-
Notifications
You must be signed in to change notification settings - Fork 4k
ARROW-12851: [Go][Parquet] Add Golang Parquet encoding package #10379
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Changes from all commits
e9e8461
ade330c
4f6318c
dc8537f
163fc08
902495c
5f607a7
db9d93b
041295b
e6c239c
99ef6bb
1808793
03224b8
a2dffb1
5d54a30
da6a92b
fba1f23
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,101 @@ | ||
| // Licensed to the Apache Software Foundation (ASF) under one | ||
| // or more contributor license agreements. See the NOTICE file | ||
| // distributed with this work for additional information | ||
| // regarding copyright ownership. The ASF licenses this file | ||
| // to you under the Apache License, Version 2.0 (the | ||
| // "License"); you may not use this file except in compliance | ||
| // with the License. You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, software | ||
| // distributed under the License is distributed on an "AS IS" BASIS, | ||
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
|
|
||
| package encoding | ||
|
|
||
| import ( | ||
| "github.com/apache/arrow/go/arrow/bitutil" | ||
| "github.com/apache/arrow/go/parquet" | ||
| "github.com/apache/arrow/go/parquet/internal/utils" | ||
| "golang.org/x/xerrors" | ||
| ) | ||
|
|
||
| // PlainBooleanDecoder is for the Plain Encoding type, there is no | ||
| // dictionary decoding for bools. | ||
| type PlainBooleanDecoder struct { | ||
| decoder | ||
|
|
||
| bitOffset int | ||
| } | ||
|
|
||
| // Type for the PlainBooleanDecoder is parquet.Types.Boolean | ||
| func (PlainBooleanDecoder) Type() parquet.Type { | ||
| return parquet.Types.Boolean | ||
| } | ||
|
|
||
| // Decode fills out with bools decoded from the data at the current point | ||
| // or until we reach the end of the data. | ||
| // | ||
| // Returns the number of values decoded | ||
| func (dec *PlainBooleanDecoder) Decode(out []bool) (int, error) { | ||
| max := utils.MinInt(len(out), dec.nvals) | ||
|
|
||
| unalignedExtract := func(start, end, curBitOffset int) int { | ||
| i := start | ||
| for ; curBitOffset < end; i, curBitOffset = i+1, curBitOffset+1 { | ||
| out[i] = (dec.data[0] & byte(1<<curBitOffset)) != 0 | ||
| } | ||
| return i // return the number of bits we extracted | ||
| } | ||
|
|
||
| // if we aren't at a byte boundary, then get bools until we hit | ||
| // a byte boundary with the bit offset. | ||
| i := 0 | ||
| if dec.bitOffset != 0 { | ||
| i = unalignedExtract(0, 8, dec.bitOffset) | ||
| dec.bitOffset = 0 | ||
| } | ||
|
|
||
| // determine the number of full bytes worth of bits we can decode | ||
| // given the number of values we want to decode. | ||
| bitsRemain := max - i | ||
| batch := bitsRemain / 8 * 8 | ||
| if batch > 0 { // only go in here if there's at least one full byte to decode | ||
| if i > 0 { // skip our data forward if we decoded anything above | ||
| dec.data = dec.data[1:] | ||
| out = out[i:] | ||
| } | ||
| // determine the number of aligned bytes we can grab using SIMD optimized | ||
| // functions to improve performance. | ||
| alignedBytes := bitutil.BytesForBits(int64(batch)) | ||
| utils.BytesToBools(dec.data[:alignedBytes], out) | ||
| dec.data = dec.data[alignedBytes:] | ||
| out = out[alignedBytes*8:] | ||
| } | ||
|
|
||
| // grab any trailing bits now that we've got our aligned bytes. | ||
| dec.bitOffset += unalignedExtract(dec.bitOffset, bitsRemain-batch, dec.bitOffset) | ||
|
|
||
| dec.nvals -= max | ||
| return max, nil | ||
| } | ||
|
|
||
| // DecodeSpaced is like Decode except it expands the values to leave spaces for null | ||
| // as determined by the validBits bitmap. | ||
