[
https://issues.apache.org/jira/browse/SPARK-19507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15858565#comment-15858565
]
David Gingrich edited comment on SPARK-19507 at 2/8/17 9:37 PM:
----------------------------------------------------------------
I'd advocate making it public, but I was saving that for a separate discussion
sine it's a bigger change :)
{{_verfy_type}} is called from {{pyspark.sql.session.createDataFrame}} if
{{verifySchema is True}}, so better debug messages would help that usecase.
I think the main reason to add it is that otherwise the output isn't useful for
debugging. The change is also simple (adding a {{name}} parameter that is used
in the exception messages and augmented for recursive calls). I have a patch
but I wanted to add tests before submitting the PR.
was (Author: dgingrich):
I'd advocate that it be made public, but I was saving that for a separate
discussion sine it's a bigger change :)
{{_verfy_type}} is called from {{pyspark.sql.session.createDataFrame}} if
{{verifySchema is True}}, so better debug messages would help that usecase.
I think the main reason to add it is that otherwise the output isn't useful for
debugging. The change is also simple (adding a {{name}} parameter that is used
in the exception messages and augmented for recursive calls. I have a patch
but I wanted to add tests before submitting the PR.
> pyspark.sql.types._verify_type() exceptions too broad to debug collections or
> nested data
> -----------------------------------------------------------------------------------------
>
> Key: SPARK-19507
> URL: https://issues.apache.org/jira/browse/SPARK-19507
> Project: Spark
> Issue Type: New Feature
> Components: PySpark
> Affects Versions: 2.1.0
> Environment: macOS Sierra 10.12.3
> Spark 2.1.0, installed via Homebrew
> Reporter: David Gingrich
> Priority: Trivial
>
> The private function pyspark.sql.types._verify_type() recursively checks an
> object against a datatype, raising an exception if the object does not
> satisfy the type. These messages are not specific enough to debug a data
> error in a collection or nested data, for instance:
> {quote}
> >>> import pyspark.sql.types as typ
> >>> schema = typ.StructType([typ.StructField('nest1',
> >>> typ.MapType(typ.StringType(), typ.ArrayType(typ.FloatType())))])
> >>> typ._verify_type({'nest1': {'nest2': [1]}}, schema)
> Traceback (most recent call last):
> File "<stdin>", line 1, in <module>
> File "/Users/david/src/3p/spark/python/pyspark/sql/types.py", line 1355, in
> _verify_type
> _verify_type(obj.get(f.name), f.dataType, f.nullable, name=new_name)
> File "/Users/david/src/3p/spark/python/pyspark/sql/types.py", line 1349, in
> _verify_type
> _verify_type(v, dataType.valueType, dataType.valueContainsNull,
> name=new_name)
> File "/Users/david/src/3p/spark/python/pyspark/sql/types.py", line 1342, in
> _verify_type
> _verify_type(i, dataType.elementType, dataType.containsNull,
> name=new_name)
> File "/Users/david/src/3p/spark/python/pyspark/sql/types.py", line 1325, in
> _verify_type
> % (name, dataType, obj, type(obj)))
> TypeError: FloatType can not accept object 1 in type <class 'int'>
> {quote}
> Passing and printing a field name would make debugging easier.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
