Changelog

All notable changes to this project will be documented in this file.

[1.2.15] - 2026-05-20

Build

• Drop 32-bit linux/arm binary. Apache Thrift v0.23.0 (required to fix CVE-2026-41602) uses math.MaxUint32 as an untyped int constant in framed_transport.go, which overflows on 32-bit architectures. 32-bit ARM is rarely used to run Grafana (RPi 4+ is arm64), so the artifact is dropped rather than staying on the vulnerable Thrift version. A custom BuildAll target in Magefile.go builds the remaining five platforms (linux amd64/arm64, darwin amd64/arm64, windows amd64).
• Refresh grafana-plugin-sdk-go to v0.292.0 to clear the validator's "SDK older than 2 months" warning.

[1.2.13] - 2026-05-20

Bug Fixes

• S3 list pagination: listPrefixes and listFiles now page through ListObjectsV2 results so buckets with more than 1000 objects are not silently truncated.

Improvements

• Schema discovery without full read: SELECT * FROM parquet LIMIT 0 (used by the frontend to fetch column names) now reads only the Parquet footer via arrowReader.Schema() instead of downloading the entire file from S3.
• Renamed pkg/duckdb → pkg/sqlexec: the package never used DuckDB; the new name reflects what it actually is — a small in-memory SQL executor.
• Plugin README: replaced the @grafana/create-plugin template src/README.md with proper, screenshot-rich documentation that is bundled into the published plugin.

[1.2.0] - 2026-02-14

Bug Fixes

• Fix LARGE_STRING Parquet panic: Pandas-generated Parquet files use LARGE_STRING Arrow type, which caused a panic in the Parquet reader. Separated arrow.LARGE_STRING and arrow.STRING handling with proper type assertions.
• Fix GROUP BY not aggregating: GROUP BY queries returned one row per record instead of grouped results. The SQL executor was using pointer addresses as group keys (*int64, *string). Added formatValue() helper to dereference pointers before grouping.
• Fix Titanic dashboard bar charts: Grafana barchart panels require a string x-axis but Pclass is int64. Changed to table panels with gauge cell rendering.

Improvements

• Added Playwright E2E tests verifying all 4 provisioned dashboards load with correct data
• Added Docker Compose dev environment (docker/docker-compose.yml) with MinIO and sample data generation
• Updated all provisioned dashboard JSON files for Grafana compatibility
• Added CLAUDE.md project documentation

[1.1.0] - 2026-02-02

Features

• SQL Query Support: Built-in, in-memory SQL engine (a custom lightweight executor, not DuckDB)
  • SELECT, WHERE, GROUP BY, ORDER BY, LIMIT clauses
  • Aggregation functions (COUNT, SUM, AVG, MIN, MAX)
  • Filtering and sorting
• Visual Query Builder: PostgreSQL-style query builder interface
  • Column selection with aggregations
  • Filter toggle with condition builder
  • Group By toggle with multi-column selection
  • Order By toggle with ASC/DESC
  • SQL Preview panel showing generated query
• Template Variables: Support for dashboard template variables
  • List files in bucket (with regex filtering)
  • List prefixes/folders
  • SQL-based variable queries
• Explore View: Enhanced query editor for Grafana Explore
  • File selector with regex filtering
  • Refresh button to reload file list
  • Builder and Code modes
• Sample Dashboards: Pre-built dashboards demonstrating plugin capabilities
  • Iris Dataset dashboard
  • Titanic Survival Dataset dashboard
  • Time Series Metrics dashboard

Improvements

• Reduced Grafana version requirement to 11.0.0+
• Better error handling and logging
• Improved path-style routing for S3-compatible storage
• ARM64 binary support for Apple Silicon

[1.0.0] - 2025-12-22

Features

• Initial Release: Parquet-S3-Datasource plugin for Grafana
• S3 Connectivity: Support for Amazon S3 and S3-compatible storage providers:
  • Amazon S3
  • MinIO
  • Wasabi
  • DigitalOcean Spaces
  • Any S3-compatible storage
• Direct Parquet Querying: Read Apache Parquet files directly from S3 without intermediate databases
• Apache Arrow Integration: Efficient columnar data processing using Apache Arrow
• Custom Endpoints: Configurable S3 endpoints for private cloud deployments
• Path-Style Routing: Automatic configuration for storage systems requiring path-style URLs
• Data Types Support: All Parquet primitive types, nested structures, and compression codecs
• Grafana 11.6+: Fully compatible with Grafana 11.6.0 and above

Components

• React-based frontend with TypeScript
• Go backend utilizing AWS SDK v2
• Docker Compose setup for local development with MinIO
• Comprehensive documentation and configuration examples
• Demo video and screenshots

Configuration

• Region selection
• Bucket specification
• Custom endpoint URLs
• Secure credential management (Access Key/Secret Key)

Development

• Provisioned datasource for quick Docker setup
• Sample Parquet file generation script
• Minio upload utilities
• End-to-end testing capabilities