| func (dec *PlainBooleanDecoder) DecodeSpaced(out []bool, nullCount int, validBits []byte, validBitsOffset int64) (int, error) { | ||
| if nullCount > 0 { | ||
| toRead := len(out) - nullCount | ||
| valuesRead, err := dec.Decode(out[:toRead]) | ||
| if err != nil { | ||
| return 0, err | ||
| } | ||
| if valuesRead != toRead { | ||
| return valuesRead, xerrors.New("parquet: boolean decoder: number of values / definition levels read did not match") | ||
| } | ||
| return spacedExpand(out, nullCount, validBits, validBitsOffset), nil | ||
| } | ||
| return dec.Decode(out) | ||
| } |
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,84 @@ | ||
| // Licensed to the Apache Software Foundation (ASF) under one | ||
| // or more contributor license agreements. See the NOTICE file | ||
| // distributed with this work for additional information | ||
| // regarding copyright ownership. The ASF licenses this file | ||
| // to you under the Apache License, Version 2.0 (the | ||
| // "License"); you may not use this file except in compliance | ||
| // with the License. You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, software | ||
| // distributed under the License is distributed on an "AS IS" BASIS, | ||
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
|
|
||
| package encoding | ||
|
|
||
| import ( | ||
| "github.com/apache/arrow/go/arrow/bitutil" | ||
| "github.com/apache/arrow/go/parquet" | ||
| "github.com/apache/arrow/go/parquet/internal/utils" | ||
| ) | ||
|
|
||
| const ( | ||
| boolBufSize = 1024 | ||
| boolsInBuf = boolBufSize * 8 | ||
| ) | ||
|
|
||
| // PlainBooleanEncoder encodes bools as a bitmap as per the Plain Encoding | ||
| type PlainBooleanEncoder struct { | ||
| encoder | ||
| bitsBuffer []byte | ||
| wr utils.BitmapWriter | ||
| } | ||
|
|
||
| // Type for the PlainBooleanEncoder is parquet.Types.Boolean | ||
| func (PlainBooleanEncoder) Type() parquet.Type { | ||
| return parquet.Types.Boolean | ||
| } | ||
|
|
||
| // Put encodes the contents of in into the underlying data buffer. | ||
| func (enc *PlainBooleanEncoder) Put(in []bool) { | ||
| if enc.bitsBuffer == nil { | ||
| enc.bitsBuffer = make([]byte, boolBufSize) | ||
| } | ||
| if enc.wr == nil { | ||
| enc.wr = utils.NewBitmapWriter(enc.bitsBuffer, 0, boolsInBuf) | ||
| } | ||
|
|
||
| n := enc.wr.AppendBools(in) | ||
| for n < len(in) { | ||
| enc.wr.Finish() | ||
| enc.append(enc.bitsBuffer) | ||
| enc.wr.Reset(0, boolsInBuf) | ||
| in = in[n:] | ||
| n = enc.wr.AppendBools(in) | ||
| } | ||
| } | ||
|
|
||
| // PutSpaced will use the validBits bitmap to determine which values are nulls | ||
| // and can be left out from the slice, and the encoded without those nulls. | ||
| func (enc *PlainBooleanEncoder) PutSpaced(in []bool, validBits []byte, validBitsOffset int64) { | ||
| bufferOut := make([]bool, len(in)) | ||
| nvalid := spacedCompress(in, bufferOut, validBits, validBitsOffset) | ||
| enc.Put(bufferOut[:nvalid]) | ||
| } | ||
|
|
||
| // EstimatedDataEncodedSize returns the current number of bytes that have | ||
| // been buffered so far | ||
| func (enc *PlainBooleanEncoder) EstimatedDataEncodedSize() int64 { | ||
| return int64(enc.sink.Len() + int(bitutil.BytesForBits(enc.wr.Pos()))) | ||
| } | ||
|
|
||
| // FlushValues returns the buffered data, the responsibility is on the caller | ||
| // to release the buffer memory | ||
| func (enc *PlainBooleanEncoder) FlushValues() Buffer { | ||
| if enc.wr.Pos() > 0 { | ||
| toFlush := int(enc.wr.Pos()) | ||
| enc.append(enc.bitsBuffer[:bitutil.BytesForBits(int64(toFlush))]) | ||
| } | ||
|
|
||
| return enc.sink.Finish() | ||
| } |
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,88 @@ | ||
