-
15. Re: Multiple columns for the multisource discriminator column.
> Not a good way. The working deployment location is managed by JBoss and is not expected to be directly manipulated.
No worries. I was curiously mainly for a better testing experience. I'm pretty sure I can handle any additional security requirements through the translator.
Back to the original discussion, I wrote a quick test program that disentangles XML packed data in the source column. Teiid ends up querying both sources when only one would have data. In a different case, querying directly against the table using the packed source column also ends up invoking both sources.
I have attached the test program to this thread.
-
16. Re: Multiple columns for the multisource discriminator column.
The first set of issues is that the xml source names you have are invalid. If the execution actually returned rows and used select * against the veiw, you would see exceptions from the xpathValue function.
The next issue is that the criteria is not making it into the proper position for the multisource logic to have an effect. It is implicit that multisource required a translator that supported equality comparison at least. In this case adding the xpathValue function, because of weaknesses with the detection of a fully evaluatable, is then inhibiting the proper placement. That last one is easy to fix and I can ensure that more complex cases are covered as well.
-
17. Re: Multiple columns for the multisource discriminator column.
> That last one is easy to fix and I can ensure that more complex cases are covered as well.
Do you want me to add a jira for it?
-
18. Re: Multiple columns for the multisource discriminator column.
Yes, something along the lines of "ensure pushdown of multisource column expressions".
-
20. Re: Multiple columns for the multisource discriminator column.
With TEIID-2748 resolved your scenario with corrected source names should work against master.
-