[
https://issues.apache.org/jira/browse/SPARK-34463?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17286478#comment-17286478
]
David Li commented on SPARK-34463:
----------------------------------
I'll take a look.
The reason for the three options is as follows:
* use_threads=False - convert each column sequentially to minimize memory
usage (overhead = 1 column's worth of memory at any time instead of N columns
where N = parallelism level)
* split_blocks=True - create a separate Pandas block for each column. This is
an internal implementation detail, but it makes it more likely for PyArrow to
find zero-copy opportunities and reduce memory usage/conversion overhead even
more
* self_destruct=True - free the PyArrow array after each conversion to save
memory (technically it's just decrementing a refcount so if you have any other
references to the array, no memory is freed)
> toPandas failed with error: buffer source array is read-only
> ------------------------------------------------------------
>
> Key: SPARK-34463
> URL: https://issues.apache.org/jira/browse/SPARK-34463
> Project: Spark
> Issue Type: Bug
> Components: PySpark
> Affects Versions: 3.0.2
> Reporter: Weichen Xu
> Priority: Major
>
> Environment:
> apache/spark master
> pandas version > 1.0.5
> Reproduce code:
> {code}
> spark.conf.set('spark.sql.execution.arrow.pyspark.enabled', True)
> spark.conf.set('spark.sql.execution.arrow.pyspark.selfDestruct.enabled',
> True) spark.createDataFrame(sc.parallelize([(i,) for i
> in range(13)], 1), 'id long').selectExpr('IF(id % 3==0, id+1, NULL) AS f1',
> '(id+1) % 2 AS label').toPandas()['label'].value_counts()
> {code}
> Get error like:
> {quote}Traceback (most recent call last):
>
> File "<stdin>", line 1, in <module>
> File
> "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pandas/core/base.py",
> line 1033, in value_counts
> dropna=dropna,
> File
> "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pandas/core/algorithms.py",
> line 820, in value_counts
> keys, counts = value_counts_arraylike(values, dropna)
> File
> "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/pandas/core/algorithms.py",
> line 865, in value_counts_arraylike
> keys, counts = f(values, dropna)
> File "pandas/_libs/hashtable_func_helper.pxi", line 1098, in
> pandas._libs.hashtable.value_count_int64
> File "stringsource", line 658, in View.MemoryView.memoryview_cwrapper
> File "stringsource", line 349, in View.MemoryView.memoryview.__cinit__
> ValueError: buffer source array is read-only
> {quote}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
