[teiid-issues] [JBoss JIRA] Resolved: (TEIID-806) JDBC connector handling of convert is inconsistent

Mon Aug 31 17:47:23 EDT 2009

     [ https://jira.jboss.org/jira/browse/TEIID-806?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Steven Hawkins resolved TEIID-806.
----------------------------------

    Resolution: Done

Created a better base class (ConvertModifier) to represent the convert function.  Also ensured that type mappings more closely reflect Teiid type semantics.  For example integral types in oracle now use the trunc function to ensure that results are more consistent (this will need to be done more pervasively - for example oracle division of Teiid int types will actually be done as a number type and should also be converted/truncated).  Also fixed the use default decimal/varchar types that had default length/precision/scale insufficient to hold all possible values.  However to do this it is assumed that bigdecimal maps to decimal(38, 19) and biginteger maps to decimal(38, 0) - or whatever the maximum precision supported by the system is.  This is inconsistent with our internal type system, since we assume float/double are implicitly convertable to bigdecimal and biginteger is implicitly convertable to bigdecimal.  This should be reexamined by TEIID-65/TEIID-7. 

> JDBC connector handling of convert is inconsistent
> --------------------------------------------------
>
>                 Key: TEIID-806
>                 URL: https://jira.jboss.org/jira/browse/TEIID-806
>             Project: Teiid
>          Issue Type: Quality Risk
>          Components: JDBC Connector
>    Affects Versions: 6.0.0
>            Reporter: Steven Hawkins
>            Assignee: Steven Hawkins
>             Fix For: 6.2.0
>
>
> If a source lacks the exact representation of a Teiid type we approximate it with the closest source type.  Pushed converts will typically not have the same affect as if they are evaluated in the engine.  For example sources lacking byte or tinyint types are having conversions dropped rather, but the actual effect should be modulo the max byte value.  We also are not ensuring that the truncation or rounding behavior is the same for all conversions - such as from a floating point number to a fixed point.

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: https://jira.jboss.org/jira/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira