Skip to end of metadata
Go to start of metadata

Agenda: 1) hsql 2) spatial db in a box 3) snapshot vs maven 4) davidB 5) Factory vs Query 6) postgis fid update

jgarnett Hey guys I would like to set up an agenda if that is allright ..
dblasby its easier to make higher quality data pretty because it usually has more info in it that you can use to make it pretty
jgarnett anything people want to talk about?
1) hsql
2) spatial db in a box
polio 3) Martin Desruisseaux
jgarnett 3) I was going to say GeoAPI snapshot update process - but that will suffice
jiki btw I am a very curious guy, am I witnessing an open source development meeting now?
jgarnett Yep
I was going to start by thanking everyone for all their hard work switching CRS implementations.
polio 4) access for daveB?
jgarnett Everything is getting better everyday.
Ah good one polio.
jgarnett 5) CoordinateSequence for Query
bleck I mean CoordinateSequenceFactory?
cholmes 6) fill me in on postgis fid details - I realized I have patches waiting, on somewhat related stuff, that I want to integrate today
jgarnett Cool. Shall we start?
1) hsql
David Blasby has been a source of inspriration and I would like to start trying to set up a hsql datastore.
cholmes But we need to spatialize it first, no?
jgarnett I would like to make him a co-module maintainer (but he does not have commit access).
See 2)
I would like to start a ext/hsql module if I may.
cholmes How about you guys actually do the work first (wink)
jgarnett And tag david with the other stuff when he passes a code review.
+1 (or is it not cool of me to vote on my own suggestion)
polio +1
cholmes Oh yeah, an official introduction is in order, though most people already know him.
David Blasby was just hired by The Open Planning Project
jgarnett Do we have enough module maintainers to vote on this ?
cholmes to take over my job as lead developer of GeoServer
jgarnett the next meeting item is for him, lets hold on a sec.
jgarnett Hi James
jmacgill Hi, sorry I'm late
jgarnett we are voting on a ext/hsql datastore.
jiki David Blasby is the guy behind postgresql's geo stuff?
jgarnett We have +3 and no negative.
Yeah that dave guy gets around.
So I will consider that carried, and can work on it as the meeting goes on.
2) spatial db in a box
cholmes That's enough, go and set it up.
jmacgill +1 from me anyway
dblasby Its a generic method for adding spatial datatypes to databases
Its based on using JTS as the backend "smarts"
I have bindings for derby (cloudbase) and postgresql
theres an article there on how to compile jts so you can run it natively
w/o a jvm
It works well enough that it should be possible to add a geotools Datastore proxy to UMN Mapserver
but it would be huge
i'm currently working on adding hsql bindings
jmacgill so you have a spatial cloudscape already?
dblasby most of the code for the individual databases is auto-generated
cholmes Yeah, at this point ogr still is far better than geotools, so there's not much gain in a geotools datastore.
We don't really support any data formats better than them yet.
dblasby there's some problems with cloudscape - it allows custom functions but not custom types
nor can you have function that work off non-fixed length datatypes
chris - then add a ogr datastore to geotools?
using JNI?
jmacgill it has been discussed in the past - their may even be a JIRA issue open against that
jgarnett chris holmes what is OGR?
cholmes Yeah, I think Frank Wamerdam voluntered on the openSDI list to do that, or at least help out.
I think that's a much higher priority. Would be even better to get gdal - but I think we need to get our raster handling a bit better first.
What mapserver uses to get its vector data.
jgarnett cool thanks - I am out of the map server loop.
BTW package all the contents have the wrong package?
CIA-11 dzwiers * r11388 geotools/gt/module/main/src/org/geotools/geometry/ JTS update ... new env
jgarnett 9package
jmacgill not again, I though I fixed that
how on earth are they compiling?
CIA-11 dzwiers * r11389 geotools/gt/module/main/src/org/geotools/geometry/ JTS update ... new env
jgarnett So david we are a bit off topic - what you got is very cool. And will help expand what SFS can do.
Can I get the hsql bindings generated from you?
And do you mind the generated code being placed under the geotools copywrite in that datastore?
(Or will you produce a jar I can make use of?)
Or do you need to talk to chris first?
Ohh and can you use this to make SFS for Oracle (smile)
cholmes Yeah, I think we should make more of an effort to bring the C and Java open source geo worlds together, as we can certainly learn from one another's experiences.
Wouldn't you need access to the source code for SFS for oracle?
jgarnett well I think david could generate bindings to compiled JTS for oracle.
Oracle lets you add C does it not?
cholmes I think generated code can definitely be contributed to geotools. But shouldn't we try to get these incorporated into the actual codebases of hsql, hibernate, ect.?
jmacgill I thought oracle let you add Java...?
jgarnett Note this is not oracle spatial - it would be a new thing.
dblasby oracle allows for both C and java extensions
jgarnett Well I think that may take a while, hsql for instance publishes exactly how to extend it use for your own application.
Which is what I woudl be doing for the datastore.
And ifyou use that SDO class from oracle datastore you could use SFS functions such as outline on oracle spatial data directly.
cholmes Cool, sounds good.
jgarnett dblasby did you get a hsql to work against yet?
ie did any of the versions you download actually work?
dblasby hsql - I download and - both appear to setup a database, but it appears they changed the default user/password so I canot login...
jgarnett I see - so we are stuck for a bit.
jmacgill note: JIRA task for OGR exists as :
jgarnett Well I can start the work against then and we will figure out username/password at some point.
