You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Marco Gaido (JIRA)" <ji...@apache.org> on 2018/12/11 13:38:00 UTC
[jira] [Commented] (SPARK-26339) Behavior of reading files that
start with underscore is confusing
[ https://issues.apache.org/jira/browse/SPARK-26339?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16717168#comment-16717168 ]
Marco Gaido commented on SPARK-26339:
-------------------------------------
The point is: files starting with underscores are hidden files in Hadoop FS. So what you are doing is the same as reading an empty folder:
- if you set the schema, an empty dataframe is returned;
- if you don't set it, the schema will be inferred from the data, but since there is no data the exception occurs.
I don't think this is a bug.
> Behavior of reading files that start with underscore is confusing
> -----------------------------------------------------------------
>
> Key: SPARK-26339
> URL: https://issues.apache.org/jira/browse/SPARK-26339
> Project: Spark
> Issue Type: Bug
> Components: SQL
> Affects Versions: 3.0.0
> Reporter: Keiichi Hirobe
> Priority: Minor
>
> Behavior of reading files that start with underscore is as follows.
> # spark.read (no schema) throws exception which message is confusing.
> # spark.read (userSpecificationSchema) succesfully reads, but content is emtpy.
> Example of files are as follows.
> The same behavior occured when I read json files.
> {code:bash}
> $ cat test.csv
> test1,10
> test2,20
> $ cp test.csv _test.csv
> $ ./bin/spark-shell --master local[2]
> {code}
> Spark shell snippet for reproduction:
> {code:java}
> scala> val df=spark.read.csv("test.csv")
> df: org.apache.spark.sql.DataFrame = [_c0: string, _c1: string]
> scala> df.show()
> +-----+---+
> | _c0|_c1|
> +-----+---+
> |test1| 10|
> |test2| 20|
> +-----+---+
> scala> val df = spark.read.schema("test STRING, number INT").csv("test.csv")
> df: org.apache.spark.sql.DataFrame = [test: string, number: int]
> scala> df.show()
> +-----+------+
> | test|number|
> +-----+------+
> |test1| 10|
> |test2| 20|
> +-----+------+
> scala> val df=spark.read.csv("_test.csv")
> org.apache.spark.sql.AnalysisException: Unable to infer schema for CSV. It must be specified manually.;
> at org.apache.spark.sql.execution.datasources.DataSource.$anonfun$getOrInferFileFormatSchema$13(DataSource.scala:185)
> at scala.Option.getOrElse(Option.scala:138)
> at org.apache.spark.sql.execution.datasources.DataSource.getOrInferFileFormatSchema(DataSource.scala:185)
> at org.apache.spark.sql.execution.datasources.DataSource.resolveRelation(DataSource.scala:373)
> at org.apache.spark.sql.DataFrameReader.loadV1Source(DataFrameReader.scala:231)
> at org.apache.spark.sql.DataFrameReader.load(DataFrameReader.scala:219)
> at org.apache.spark.sql.DataFrameReader.csv(DataFrameReader.scala:625)
> at org.apache.spark.sql.DataFrameReader.csv(DataFrameReader.scala:478)
> ... 49 elided
> scala> val df=spark.read.schema("test STRING, number INT").csv("_test.csv")
> df: org.apache.spark.sql.DataFrame = [test: string, number: int]
> scala> df.show()
> +----+------+
> |test|number|
> +----+------+
> +----+------+
> {code}
> I noticed that spark cannot read files that start with underscore after I read some codes.(I could not find any documents about file name limitation)
> Above behavior is not good especially userSpecificationSchema case, I think.
> I suggest to throw exception which message is "Path does not exist" in both cases.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org