[teiid-issues] [JBoss JIRA] (TEIID-5406) Allow usage of long datatype as the increment parameter in TimestampAdd function

Steven Hawkins (JIRA) issues at jboss.org
Fri Jul 6 08:16:01 EDT 2018


    [ https://issues.jboss.org/browse/TEIID-5406?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13601848#comment-13601848 ] 

Steven Hawkins commented on TEIID-5406:
---------------------------------------

It looks like there is some bad code handling the fractional add.  It will be easy to correct.  The expectation of an integer argument that is to match the expected behavior of most sources.  Converting to a long is possible, but it may require adding casts to existing push down logic.  Unfortunately there is no agreement on what fractional seconds mean - you can find millis, micros, and nanos.

> Allow usage of long datatype as the increment parameter in TimestampAdd function
> --------------------------------------------------------------------------------
>
>                 Key: TEIID-5406
>                 URL: https://issues.jboss.org/browse/TEIID-5406
>             Project: Teiid
>          Issue Type: Enhancement
>          Components: Query Engine
>    Affects Versions: 10.2
>         Environment: teiid-10.2.0 on WildFly Full 11.0.0.Final (WildFly Core 3.0.8.Final)
>            Reporter: dalex dalex
>            Assignee: Steven Hawkins
>             Fix For: 11.1
>
>
> Currently the TIMESTAMPADD(interval, count, timestamp) expects the count parameter to be integer. This is inconsistent with the return value of the TIMSTAMPDIFF function that is long. Thus we cannot take the fraction of the count returned by TIMSTAMPDIFF and submit it to TIMESTAMPADD. Since the SQL_TSI_FRAC_SECOND difference is too large for most human values (and it's the only way to retrieve millis, even though millions of millis), we need a way to submit this data to TIMSTAMPADD.
> 1. There is an inconsistency when adding 1 billion fractions of a second. Please execute the script below, one adds one billion minus one, the next one exactly 1bln fractions which results in two values, one of which is 1 fraction larger than expected:
> {code:sql}
> Select TimestampAdd(SQL_TSI_FRAC_SECOND, 999999999, Cast(CurDate() as timestamp))
> Union All
> Select TimestampAdd(SQL_TSI_FRAC_SECOND, 999999999 + 1, Cast(CurDate() as timestamp));;
> {code}
> 2. Another example:
> {code:sql}
> Begin
>     Declare timestamp initialDate = CurDate();
>     Declare long diff = TimestampDiff(SQL_TSI_FRAC_SECOND, initialDate, TimestampAdd(SQL_TSI_DAY, 1, initialDate));
>     Declare timestamp back = TimestampAdd(SQL_TSI_FRAC_SECOND, diff, initialDate);
>     select initialDate, back;
> End ;;
> {code}
> which leads to the following error message:
> {code:noformat}
> 2018-07-03 10:10:33,942 WARN  [org.teiid.PROCESSOR] (Worker7_QueryProcessorQueue47) 1RBg/ooz5vx7 TEIID30020 Processing exception for request 1RBg/ooz5vx7.27 'TEIID30070 The function
> 'TimestampAdd(SQL_TSI_FRAC_SECOND, VARIABLES.diff, VARIABLES.initialDate)' is a valid function form, but the arguments do not match a known type signature and cannot be converted usi
> ng implicit type conversions.'. Originally QueryResolverException ResolverVisitor.java:757.: org.teiid.api.exception.query.QueryResolverException: TEIID30070 The function 'TimestampA
> dd(SQL_TSI_FRAC_SECOND, VARIABLES.diff, VARIABLES.initialDate)' is a valid function form, but the arguments do not match a known type signature and cannot be converted using implicit
>  type conversions.
>         at org.teiid.query.resolver.util.ResolverVisitor.resolveFunction(ResolverVisitor.java:757)
>         at org.teiid.query.resolver.util.ResolverVisitor.visit(ResolverVisitor.java:392)
>         at org.teiid.query.sql.symbol.Function.acceptVisitor(Function.java:169)
>         at org.teiid.query.sql.navigator.AbstractNavigator.visitVisitor(AbstractNavigator.java:50)
>         at org.teiid.query.sql.navigator.PreOrPostOrderNavigator.postVisitVisitor(PreOrPostOrderNavigator.java:57)
>         at org.teiid.query.sql.navigator.PreOrPostOrderNavigator.visit(PreOrPostOrderNavigator.java:194)
>         at org.teiid.query.sql.symbol.Function.acceptVisitor(Function.java:169)
>         at org.teiid.query.resolver.util.ResolverVisitor.resolveLanguageObject(ResolverVisitor.java:1456)
>         at org.teiid.query.resolver.command.UpdateProcedureResolver.resolveStatement(UpdateProcedureResolver.java:234)
>         at org.teiid.query.resolver.command.UpdateProcedureResolver.resolveBlock(UpdateProcedureResolver.java:127)
>         at org.teiid.query.resolver.command.UpdateProcedureResolver.resolveCommand(UpdateProcedureResolver.java:110)
>         at org.teiid.query.resolver.QueryResolver.resolveCommand(QueryResolver.java:282)
>         at org.teiid.query.resolver.QueryResolver.resolveCommand(QueryResolver.java:128)
>         at org.teiid.dqp.internal.process.Request.resolveCommand(Request.java:282)
>         at org.teiid.dqp.internal.process.Request.generatePlan(Request.java:418)
>         at org.teiid.dqp.internal.process.Request.processRequest(Request.java:486)
>         at org.teiid.dqp.internal.process.RequestWorkItem.processNew(RequestWorkItem.java:660)
>         at org.teiid.dqp.internal.process.RequestWorkItem.process(RequestWorkItem.java:339)
>         at org.teiid.dqp.internal.process.AbstractWorkItem.run(AbstractWorkItem.java:47)
>         at org.teiid.dqp.internal.process.RequestWorkItem.run(RequestWorkItem.java:276)
>         at org.teiid.dqp.internal.process.DQPWorkContext.runInContext(DQPWorkContext.java:277)
>         at org.teiid.dqp.internal.process.ThreadReuseExecutor$RunnableWrapper.run(ThreadReuseExecutor.java:115)
>         at org.teiid.dqp.internal.process.ThreadReuseExecutor$3.run(ThreadReuseExecutor.java:206)
>         at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
>         at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
>         at java.lang.Thread.run(Thread.java:745)
> {code}



--
This message was sent by Atlassian JIRA
(v7.5.0#75005)


More information about the teiid-issues mailing list