[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070933#comment-16070933 ] Kazuaki Ishizaki commented on SPARK-21271: -- In {{UnsafeRow}}, the length of fixed part should be a multiple of 8. However, the length of variable part may not require this regulation. [~cloud_fan] What do you think? > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070944#comment-16070944 ] Wenchen Fan commented on SPARK-21271: - We do have this regulation for var-length part in UnsafeRow, because var-length data is always word-aligned. This is also why we store the length information, otherwise we can just calculate the length by subtracting 2 adjacent offsets. > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070961#comment-16070961 ] Kazuaki Ishizaki commented on SPARK-21271: -- I see. For var-length part, its regulation (or specification) is a multiple of 4. Should we use {{hashUnsafeBytes}} instead of {{Murmur3_x86_32.hashUnsafeWords}} for {{UnsafeRow.hashCode()}}? > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070963#comment-16070963 ] Wenchen Fan commented on SPARK-21271: - For word-aligned I mean 8-bytes aligned, so the size of var-lengh data is always a multiple of 8. > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070966#comment-16070966 ] Kazuaki Ishizaki commented on SPARK-21271: -- I see. This issue comes from the violation such as [a code|https://github.com/apache/spark/blob/master/sql/catalyst/src/main/java/org/apache/spark/sql/catalyst/expressions/FixedLengthRowBasedKeyValueBatch.java#L65] (e.g. length + 4). Should we detect all of these code pieces by inserting an {{assert(length % 8 == 0)}} at {{pointTo()}}? > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070969#comment-16070969 ] Wenchen Fan commented on SPARK-21271: - yea we should. BTW the code seems wrong to be, the length of value row should be {{vlen}} instead of {{vlen + 4}} > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16070973#comment-16070973 ] Kazuaki Ishizaki commented on SPARK-21271: -- I see. I will work for this. Thank you for letting us how to fix it. I has been thinking about {{ + 4 }}. I also saw the similar code in [here|https://github.com/apache/spark/blob/master/sql/catalyst/src/main/java/org/apache/spark/sql/catalyst/expressions/VariableLengthRowBasedKeyValueBatch.java#L68] > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-21271) UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8
[ https://issues.apache.org/jira/browse/SPARK-21271?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16071551#comment-16071551 ] Apache Spark commented on SPARK-21271: -- User 'kiszk' has created a pull request for this issue: https://github.com/apache/spark/pull/18503 > UnsafeRow.hashCode assertion when sizeInBytes not multiple of 8 > --- > > Key: SPARK-21271 > URL: https://issues.apache.org/jira/browse/SPARK-21271 > Project: Spark > Issue Type: Bug > Components: SQL >Affects Versions: 2.2.0 >Reporter: Bogdan Raducanu > > The method is: > {code} > public int hashCode() { > return Murmur3_x86_32.hashUnsafeWords(baseObject, baseOffset, > sizeInBytes, 42); > } > {code} > but sizeInBytes is not always a multiple of 8 (in which case hashUnsafeWords > throws assertion) - for example here: > {code}FixedLengthRowBasedKeyValueBatch.appendRow{code} > The fix could be to use hashUnsafeBytes or to use hashUnsafeWords but on a > prefix that is multiple of 8. -- This message was sent by Atlassian JIRA (v6.4.14#64029) - To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org