Changelog

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

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

Fixed issues where fields in structs that were not defined in the schema raised errors (array, slice, time, map)
Added support for reflect encoding/decoding of additional types to/from parquet types
- Go type <-> Parquet type
- int <-> int64
- int32 <-> int64
- int16 <-> int64
- int8 <-> int64
- uint <-> int64
- uint16 <-> int64
- uint8 <-> int64
- int64 <-> int32
- uint64 <-> int32
- uint32 <-> int32
Removed some inconsistent api behaviors
- reflect marshaling/unmarshalling now ignores fields not defined in the schema (this already happens when the marshaller is used by floor.Writer and goparquet.FileReader)
- int32PlainDecoder and int32PlainEncoder now no longer support uint values
Added function IsAfterUnixEpoch to allow test whether a timestamp can be written as Julian date.
Allowed setting selected columns after opening a parquet file through SetSelectedColumns method.
Fixed dictStore to correctly reset valueSize which should reduce size of written files.
Fixed reflection-based marshal/unmarshaling for some basic types.
Fixed DECIMALs computation of maximum amount of digits.
Improved parsing of legacy timestamps.
Allowed reading of file metadata before properly opening a parquet file.
Added method SeekToRowGroup to allow seeking to specific row groups.

Added examples how to use the low-level and high-level APIs.
Added backwards compability in floor reader and writer for lists as produced by Athena.
Fixed many crash issues found during fuzzing of the parquet reader.

Added csv2parquet command to convert CSV files into parquet files
Fixed issue in parquet-tool cat that wouldn't print any records if no limit was provided
Added support for backward compatible MAPs and LISTs in schema validation
Added ValidateStrict method to validate without regard for backwards compatibility