zeroshade commented on code in PR #38367:
URL: https://github.com/apache/arrow/pull/38367#discussion_r1376402450
##########
go/parquet/internal/encoding/boolean_decoder.go:
##########
@@ -109,3 +114,76 @@ func (dec *PlainBooleanDecoder) DecodeSpaced(out []bool,
nullCount int, validBit
}
return dec.Decode(out)
}
+
+type RleBooleanDecoder struct {
+ decoder
+
+ rleDec *utils.RleDecoder
+}
+
+func (RleBooleanDecoder) Type() parquet.Type {
+ return parquet.Types.Boolean
+}
+
+func (dec *RleBooleanDecoder) SetData(nvals int, data []byte) error {
+ dec.nvals = nvals
+
+ if len(data) < 4 {
+ return fmt.Errorf("invalid length - %d (corrupt data page?)",
len(data))
+ }
+
+ // load the first 4 bytes in little-endian which indicates the length
+ nbytes := binary.LittleEndian.Uint32(data[:4])
+ if nbytes > uint32(len(data)-4) {
+ return fmt.Errorf("received invalid number of bytes - %d
(corrupt data page?)", nbytes)
+ }
+
+ dec.data = data[4:]
+ if dec.rleDec == nil {
+ dec.rleDec = utils.NewRleDecoder(bytes.NewReader(dec.data), 1)
+ } else {
+ dec.rleDec.Reset(bytes.NewReader(dec.data), 1)
+ }
+ return nil
+}
+
+func (dec *RleBooleanDecoder) Decode(out []bool) (int, error) {
+ max := shared_utils.MinInt(len(out), dec.nvals)
+
+ var (
+ buf [1024]uint64
+ n = max
+ )
+
+ for n > 0 {
+ batch := shared_utils.MinInt(len(buf), n)
+ decoded := dec.rleDec.GetBatch(buf[:batch])
+ if decoded != batch {
+ return max - n, io.ErrUnexpectedEOF
Review Comment:
No, that wouldn't happen because parquet doesn't include nulls in the
underlying data. If the data is
```
nullable: [false, false, false, false, null]`
```
Then the *physical data in the file* is just `[false, false, false, false]`
and the def levels would have `[1, 1, 1, 1, 0]`.
Decode(5) would still get `4, nil` because only 4 *physical values* were
decoded. This scenario is also tested in the unit tests.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