| // Licensed to the Apache Software Foundation (ASF) under one | ||
| // or more contributor license agreements. See the NOTICE file | ||
| // distributed with this work for additional information | ||
| // regarding copyright ownership. The ASF licenses this file | ||
| // to you under the Apache License, Version 2.0 (the | ||
| // "License"); you may not use this file except in compliance | ||
| // with the License. You may obtain a copy of the License at | ||
| // | ||
| // http://www.apache.org/licenses/LICENSE-2.0 | ||
| // | ||
| // Unless required by applicable law or agreed to in writing, software | ||
| // distributed under the License is distributed on an "AS IS" BASIS, | ||
| // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. | ||
| // See the License for the specific language governing permissions and | ||
| // limitations under the License. | ||
|
|
||
| package encoding | ||
|
|
||
| import ( | ||
| "encoding/binary" | ||
|
|
||
| "github.com/apache/arrow/go/parquet" | ||
| "github.com/apache/arrow/go/parquet/internal/utils" | ||
| "golang.org/x/xerrors" | ||
| ) | ||
|
|
||
| // PlainByteArrayDecoder decodes a data chunk for bytearrays according to | ||
| // the plain encoding. The byte arrays will use slices to reference the | ||
| // data rather than copying it. | ||
| // | ||
| // The parquet spec defines Plain encoding for ByteArrays as a 4 byte little | ||
| // endian integer containing the length of the bytearray followed by that many | ||
| // bytes being the raw data of the byte array. | ||
| type PlainByteArrayDecoder struct { | ||
| decoder | ||
| } | ||
|
|
||
| // Type returns parquet.Types.ByteArray for this decoder | ||
| func (PlainByteArrayDecoder) Type() parquet.Type { | ||
| return parquet.Types.ByteArray | ||
| } | ||
|
|
||
| // Decode will populate the slice of bytearrays in full or until the number | ||
| // of values is consumed. | ||
| // | ||
| // Returns the number of values that were decoded. | ||
| func (pbad *PlainByteArrayDecoder) Decode(out []parquet.ByteArray) (int, error) { | ||
| max := utils.MinInt(len(out), pbad.nvals) | ||
|
|
||
| for i := 0; i < max; i++ { | ||
| // there should always be at least four bytes which is the length of the | ||
| // next value in the data. | ||
| if len(pbad.data) < 4 { | ||
| return i, xerrors.New("parquet: eof reading bytearray") | ||
| } | ||
|
|
||
| // the first 4 bytes are a little endian int32 length | ||
| byteLen := int32(binary.LittleEndian.Uint32(pbad.data[:4])) | ||
| if byteLen < 0 { | ||
| return i, xerrors.New("parquet: invalid BYTE_ARRAY value") | ||
| } | ||
|
|
||
| if int64(len(pbad.data)) < int64(byteLen)+4 { | ||
| return i, xerrors.New("parquet: eof reading bytearray") | ||
| } | ||
|
|
||
| out[i] = pbad.data[4 : byteLen+4 : byteLen+4] | ||
| pbad.data = pbad.data[byteLen+4:] | ||
| } | ||
|
|
||
| pbad.nvals -= max | ||
| return max, nil | ||
| } | ||
|
|
||
| // DecodeSpaced is like Decode, but expands the slice out to leave empty values | ||
| // where the validBits bitmap has 0s | ||
| func (pbad *PlainByteArrayDecoder) DecodeSpaced(out []parquet.ByteArray, nullCount int, validBits []byte, validBitsOffset int64) (int, error) { | ||
|
||
| toRead := len(out) - nullCount | ||
| valuesRead, err := pbad.Decode(out[:toRead]) | ||
| if err != nil { | ||
| return valuesRead, err | ||
| } | ||
| if valuesRead != toRead { | ||
| return valuesRead, xerrors.New("parquet: number of values / definition levels read did not match") | ||
| } | ||
|
|
||
| return spacedExpand(out, nullCount, validBits, validBitsOffset), nil | ||
| } | ||
Uh oh!
There was an error while loading. Please reload this page.