Steve Loughran created PARQUET-2151:
---------------------------------------
Summary: parquet-hadoop to drop Hadoop 1 input stream support
Key: PARQUET-2151
URL: https://issues.apache.org/jira/browse/PARQUET-2151
Project: Parquet
Issue Type: Improvement
Components: parquet-mr
Affects Versions: 1.13.0
Reporter: Steve Loughran
Parquet uses reflection to load a hadoop2 input stream, falling back to a
hadoop-1 compatible client if not found.
All hadoop 2.0.2+ releases work with H2SeekableInputStream, so
H1SeekableInputStream can be cut and the binding to H2SeekableInputStream
reworked to avoid needing reflection. This would make it a lot easier to probe
for/use the bytebuffer input, and line the code up for more recent hadoop
releases.
One thing H1SeekableInputStream does do is read into a temp array if the
FSDataInputStream doesn't support , that is, doesn't implement
ByteBufferReadable.
but FSDataInputStream simply forwards that to the inner stream, if it too
implements ByteBufferReadable. Filesystems which don't (the cloud stores) can't
be read through H2SeekableInputStream.read(ByteBufferReadable). If this
desired, H2SeekableInputStream will need to dynamically downgrade to
DelegatingSeekableInputStream's base methods if a call to
FSDataInputStream.read(ByteBuffer) fails.
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
