[
https://issues.apache.org/jira/browse/SPARK-35324?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17709352#comment-17709352
]
Willi Raschkowski commented on SPARK-35324:
-------------------------------------------
I understand this better now:
* When calling {{SQLConf.get}} on executors, the configs are read from the
local properties on the {{{}TaskContext{}}}. The local properties are populated
driver-side when scheduling the job, using the properties found in
{{{}sparkContext.localProperties{}}}.
* For RDD actions like {{{}rdd.count{}}}, nothing populates moves driver-side
SQL configs into the SparkContext's local properites.
* For datasets, all actions incl. {{{}dataset.count{}}}, are wrapped in an
{{withAction}} call
[(e.g.)|https://github.com/apache/spark/blob/v3.3.2/sql/core/src/main/scala/org/apache/spark/sql/Dataset.scala#L3160].
* {{withAction}} wraps the action in {{SQLExecution.withNewExecutionId}},
which in turn wraps it in {{SQLExecution.withSQLConfPropagated}}. This latter
method copies SQL configs into the SparkContext's local properties.
So in summary, all actions on datasets get wrapped in {{withSQLConfPropagated}}
while actions on RDDs aren't. That's why {{df.count}} works but
{{df.rdd.count}} doesn't. With {{count}} the answer is to just use
{{Dataset.count}}. But, e.g., {{df.toLocalIterator}} has no alternative.
To fix this, Spark would have to always copy configs into local properties
(e.g. in {{submitJob}}). If maintainers likes that, I'll put up a PR. If not,
feel free to close.
In the meantime, my work-around is to call
{{SQLExecution.withSQLConfPropagated}} myself.
{code}
scala> spark.conf.set("spark.sql.legacy.timeParserPolicy", "legacy")
scala> spark.read.schema("date date").option("dateFormat",
"MM/dd/yy").csv(Seq("2/6/18").toDS()).toLocalIterator.next
23/04/06 13:58:26 ERROR Executor: Exception in task 0.0 in stage 0.0 (TID 0)
org.apache.spark.SparkUpgradeException:
[INCONSISTENT_BEHAVIOR_CROSS_VERSION.PARSE_DATETIME_BY_NEW_PARSER]
...
scala> SQLExecution.withSQLConfPropagated(spark) {
| spark.read.schema("date date").option("dateFormat",
"MM/dd/yy").csv(Seq("2/6/18").toDS()).toLocalIterator.next
| }
res2: org.apache.spark.sql.Row = [2018-02-06]
{code}
> Spark SQL configs not respected in RDDs
> ---------------------------------------
>
> Key: SPARK-35324
> URL: https://issues.apache.org/jira/browse/SPARK-35324
> Project: Spark
> Issue Type: Bug
> Components: Input/Output, SQL
> Affects Versions: 3.0.2, 3.1.1
> Reporter: Willi Raschkowski
> Priority: Major
> Attachments: Screen Shot 2021-05-06 at 00.33.10.png, Screen Shot
> 2021-05-06 at 00.35.10.png
>
>
> When reading a CSV file, {{spark.sql.timeParserPolicy}} is respected in
> actions on the resulting dataframe. But it's ignored in actions on
> dataframe's RDD.
> E.g. say to parse dates in a CSV you need {{spark.sql.timeParserPolicy}} to
> be set to {{LEGACY}}. If you set the config, {{df.collect}} will work as
> you'd expect. However, {{df.collect.rdd}} will fail because it'll ignore the
> override and read the config value as {{EXCEPTION}}.
> For instance:
> {code:java|title=test.csv}
> date
> 2/6/18
> {code}
> {code:java}
> scala> spark.conf.set("spark.sql.legacy.timeParserPolicy", "legacy")
> scala> val df = {
> | spark.read
> | .option("header", "true")
> | .option("dateFormat", "MM/dd/yy")
> | .schema("date date")
> | .csv("/Users/wraschkowski/Downloads/test.csv")
> | }
> df: org.apache.spark.sql.DataFrame = [date: date]
> scala> df.show
> +----------+
>
> | date|
> +----------+
> |2018-02-06|
> +----------+
> scala> df.count
> res3: Long = 1
> scala> df.rdd.count
> 21/05/06 00:06:18 ERROR Executor: Exception in task 0.0 in stage 4.0 (TID 4)
> org.apache.spark.SparkUpgradeException: You may get a different result due to
> the upgrading of Spark 3.0: Fail to parse '2/6/18' in the new parser. You can
> set spark.sql.legacy.timeParserPolicy to LEGACY to restore the behavior
> before Spark 3.0, or set to CORRECTED and treat it as an invalid datetime
> string.
