JDBC FeatureWriter remove skips

Sean Geoghegan added a comment - 29/Dec/03 7:22 PM This is a weird one since it works fine with Oracle, however they both use exactly the same code. Jody have you confirmed that Oracle works okay for deletes? Im sort of worried it could be a difference in the drivers. Maybe we need a couple of result set level tests to ensure that both drivers behave the same. If you guys need more help with this I will be back at work on Tuesday (6th Jan).

Jody Garnett added a comment - 20/Jan/04 4:14 AM EXTERNAL MESSAGE: SUBJECT: Re: [jira] Commented: ( GEOT-75 ) JDBC FeatureWriter remove skips jira@codehaus.org wrote: >The following comment has been added to this issue: > > Author: Sean Geoghegan > Created: Mon, 29 Dec 2003 7:22 PM > Body: >This is a weird one since it works fine with Oracle, however they both use exactly the same code. Jody have you confirmed that Oracle works okay for deletes? > > No I have not confirmed this yet, I will put it on my todo list though. Just got back from a break. >Im sort of worried it could be a difference in the drivers. Maybe we need a couple of result set level tests to ensure that both drivers behave the same. > > Oh joy, write once test everywhere >If you guys need more help with this I will be back at work on Tuesday (6th Jan). > > That is cool, I will try and confirm it before then :-)

Jody Garnett added a comment - 05/Feb/04 3:00 PM This is blocking developement on Strong Transaction Support. Do we have any idea what is wrong with this one? Last we let off I think it was stuck behind QueryData next(). TO be clear this only happens with the FeatureType is in the following form: attrb1, attrb2, ..., GeometryAttrib, attrb N; Everything works fine for: attrb1, ... attrb N, GeomtryAttrib; That is to say this occurs when we have several AttributeIO spans in the same FeatureRead. This is easy to force by placing a geometry column in the middle of your attributes (usually postgis appends Geometry columns at the end so this bug does not appear). A couple of tests to try: Hand construct a Feature with three AttributeTypeIO spans (rather than two as tested) and see if we skip every three rows. Also note: the exact same code seems to work on Oracle, so we may be in some kind of differences between JDBC drivers here, or we may not have performed the same test (multiple attributeIO spans) in Oracle. We may need to assign this to someone other than Chris, as progress is needed, and I think he is short of time these days? Comments (and Ideas) please.

Chris Holmes added a comment - 05/Feb/04 3:14 PM EXTERNAL MESSAGE: SUBJECT: Re: [jira] Updated: ( GEOT-75 ) JDBC FeatureWriter remove skips Yeah, I'm a bit short on time, best to reassign it. If it only appears with geometry rows in odd places then it seems like it would have to be a problem with the calling of next. --

Jody Garnett added a comment - 05/Feb/04 4:21 PM I need to be able to trust the FeatureWrtier before I can proceed. Please note that having the FeatureWriter broken on remove - is really bad. The use of an optimized FeatureSource has hidden this bug from GeoServer - but we cannot pretend that the PostgisDataStore is working at the moment. If a user tries to use the low-level api (or even the non optimized FeatureSource) they will run into our inconsistent FeatureWriter.

Andrea Aime added a comment - 24/Apr/04 11:02 AM This bug has been solved in fid_exp. It is due to a bug into the jdbc driver I think, and I've solved by opening the result set in read only mode and creating explicit queries for delete and update (explicit ones for insert where already there)

Jody Garnett added a comment - 04/Oct/04 4:17 PM Andrea's fid-exp fixed this ages ago.