You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@spark.apache.org by "Qionghui Zhang (Jira)" <ji...@apache.org> on 2020/07/07 07:54:00 UTC
[jira] [Updated] (SPARK-32206) Enable multi-line true could break
the read csv in Azure Data Lake Storage gen2
[ https://issues.apache.org/jira/browse/SPARK-32206?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Qionghui Zhang updated SPARK-32206:
-----------------------------------
Description:
I'm using azure data lake gen2, when I'm loading data frame with certain options:
var df = spark.read.format("csv")
.option("ignoreLeadingWhiteSpace", "true")
.option("ignoreTrailingWhiteSpace", "true")
.option("parserLib", "UNIVOCITY")
.option("multiline", "true")
.option("inferSchema", "true")
.option("mode", "PERMISSIVE")
.option("quote", "\"")
.option("escape", "\"")
.option("timeStampFormat", "M/d/yyyy H:m:s a")
.load("abfss://\{containername}@\{storage}.dfs.core.windows.net/\{DirectoryWithoutColon}")
.limit(1)
It will load data correctly.
But if I use \{DirectoryWithColon}, it will thrown error:
java.lang.IllegalArgumentException: java.net.URISyntaxException: Relative path in absolute URI: \{somefilesnapshotname}yyyy-MM-dd'T'hh:mm:ss
Then if I remove .option("multiline", "true"), data can be loaded, but for sure that the dataframe is not handled correctly because there are newline character.
So I believe it is a bug.
And since our production is running correctly if we enable spark.read.schema(\{SomeSchemaList}).format("csv"), and we want to use inferschema feature on those file path with colon or other special characters, could you help fix this issue?
was:
I'm using azure data lake gen2, when I'm loading data frame with certain options:
var df = spark.read.format("csv")
.option("ignoreLeadingWhiteSpace", "true")
.option("ignoreTrailingWhiteSpace", "true")
.option("parserLib", "UNIVOCITY")
.option("multiline", "true")
.option("inferSchema", "true")
.option("mode", "PERMISSIVE")
.option("quote", "\"")
.option("escape", "\"")
.option("timeStampFormat", "M/d/yyyy H:m:s a")
.load("abfss://\{containername}@\{storage}.dfs.core.windows.net/\{DirectoryWithoutColon}")
.limit(1)
It will load data correctly.
But if I use \{DirectoryWithColon}, it will thrown error:
java.lang.IllegalArgumentException: java.net.URISyntaxException: Relative path in absolute URI: Query.csv@snapshot=2020-07-07T07:30:12.1000093Z.
Then if I remove .option("multiline", "true"), data can be loaded, but for sure that the dataframe is not handled correctly because there are newline character.
So I believe it is a bug.
And since our production is running correctly if we enable spark.read.schema(\{SomeSchemaList}).format("csv"), and we want to use inferschema feature on those file path with colon or other special characters, could you help fix this issue?
> Enable multi-line true could break the read csv in Azure Data Lake Storage gen2
> -------------------------------------------------------------------------------
>
> Key: SPARK-32206
> URL: https://issues.apache.org/jira/browse/SPARK-32206
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
> Affects Versions: 2.3.2, 2.4.5
> Reporter: Qionghui Zhang
> Priority: Major
>
> I'm using azure data lake gen2, when I'm loading data frame with certain options:
> var df = spark.read.format("csv")
> .option("ignoreLeadingWhiteSpace", "true")
> .option("ignoreTrailingWhiteSpace", "true")
> .option("parserLib", "UNIVOCITY")
> .option("multiline", "true")
> .option("inferSchema", "true")
> .option("mode", "PERMISSIVE")
> .option("quote", "\"")
> .option("escape", "\"")
> .option("timeStampFormat", "M/d/yyyy H:m:s a")
> .load("abfss://\{containername}@\{storage}.dfs.core.windows.net/\{DirectoryWithoutColon}")
> .limit(1)
> It will load data correctly.
>
> But if I use \{DirectoryWithColon}, it will thrown error:
> java.lang.IllegalArgumentException: java.net.URISyntaxException: Relative path in absolute URI: \{somefilesnapshotname}yyyy-MM-dd'T'hh:mm:ss
>
> Then if I remove .option("multiline", "true"), data can be loaded, but for sure that the dataframe is not handled correctly because there are newline character.
>
> So I believe it is a bug.
>
> And since our production is running correctly if we enable spark.read.schema(\{SomeSchemaList}).format("csv"), and we want to use inferschema feature on those file path with colon or other special characters, could you help fix this issue?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscribe@spark.apache.org
For additional commands, e-mail: issues-help@spark.apache.org