> at
> org.apache.spark.sql.catalyst.util.DateTimeFormatterHelper$$anonfun$checkParsedDiff$1.applyOrElse(DateTimeFormatterHelper.scala:150)
> at
> org.apache.spark.sql.catalyst.util.DateTimeFormatterHelper$$anonfun$checkParsedDiff$1.applyOrElse(DateTimeFormatterHelper.scala:141)
> at
> scala.runtime.AbstractPartialFunction.apply(AbstractPartialFunction.scala:38)
> at
> org.apache.spark.sql.catalyst.util.Iso8601DateFormatter.$anonfun$parse$1(DateFormatter.scala:61)
> at
> scala.runtime.java8.JFunction0$mcI$sp.apply(JFunction0$mcI$sp.java:23)
> at scala.Option.getOrElse(Option.scala:189)
> at
> org.apache.spark.sql.catalyst.util.Iso8601DateFormatter.parse(DateFormatter.scala:58)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser.$anonfun$makeConverter$21(UnivocityParser.scala:202)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser.nullSafeDatum(UnivocityParser.scala:238)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser.$anonfun$makeConverter$20(UnivocityParser.scala:200)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser.org$apache$spark$sql$catalyst$csv$UnivocityParser$$convert(UnivocityParser.scala:291)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser.$anonfun$parse$2(UnivocityParser.scala:254)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser$.$anonfun$parseIterator$1(UnivocityParser.scala:396)
> at
> org.apache.spark.sql.catalyst.util.FailureSafeParser.parse(FailureSafeParser.scala:60)
> at
> org.apache.spark.sql.catalyst.csv.UnivocityParser$.$anonfun$parseIterator$2(UnivocityParser.scala:400)
> at scala.collection.Iterator$$anon$11.nextCur(Iterator.scala:484)
> at scala.collection.Iterator$$anon$11.hasNext(Iterator.scala:490)
> at scala.collection.Iterator$$anon$10.hasNext(Iterator.scala:458)
> at
> org.apache.spark.sql.execution.datasources.FileScanRDD$$anon$1.hasNext(FileScanRDD.scala:93)
> at
> org.apache.spark.sql.execution.datasources.FileScanRDD$$anon$1.nextIterator(FileScanRDD.scala:173)
> at
> org.apache.spark.sql.execution.datasources.FileScanRDD$$anon$1.hasNext(FileScanRDD.scala:93)
> at scala.collection.Iterator$$anon$10.hasNext(Iterator.scala:458)
> at scala.collection.Iterator$$anon$10.hasNext(Iterator.scala:458)
> at scala.collection.Iterator$$anon$10.hasNext(Iterator.scala:458)
> at org.apache.spark.util.Utils$.getIteratorSize(Utils.scala:1866)
> at org.apache.spark.rdd.RDD.$anonfun$count$1(RDD.scala:1253)
> at org.apache.spark.rdd.RDD.$anonfun$count$1$adapted(RDD.scala:1253)
> at
> org.apache.spark.SparkContext.$anonfun$runJob$5(SparkContext.scala:2242)
> at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:90)
> at org.apache.spark.scheduler.Task.run(Task.scala:131)
> at
> org.apache.spark.executor.Executor$TaskRunner.$anonfun$run$3(Executor.scala:497)
> at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1439)
> at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:500)
> at
> java.base/java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128)
> at
> java.base/java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628)
> at java.base/java.lang.Thread.run(Thread.java:834)
> Caused by: java.time.format.DateTimeParseException: Text '2/6/18' could not
> be parsed at index 0
> at
> java.base/java.time.format.DateTimeFormatter.parseResolved0(DateTimeFormatter.java:2046)
> at
> java.base/java.time.format.DateTimeFormatter.parse(DateTimeFormatter.java:1874)
> at
> org.apache.spark.sql.catalyst.util.Iso8601DateFormatter.$anonfun$parse$1(DateFormatter.scala:59)
> ... 32 more
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
