HeartSaVioR commented on a change in pull request #24173: [SPARK-27237][SS]
Introduce State schema validation among query restart
URL: https://github.com/apache/spark/pull/24173#discussion_r268171041
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/StateStore.scala
##########
@@ -477,5 +486,82 @@ object StateStore extends Logging {
None
}
}
-}
+ private class StateSchemaCompatibilityChecker(
+ providerId: StateStoreProviderId,
+ hadoopConf: Configuration) {
+
+ private val storeCpLocation = providerId.storeId.storeCheckpointLocation()
+ private val fm = CheckpointFileManager.create(storeCpLocation, hadoopConf)
+ private val schemaFileLocation = schemaFile(storeCpLocation)
+
+ fm.mkdirs(schemaFileLocation.getParent)
+
+ def check(keySchema: StructType, valueSchema: StructType): Unit = {
+ if (fm.exists(schemaFileLocation)) {
+ logDebug(s"Schema file for provider $providerId exists. Comparing with
provided schema.")
+ val (storedKeySchema, storedValueSchema) = readSchemaFile()
+
+ def typesEq(schema1: StructType, schema2: StructType): Boolean = {
+ val fieldToType: StructField => (DataType, Boolean) = f =>
(f.dataType, f.nullable)
+ (schema1.length == schema2.length) &&
+ schema1.map(fieldToType).equals(schema2.map(fieldToType))
+ }
+
+ def namesEq(schema1: StructType, schema2: StructType): Boolean = {
+ schema1.fieldNames.toList.equals(schema2.fieldNames.toList)
+ }
+
+ val errorMsg = "Provided schema doesn't match to the schema for
existing state! " +
+ "Please note that Spark allow difference of field name: check count
of fields " +
+ "and data type of each field.\n" +
+ s"- provided schema: key $keySchema value $valueSchema\n" +
+ s"- existing schema: key $storedKeySchema value $storedValueSchema\n"
+
+ if (!typesEq(keySchema, storedKeySchema) || !typesEq(valueSchema,
storedValueSchema)) {
+ logError(errorMsg)
+ throw new IllegalStateException(errorMsg)
+ } else if (!namesEq(keySchema, storedKeySchema) ||
+ !namesEq(valueSchema, storedValueSchema)) {
+ logInfo("Detected the change of field name, will overwrite schema
file to new.")
Review comment:
End users can change the column of aggregated column like
`sum("field").as("sum_val")` to `sum("field").as("sum_value")` which doesn't
change semantic. One counter case is swapping columns for min and max.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
