I encountered the following problem when using a filter on a clientdataset.
Wenn I use a filter like "(group = 1) AND (comment = 'something')", a
record is still in the resultset when at least one of the fields value is
NULL. The problem does not occur when the filter has only one expression
like "(group = 1)" or "(comment = 'something')".
I can avoid the problem when I set the filter to "(group = 1) AND (comment
= 'something') AND (comment IS NOT NULL) AND (group IS NOT NULL)". But I
think this can't be the way it's meant to be.
I found the problem in D6.02, D7.11 and D2005.3. Is there a working
workaround? Is it a known problem?
Thanks in advance and
greetings from germany,
Daniel
This is a known problem (QC 13201). Drop your votes in there to increase
chances of getting fixed. As a work around do what you already
suggested. Be explicit regarding NULL values in your fiter expression.
> This is a known problem (QC 13201). Drop your votes in there to increase
> chances of getting fixed. As a work around do what you already
> suggested. Be explicit regarding NULL values in your fiter expression.
Thanks Kostas.
I dropped my vote. Hey, now there are already 2 votes for fixing the bug on
this basic feature. Does no one else need this fixed or are they only too
lazy to vote? ;)
The bug really occurs since D6. Unbelievable that it is still unfixed.
Daniel
Look Daniel, we have a maximum of 5 votes IIRC and I didn't drop my vote
on this because there is a work around already there (even if not so
elegant). There are other midas bugs that are older, open (and more
difficult to fix AFAIK) and there is no way around them (such as 11761/
5603. If one is is trying to reduce network traffic taking advantage of
poPropogateChanges, this bug corrupts M/D relationships and the
developer might not even notice this until deploying his application
(too late). The only way around this bug is to redesign the app and
Refresh data after applying updates loosing any advantage of
poPropagateChanges).