Windows test

Author: satvshr

.github/workflows/test.yml

@@ -74,7 +74,15 @@ jobs:
       with:
         python-version: ${{ matrix.python-version }}
 
-    - name: Install test dependencies and scikit-learn
+    - name: Checkout server-api and patch Docker path
+          if: runner.os == 'Linux'
+          shell: bash
+          run: |
+            git clone --depth 1 https://github.com/openml/server-api.git server-api
+            sed -i 's|\.\./server-api|./server-api|g' docker-compose.yml
+
+    - name: Install test dependencies, scikit-learn, and optional pandas
+      shell: bash
       run: |
         python -m pip install --upgrade pip
         pip install -e .[test] scikit-learn==${{ matrix.scikit-learn }}

docker-compose.yml

@@ -0,0 +1,53 @@
+services:
+  database:
+    image: "openml/test-database:20240105"
+    container_name: "openml-test-db-ci"
+    environment:
+      MYSQL_ROOT_PASSWORD: ok
+    ports:
+      - "33060:3306"
+    healthcheck:
+      test: ["CMD", "mysqladmin" ,"ping", "-h", "localhost"]
+      start_period: 30s
+      interval: 5s
+      retries: 10
+
+  # SETUP WORKER
+  database-setup:
+    image: mysql
+    container_name: "openml-test-setup-ci"
+    volumes:
+      # You MUST save the update.sh content you shared earlier to this path
+      - ./docker/update.sh:/database-update.sh
+    command: /bin/sh -c "/database-update.sh"
+    depends_on:
+      database:
+        condition: service_healthy
+
+  php-api:
+    image: "openml/php-rest-api:v1.2.2"
+    container_name: "openml-php-api-ci"
+    ports:
+      - "9002:80"
+    depends_on:
+      database:
+        condition: service_started
+    environment:
+      - DB_HOST_OPENML=database:3306
+      - DB_HOST_EXPDB=database:3306
+      - BASE_URL=http://localhost:9002/
+      - INDEX_ES_DURING_STARTUP=false
+
+  # V2 API (PYTHON)
+  python-api:
+    container_name: "openml-python-api-ci"
+    build:
+          # TODO: replace with image when available
+          context: ../server-api
+          dockerfile: docker/python/Dockerfile
+    ports:
+      - "9001:8000"
+    depends_on:
+      - database
+    environment:
+      - DATABASE_URL=mysql://root:ok@database:3306/openml

docker/update.sh

@@ -0,0 +1,31 @@
+#/bin/bash
+# Change the filepath of openml.file
+# from "https://www.openml.org/data/download/1666876/phpFsFYVN"
+# to "http://minio:9000/datasets/0000/0001/phpFsFYVN"
+mysql -hdatabase -uroot -pok -e 'UPDATE openml.file SET filepath = CONCAT("http://minio:9000/datasets/0000/", LPAD(id, 4, "0"), "/", SUBSTRING_INDEX(filepath, "/", -1)) WHERE extension="arff";'
+
+# Update openml.expdb.dataset with the same url
+mysql -hdatabase -uroot -pok -e 'UPDATE openml_expdb.dataset DS, openml.file FL SET DS.url = FL.filepath WHERE DS.did = FL.id;'
+
+
+
+
+
+# Create the data_feature_description TABLE. TODO: can we make sure this table exists already?
+mysql -hdatabase -uroot -pok -Dopenml_expdb -e 'CREATE TABLE IF NOT EXISTS `data_feature_description` (
+  `did` int unsigned NOT NULL,
+  `index` int unsigned NOT NULL,
+  `uploader` mediumint unsigned NOT NULL,
+  `date` timestamp NOT NULL DEFAULT CURRENT_TIMESTAMP,
+  `description_type` enum("plain", "ontology") NOT NULL,
+  `value` varchar(256) NOT NULL,
+  KEY `did` (`did`,`index`),
+  CONSTRAINT `data_feature_description_ibfk_1` FOREIGN KEY (`did`, `index`) REFERENCES `data_feature` (`did`, `index`) ON DELETE CASCADE ON UPDATE CASCADE
+)'
+
+# SET dataset 1 to active (used in unittests java)
+mysql -hdatabase -uroot -pok -Dopenml_expdb -e 'INSERT IGNORE INTO dataset_status VALUES (1, "active", "2024-01-01 00:00:00", 1)'
+mysql -hdatabase -uroot -pok -Dopenml_expdb -e 'DELETE FROM dataset_status WHERE did = 2 AND status = "deactivated";'
+
+# Temporary fix in case the database missed the kaggle table. The PHP Rest API expects the table to be there, while indexing.
+mysql -hdatabase -uroot -pok -Dopenml_expdb -e 'CREATE TABLE IF NOT EXISTS `kaggle` (`dataset_id` int(11) DEFAULT NULL, `kaggle_link` varchar(500) DEFAULT NULL)'

pytest.ini

@@ -0,0 +1,4 @@
+[pytest]
+env =
+    OPENML_SERVER = http://localhost:9001/api/v2
+    OPENML_API_KEY = AD000000000000000000000000000000

tests/conftest.py

@@ -24,6 +24,7 @@
 from __future__ import annotations
 
 import multiprocessing
+import sys
 
 multiprocessing.set_start_method("spawn", force=True)
 
@@ -35,6 +36,9 @@
 import pytest
 import openml_sklearn
 
+import time
+import subprocess
+import requests
 import openml
 from openml.testing import TestBase
 
@@ -296,6 +300,44 @@ def with_test_cache(test_files_directory, request):
     if tmp_cache.exists():
         shutil.rmtree(tmp_cache)
 
+# This starts the entire stack once for the whole test run
+@pytest.fixture(scope="session", autouse=True)
+def openml_docker_stack():
+    # if sys.platform == "win32":
+    #         yield
+    #         return
+    # 1. Start the containers defined in your final docker-compose.yml
+    subprocess.run(["docker", "compose", "up", "-d"], check=True)
+    
+    # 2. Wait for the database setup worker to finish its tasks
+    # This ensures update.sh has finished before we hit the APIs
+    subprocess.run(["docker", "wait", "openml-test-setup-ci"], check=True)
+    
+    # 3. Quick health check: Wait for the Python API to respond on port 9001
+    timeout = 30
+    start = time.time()
+    while time.time() - start < timeout:
+        try:
+            if requests.get("http://localhost:9001/api/v2/").status_code == 200:
+                break
+        except requests.exceptions.ConnectionError:
+            time.sleep(1)
+            
+    yield # Tests run here
+    
+    # 4. Tear everything down after tests finish to keep the machine clean
+    subprocess.run(["docker", "compose", "down", "-v"], check=True)
+
+# This resets the database state before every single test to prevent race conditions
+@pytest.fixture(scope="function", autouse=True)
+def reset_db_state():
+    # if sys.platform == "win32":
+    #         yield
+    #         return
+    # Fast restart of the database container to return to the 'baked-in' state
+    subprocess.run(["docker", "compose", "restart", "database"], check=True)
+    # Re-run the setup worker to ensure paths are still correct
+    subprocess.run(["docker", "compose", "up", "database-setup"], check=True)
 
 @pytest.fixture
 def static_cache_dir():