[jira] [Commented] (CASSANDRA-7499) Unable to update list element by index using CAS condition

["Tyler Hobbs (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/CASSANDRA-7499?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14094844#comment-14094844 ] 

Tyler Hobbs commented on CASSANDRA-7499:
----------------------------------------

I haven't completely reviewed this yet, but here are some initial review comments:

* {{StorageProxy.cas()}}: docstring needs to be updated
* {{UpdateParameters.setPrefectedLists()}}: unused and typo in method name
* {{BatchStatement.executeWithConditions()}}: 'timestamp' is unused and needs to be passed to the updates.  Test coverage for using the correct timestamps would be good.
* {{RowUpdate}}: comments explaining what's going on would be useful
* {{RowCondition}} constructor: since we weren't dividing {{now}} by 1000 before, TTL'ed cells were considered expired way too early during CAS reads? If that's true, can you create dtests for that scenario as well?  Can you add docstrings for {{RowCondition.now}} and {{Column.isLive()}} (at a minimum) to document when millis and micros are used?

As a side note, what do you think about pushing this fix to 2.1 and throwing an IRE in this situation in 2.0?  I'm a little nervous about making large changes to CAS this late in 2.0, and this hasn't worked in 2.0 at all so far.

> Unable to update list element by index using CAS condition
> ----------------------------------------------------------
>
>                 Key: CASSANDRA-7499
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7499
>             Project: Cassandra
>          Issue Type: Bug
>          Components: Core
>         Environment: Cassandra *2.0.9*, Java Driver *2.0.2* & *2.0.3*
> Client: cqlsh *3.1.8*, CQL spec *3.1.0*, Thrift protocol *19.39.0*
>            Reporter: DOAN DuyHai
>            Assignee: Sylvain Lebresne
>             Fix For: 2.0.10
>
>         Attachments: 7499.txt
>
>
> While running IT tests for *Achilles*, I ran into a strange bug:
> *With CQLSH*
> {code:sql}
> cqlsh:test> CREATE TABLE cas_update(id int PRIMARY KEY,name text,friends list<text>);
> cqlsh:test> INSERT INTO cas_update (id, name , friends ) VALUES ( 10,'John',['Paul','George']);
> cqlsh:test> SELECT * FROM cas_update WHERE id=10;
>  id | friends            | name
> ----+--------------------+------
>  10 | ['Paul', 'George'] | John
> cqlsh:test> UPDATE cas_update SET friends[0]='Helen' WHERE id=10 IF name='John';
> Bad Request: List index 0 out of bound, list has size 0
> cqlsh:test> UPDATE cas_update SET friends[0]='Helen' WHERE id=10;
> cqlsh:test> SELECT * FROM cas_update WHERE id=10;
>  id | friends             | name
> ----+---------------------+------
>  10 | ['Helen', 'George'] | John
> {code}
> It seems that we cannot update list element by index with a CAS condition.
> *With Java driver 2.0.2 or 2.0.3*
> {code:java}
>  ACHILLES_DML_STATEMENT@:writeDMLStatementLog Prepared statement : [INSERT INTO CompleteBean(id,followers,friends,name,preferences) VALUES (:id,:followers,:friends,:name,:preferences) USING TTL :ttl;] with CONSISTENCY LEVEL [ONE] 
>  ACHILLES_DML_STATEMENT@:writeDMLStatementLog    bound values : [621309709026375591, [], [Paul, Andrew], John, {}, 0] 
>  ACHILLES_DML_STATEMENT@:writeDMLStartBatch  
>  ACHILLES_DML_STATEMENT@:writeDMLStartBatch  
>  ACHILLES_DML_STATEMENT@:writeDMLStartBatch ****** BATCH UNLOGGED START ****** 
>  ACHILLES_DML_STATEMENT@:writeDMLStartBatch  
>  ACHILLES_DML_STATEMENT@:writeDMLStatementLog Parameterized statement : [UPDATE CompleteBean USING TTL 100 SET friends[0]=? WHERE id=621309709026375591 IF name=?;] with CONSISTENCY LEVEL [ONE] 
>  ACHILLES_DML_STATEMENT@:writeDMLStatementLog    bound values : [100, 0, Helen, 621309709026375591, John] 
>  ACHILLES_DML_STATEMENT@:writeDMLStatementLog Parameterized statement : [UPDATE CompleteBean USING TTL 100 SET friends[1]=null WHERE id=621309709026375591 IF name=?;] with CONSISTENCY LEVEL [ONE] 
>  ACHILLES_DML_STATEMENT@:writeDMLStatementLog    bound values : [100, 1, null, 621309709026375591, John] 
>  ACHILLES_DML_STATEMENT@:writeDMLEndBatch  
>  ACHILLES_DML_STATEMENT@:writeDMLEndBatch   ****** BATCH UNLOGGED END with CONSISTENCY LEVEL [DEFAULT] ****** 
>  ACHILLES_DML_STATEMENT@:writeDMLEndBatch  
>  ACHILLES_DML_STATEMENT@:writeDMLEndBatch  
>  ACHILLES_DML_STATEMENT@:truncateTable   Simple query : [TRUNCATE entity_with_enum] with CONSISTENCY LEVEL [ALL] 
>  ACHILLES_DML_STATEMENT@:truncateTable   Simple query : [TRUNCATE CompleteBean] with CONSISTENCY LEVEL [ALL] 
> com.datastax.driver.core.exceptions.InvalidQueryException: List index 0 out of bound, list has size 0
>         at com.datastax.driver.core.exceptions.InvalidQueryException.copy(InvalidQueryException.java:35)
>         at com.datastax.driver.core.DefaultResultSetFuture.extractCauseFromExecutionException(DefaultResultSetFuture.java:256)
>         at com.datastax.driver.core.DefaultResultSetFuture.getUninterruptibly(DefaultResultSetFuture.java:172)
>         at com.datastax.driver.core.AbstractSession.execute(AbstractSession.java:52)
> {code}
> With Cassandra *2.0.8* and Java Driver 2.0.2 or 2.0.3, *the test passed* so it seems that there is a regression somewhere in the CAS update code



--
This message was sent by Atlassian JIRA
(v6.2#6252)