[
https://issues.apache.org/jira/browse/BEAM-9152?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17539491#comment-17539491
]
Jean-Christophe CARLES commented on BEAM-9152:
----------------------------------------------
It looks like patching `apache_beam/io/hadoopfilesystem.py` by updating the
length parameter of the method `get_range` from the class `HdfsDownloader` in
the following way fixes the issue:
```python
length=end - start + 1 --> length=end - start
```
Although it feels weird that this was not noticed more, is anyone using the
python sdk to read hdfs? Maybe we are doing something wrong...
> Hadoop Downloader Range not correct
> -----------------------------------
>
> Key: BEAM-9152
> URL: https://issues.apache.org/jira/browse/BEAM-9152
> Project: Beam
> Issue Type: Bug
> Components: io-py-tfrecord
> Reporter: Fangyuan Zhou
> Priority: P3
> Fix For: Missing
>
>
> I found that the `HdfsDownloader.get_range(self, start, end)' get range
> {color:#ff0000}[start, end]{color} rather than {color:#ff0000}[start,
> end){color}. This will cause an error while reading hdfs file.
>
> File
> "/data/anaconda3/envs/tfdv1.15/lib/python3.7/site-packages/apache_beam/io/tfrecordio.py",
> line 127, in read_record
> buf = file_handle.read(buf_length_expected)
> File
> "/data/anaconda3/envs/tfdv1.15/lib/python3.7/site-packages/apache_beam/io/filesystemio.py",
> line 123, in readinto
> b[:len(data)] = data
> ValueError: memoryview assignment: lvalue and rvalue have different
> structures
--
This message was sent by Atlassian Jira
(v8.20.7#820007)
