-) -> None:
-    """
-    A third-party ``sqlalchemy.dialects`` entry point that mutates SQLAlchemy's
-    shared, process-global ``compiler.OPERATORS`` mapping on import (as
-    ``sqlalchemy-monetdb`` does, in place, rather than subclassing) must not be
-    allowed to leak that change into every other dialect for the rest of the
-    process.
-
-    Regression test: enumerating a real "monetdb" entry point here (to build the
-    "available databases" list) silently changed ``!=`` rendering to ``<>`` for
-    postgres/mysql/sqlite/etc. too, for the remainder of the process.
-    """
-    from sqlalchemy.sql import compiler as sqla_compiler, operators
-
-    mocker.patch(
-        "superset.db_engine_specs.load_engine_specs",
-        return_value=iter([]),
-    )
-
-    pristine = dict(sqla_compiler.OPERATORS)
-    assert pristine[operators.ne] != " <> "
-
-    class MisbehavingDialect(DefaultDialect):
-        name = "misbehaving"
-        driver = "misbehaving_driver"
-
-    def load_and_mutate_globally() -> type[MisbehavingDialect]:
-        # Mirrors sqlalchemy-monetdb's `base.py`: grabs a reference to the
-        # shared dict (not a copy) and mutates it in place.
-        sqla_compiler.OPERATORS[operators.ne] = " <> "
-        return MisbehavingDialect
-
-    entry_point = mocker.MagicMock()
-    entry_point.name = "misbehaving"
-    entry_point.load.side_effect = load_and_mutate_globally
-    mocker.patch(
-        "superset.db_engine_specs.entry_points",
-        return_value=[entry_point],
-    )
-
-    try:
-        get_available_engine_specs()
-        assert sqla_compiler.OPERATORS[operators.ne] == pristine[operators.ne]
-    finally:
-        sqla_compiler.OPERATORS.clear()
-        sqla_compiler.OPERATORS.update(pristine)
-
-
 @pytest.mark.parametrize(
     "app",
     [{"DBS_AVAILABLE_DENYLIST": {"databricks": {"pyhive", "pyodbc"}}}],