v1.2.3

What's Changed

Fixes

• Fix SQL identifier injection in table summary queries — column and table names are now properly quoted
• Fix error handling: RowsAffected, CSV write errors, queryOutputFile.Close(), Float64Time null JSON
• Propagate parent context in timeout wrappers instead of using context.Background()
• Add HTTP client timeout (30s) for Pulumi API calls
• Fix unused local variable (#82)
• Fix explicit returns mixed with implicit (fall through) returns (#81)
• Fix delete_run query

Improvements

• Add pre-commit hook for fmt, vet, staticcheck, tests, and coverage
• Exclude benchmarks from release tar

Dependencies

• Update presto-go-client to v2.1.1 (renamed query_json → queryjson)
• Bump filippo.io/edwards25519, golang.org/x/crypto, and other transitive deps

Performance

• Add new version of stream run for 10TB from perf team (#78)

1.2.2

Switch to prestodb/presto-go-client/v2 (v2.0.0). Bump Go to 1.26.1. Bump GitHub Actions to Node.js 24 compatible versions.

1.2.1

genconfig Improvements

• Generalized to use generic maps instead of hardcoded Go structs — adding new config fields no longer requires Go code changes, just JSON and template updates
• .genconfigignore support — directories containing this marker file are skipped during config generation, allowing manually maintained configs to coexist with auto-generated ones
• Skip empty template outputs — templates that render no output no longer create empty files
• New template functions: mul, add, sub, div, min, max, floor, ceil, set, default, seq, hasSuffix, hasPrefix, contains
• Repeatable -p flag for stacking multiple parameter files (later overrides earlier)
• Computation logic moved from Go into .prelude templates using set(), making config generation fully data-driven
• Spark template improvements: conditional generate_spark_config flag, SSD local dirs, GPU support

Bug Fixes

• Disable CGO for cross-compiled builds to produce static binaries that run on older Linux distros (e.g., Debian Bullseye with glibc 2.31)

1.2

New Features

• Parallel stream execution — stream_count parameter runs multiple parallel instances of a stage, each with its own random seed for deterministic replay
• Directory expansion in query_files — entries pointing to directories are expanded to their contained files at execution time (after pre_stage_scripts), enabling dynamic query generation workflows (e.g., dsqgen)
• no_random_duplicates — shuffled random execution that cycles through all queries before repeating any
• Shell script environment variables — PBENCH_STAGE_ID, PBENCH_OUTPUT_DIR, PBENCH_QUERY_FILE, PBENCH_QUERY_ID, PBENCH_QUERY_ERROR, etc. are injected into all script hooks
• pre_query_scripts — run scripts before each individual query execution
• Auto-detect catalog/schema changes — child stages with different catalog/schema/timezone automatically create a new client
• Custom stage IDs — id field in stage JSON
• Generalized genddl — supports TPC-H and custom workloads, not just TPC-DS; paths resolved relative to config file
• Glue catalog benchmark configs for Java/native engines

Bug Fixes

• Fix data race on RandSeedUsed in concurrent stream goroutines
• Fix post_query_cycle_scripts not running on abort
• Fix concurrency bugs in stage package (multiple race conditions)
• Fix SQL parameter mismatch in loadjson ORM
• Fix cold/warm run value initialization and inheritance
• Fix query forward tool synchronization issues
• Fix x/net security vulnerability

Improvements

• Migrated from internal presto package to presto-go library
• CI pipeline with tests, coverage (50%+ threshold), linting, vulnerability scanning, and semantic commit checks
• GitHub Actions release workflow with approval gate
• Comprehensive test coverage for genconfig, queryplan, round, genddl
• Staticcheck lint cleanup across codebase

1.1

November 2024

v1.0

Update README.md