[Go] DeltaBitPack deltaBitWidths overflow

[Hattonuri]

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

I wrote some int64 using c++ ::parquet::Encoding::DELTA_BINARY_PACKED
and then tried to read this with go but got:

panic: runtime error: index out of range [4] with length 4

goroutine 199 [running]:
github.com/apache/arrow/go/v14/parquet/internal/encoding.(*DeltaBitPackInt64Decoder).unpackNextMini(0xc1aee302a0)
        /home/dstasenko/go/pkg/mod/github.com/apache/arrow/go/v14@v14.0.0-20231023113027-37935604bf16/parquet/internal/encoding/delta_bit_packing.go:230 +0x2da
github.com/apache/arrow/go/v14/parquet/internal/encoding.(*DeltaBitPackInt64Decoder).Decode(0xc1aee302a0, {0xc1cc73b008?, 0xa7d50d?, 0xc1adb08360?})
        /home/dstasenko/go/pkg/mod/github.com/apache/arrow/go/v14@v14.0.0-20231023113027-37935604bf16/parquet/internal/encoding/delta_bit_packing.go:267 +0x118
github.com/apache/arrow/go/v14/parquet/file.(*Int64ColumnChunkReader).ReadBatch.func1(0x15a201, 0x167d01)
        /home/dstasenko/go/pkg/mod/github.com/apache/arrow/go/v14@v14.0.0-20231023113027-37935604bf16/parquet/file/column_reader_types.gen.go:93 +0xa3
github.com/apache/arrow/go/v14/parquet/file.(*columnChunkReader).readBatch(0xc1adb08360, 0x3c9c00, {0xc1d3f26000, 0x3c9c00, 0x3c9c00}, {0x0, 0x0, 0x0}, 0xc1aeb73c80)
        /home/dstasenko/go/pkg/mod/github.com/apache/arrow/go/v14@v14.0.0-20231023113027-37935604bf16/parquet/file/column_reader.go:508 +0xc2
github.com/apache/arrow/go/v14/parquet/file.(*Int64ColumnChunkReader).ReadBatch(0xc1aee2e3d0?, 0xc1aee36960?, {0xc1cbc6a000?, 0x2b21d60?, 0xc0001a9d58?}, {0xc1d3f26000?, 0x1a49d55?, 0xa?}, {0x0, 0x0, ...})

I assume that this line lacks boundaries check

arrow/go/parquet/internal/encoding/delta_bit_packing.go

Line 228 in ac581fd

d.deltaBitWidth = d.deltaBitWidths.Bytes()[int(d.miniBlockIdx)]

Component(s)

Go

[mapleFU]

Would you mind provide the test data which might trigger the bug here?

[Hattonuri]

storage_000000000.parquet.txt
added .txt ending because github does not allow to upload parquet files

[Hattonuri]

arrow/go/parquet/internal/encoding/delta_bit_packing.go

Lines 175 to 176 in ac581fd

	}
	if d.currentMiniBlockVals == 0 {

probably you forgot "if err != nil return"

[mapleFU]

I've found if err != nil return, but it's still propogate an error

C++ is able to read file like this, let me investigate it.

[mapleFU]

I've find out the reason. Let me submit a fixing

[mapleFU]

@Hattonuri I've submit a fixing here: #38413

You can have a try at the new code.

[Hattonuri]

Now it works! Thank you