Skip to content

Development and Testing ​

This page is for source and extension developers. See Service Deployment to deploy a service and Client Connections for everyday SQL use. Run commands from the repository root.

Build and Run Locally ​

Build the service and required modules:

bash
mvn package -pl sqlrec-frontend -am -DskipTests
SQL_SCHEMA_DIR="$PWD/sqlrec-demo/src/main/sql" bash bin/cli.sh

SQL-file mode does not require HMS or Minikube. Queries still need the relevant Connector and UDF JARs. CLI logs default to sqlrec-frontend/target/logs.

For remote metadata, configure HIVE_METASTORE_URI, META_DB_URL, and database credentials. Without configured endpoints, the source CLI tries development defaults from DEFAULT_TEST_IP, NODE_IP, or Minikube discovery. Explicit settings, including empty strings, are preserved.

Test an Existing Environment ​

bash
bash bin/test.sh -pl sqlrec-core -am

The script runs Maven integration tests and passes other arguments to Maven unchanged. Addresses prefer DEFAULT_TEST_IP, then NODE_IP, then Minikube discovery. Override individual component endpoints or HADOOP_CONF_DIR as needed. Maven uses the current shell's JDK; Hive/Hadoop clients use their configured Java environment.

Rebuild and Run All Tests ​

bash
bash bin/test.sh -all

--all and -all are equivalent. This mode needs an accessible existing Minikube environment to save an image cache, then deletes and rebuilds the cluster named minikube, including its node-local dynamic volume data. Host-side image caches and client directories remain; failed image caching prevents cluster deletion.

Stages include dependency deployment, readiness checks, Maven integration tests, and SQL performance, GrowthBook, Hugging Face, and model quality benchmarks. Any failed stage stops the run. The script uses configured release images without building local source images; build and select your images before testing source changes. The run requires network access, deployment permissions, and sufficient compute and storage.

Read Reports ​

The full run writes its report and stage logs under benchmark/runs/ and prints the run directory on completion. Failed runs also produce reports. Override the parent directory with TEST_REPORT_ROOT.

  • summary.md: stage status, test results, and benchmark results.
  • logs/: command output for each stage.
  • growthbook/, huggingface/, model_quality/: detailed benchmark artifacts.

See individual benchmark READMEs for standalone options:

Preview and Check Documentation ​

bash
npm ci
npm run docs:dev
npm run docs:build

Documentation sources are in docs/docs/ and docs/en/docs/. When moving pages, update both languages and docs/.vitepress/config.mts. Do not edit generated .vitepress/dist or cache files.