[GO] parquet: dict spaced eof exception

[Sunyue]

Describe the bug, including details regarding any error messages, version, and platform.

Hi, I am reading parquet from S3 using the parquet reader, basically by providing a reader with ReadAt(buf []byte, off int64) implemented with Range header like bytes=<off>-<off+len(buf)-1>. When I enable BufferedStreamEnabled, when reading certain columns parquet reader either ends in error parquet: dict spaced eof exception or panic. But when I disable BufferedStreamEnabled reading same parquet file works fine.
I can't see where the error is, could you kindly help on this? Thanks!

Panic looks like

panic: snappy: corrupt input

goroutine 87202 [running]:
github.com/apache/arrow-go/v18/parquet/compress.snappyCodec.Decode(...)

Component(s)

Parquet

[Sunyue]

Reading from an os.File can also end in same error. Please find program to reproduce this issue below:

package main

import (
	"context"
	"log"
	"os"

	"github.com/apache/arrow-go/v18/arrow/memory"
	"github.com/apache/arrow-go/v18/parquet"
	"github.com/apache/arrow-go/v18/parquet/file"
	"github.com/apache/arrow-go/v18/parquet/pqarrow"
)

func main() {
	pFile, err := os.Open("<path_to_file>")
	if err != nil {
		log.Fatal(err)
	}

	ctx := context.Background()
	opts := &parquet.ReaderProperties{
		BufferSize:            1024,
		FileDecryptProps:      nil,
		BufferedStreamEnabled: true,
	}

	base, err := file.NewParquetReader(pFile, file.WithReadProps(opts))
	if err != nil {
		log.Fatal(err)
	}

	pReader, err := pqarrow.NewFileReader(base, pqarrow.ArrowReadProperties{
		Parallel:  true,
		BatchSize: 3000,
	}, memory.DefaultAllocator)
	if err != nil {
		log.Fatal(err)
	}

	numOfRowGroups := pReader.ParquetReader().NumRowGroups()

	for rgIdx := range numOfRowGroups {
		rReader, err := pReader.GetRecordReader(ctx, nil, []int{rgIdx})
		if err != nil {
			log.Fatal(err)
		}

		for rReader.Next() {
			batch := rReader.RecordBatch()
			log.Printf("got batch: %v records", batch.NumRows())
		}
	}
}

[Sunyue]

test2.snappy.parquet.zip You can reproduce using this file with above test program.
BTW, just tired with v18.4.1 and it worked fine, seems to be an regression introduced with v18.5.0