[C++][Parquet] Read string columns directly into STRING_VIEW arrays and cast to LARGE_STRING if necessary

[felipecrv]

Describe the enhancement requested

This would fix two issues for the price of one:

1. Reading from Parquet into schemas that use the new STRING_VIEW type
2. Reading LARGE_STRING_ARRAY from Parquet ([C++] Parquet reader is unable to read LargeString columns #39682)

This issue also depends on:

• [C++] Support casting to and from utf8_view/binary_view #42247
• [C++][Parquet] Reading/Writing string/binary types as/with the corresponding arrow view type in Parquet #43041

Component(s)

C++, Parquet

[mapleFU]

Related: apache/arrow-rs#5530

This can also applying "zero-copy" here for non Delta string encoding

[zzl200012]

Is there any updates about this feature? We have lots of parquet files containing plain encoded binary data, I think the reading performance will be much better if we can read the binary columns directly into binary_view arrays while applying the zero-copy method which arrow-rs already has.

[mapleFU]

Currently not. I'm glad to review this part of code but currently I don't have enough bandwidth to working on this

This might need a minor refactor on BinaryRecordReader (

arrow/cpp/src/parquet/column_reader.h

Line 441 in 2d76789

class BinaryRecordReader : virtual public RecordReader {

) and Decoder api