AlenkaF commented on PR #36266:
URL: https://github.com/apache/arrow/pull/36266#issuecomment-1818837257
I have found out the cause for the random flakey failures in the CI.
The code in the tests is generating a numpy array with random strings of
length 4. They can also include the string `"NULL"` which currently gets
converted to `None`.
<details>
<summary>Reproducible example</summary>
```python
import pyarrow as pa
import numpy as np
import os
from pyarrow.csv import (read_csv, write_csv,
ConvertOptions, ParseOptions,
ReadOptions, WriteOptions)
file_name = "fixedsize_NULL"
nparr = np.array([b'NULL'])
fixed_arr = pa.array(nparr, pa.binary(4))
fixed_table = pa.Table.from_arrays([fixed_arr], names=['fixedsize'])
write_options = WriteOptions(include_header=True, batch_size=2048,
delimiter='|')
write_csv(fixed_table, file_name, write_options=write_options)
assert os.path.exists(file_name)
parse_options = ParseOptions(delimiter="|")
convert_options = ConvertOptions(column_types={"fixedsize": pa.binary(4)})
read_options = ReadOptions(block_size=2048)
res_table = read_csv(file_name, parse_options=parse_options,
convert_options=convert_options,
read_options=read_options)
```
```python
>>> nparr
array([b'NULL'], dtype='|S4')
>>> fixed_table
pyarrow.Table
fixedsize: fixed_size_binary[4]
----
fixedsize: [[4E554C4C]]
>>> res_table
pyarrow.Table
fixedsize: fixed_size_binary[4]
----
fixedsize: [[null]]
```
</details>
We should use `convert_options = ConvertOptions(column_types={"fixedsize":
pa.binary(4)}, quoted_strings_can_be_null=False)` so that `"NULL"` is not
converted to `None`.
Will add the change via suggestions.
--
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]