Can I go on with the meeting?
cholmes (warning, I may crash my browser, I have to test something)
jgarnett 3) I was going to say GeoAPI snapshot update process
We have had the build taken out a couple times last week.
And I wanted to ask martin to run maven build before updating the snapshot.
polio I though we went over this a couple months ago ...
jgarnett We did. I don't know why he does not run maven (sad)
jmacgill Yes, so I have set up a new procedure...
jgarnett Oh good.
jmacgill SNAPSHOT is no longer used
there is now a dated dep
jgarnett Can we still get our email to the list when a new dated geoapi is available?
jmacgill it is easy to update as it is only in the top level project.xml file
polio yes, but when this last came up you also asked for an email to the list with a set of changes ... our first notification seems to be a failed build report
jmacgill i.e DO NOT ADD A DEP ON GEOAPI to any plugin or extention
(or any module for that matter)
jgarnett Cool - I was still worried that martin was not running maven build. I would like to see the snapshot tested before the toplevel project.xml file is commited.
But yeah that change sounds good james.
jmacgill are we off JTS-1.6-SNAPSHOT now? Did I hear that it was out...?
jgarnett I have not seen the email about that?
polio can we ask for a review for each new geoapi jar just like any other change to the api, my argument for this is the effective changes are the same
jgarnett apparently we are no longer using the snapshot - I just did a clean maven build and geoapi-snapshot was not downloaded.
jeichar YOu're right James. JTS-16 is out. I'll update the xml files today.
jgarnett So I think your system is working james, I just did not see the change in policy actually go out.
polio that's not the point jody ... the api still changes whether the names are the same of different
jmacgill indeed, its just a bit clearer that something has happend - i.e. the project.xml will change.
It might be time to setup CruiseControl
I've done it a few times for inhouse projects
jgarnett I tried last year, but yeah cruisecontrol is great.
polio or we could enforce the coding standards ... run maven before commiting ...
jmacgill we have been trying to do that for a while.
CIA-11 jgarnett * r11390 geotools/gt/ext/hsql/: Initial setup of hsql datastore
jmacgill there was a time when maven was a major pain to use, now I think it is fairly easy to use at least as a final test
jgarnett agreed. cruise control should only be a fallback of last resort.
polio yes, I went though and maven it run headless, so if you 'nice' it then your machine is usable
jgarnett Okay next agenda item ...
4) access for daveB?
jmacgill +10
polio +1
jgarnett Hold on we have a procedure.
cholmes I think we should wait until he actually writes some code...
jgarnett Get chris to do a code review ...
jmacgill sorry, yes.
jgarnett lol
polio oh, yah (think some of us have seen his code before ... )
jgarnett however we can do the staged please only work in ext/hsql until chris things you are safe.
Now just because he hacks fast does no mean he is incapable of playing nice and writing his test cases.
cholmes Well, we need geotools code - to make sure he follows conventions and all, jalopies the code, has the setting right, good comments, ect.
jgarnett chances are he is coding write now and ignoring this meeting.
cholmes Yeah, if it's easier I'm fine for +1 on ext/hsql access.
CIA-11 rgould * r11391 udig/plugins/net.refractions.udig.printing.ui/plugin.xml:
Fixed page loading by re-writing the extension point processor. This required
exposing PageImpl's and MapImpl's zero-argument constructor. The extension point
argument must now point to the concrete class, not an interface.
dblasby lol - no, I'm trying to figure out how to connect to hsql
CIA-11 rgould * r11392 udig/plugins/net.refractions.udig.project/src/net/refractions/udig/project/impl/
Fixed page loading by re-writing the extension point processor. This required
exposing PageImpl's and MapImpl's zero-argument constructor. The extension point
argument must now point to the concrete class, not an interface.
rgould * r11393 udig/plugins/net.refractions.udig.project.ui/ (2 files in 2 dirs):
Fixed page loading by re-writing the extension point processor. This required
exposing PageImpl's and MapImpl's zero-argument constructor. The extension point
argument must now point to the concrete class, not an interface.
rgould * r11394 udig/plugins/net.refractions.udig.printing.model/src/net/refractions/udig/printing/model/impl/
Fixed page loading by re-writing the extension point processor. This required
exposing PageImpl's and MapImpl's zero-argument constructor. The extension point
argument must now point to the concrete class, not an interface.
jgarnett I will ask him to review/patch the SFSQL Generator for hsql and chris can review that.
jgarnett rgould is supposed to be sick today (must be working from home)
bad richard, rest if you are sick.
+1 in hsql access
jmacgill sorry, lost connection for a bit
cholmes He could you refractions guys bug paul about geoserver on the geotools svn, and the geoserver commits list - the last one is annoying me, since I can no longer see who else is doing geoserver work, and I know gabriel has done some lately.
jgarnett confused? what are you talking about chris?
the lists stopped working? or something?
cholmes two things
1) geoserver svn is here:
It should be here:
or really here:
polio cholmes - will bug the sys admin
cholmes Paul said he'd move it after a month, if everything was in order.
2) This list: is no longer updated
Because of the move to svn.
jgarnett dblasby do you have an offical topp email for the project.xml file?
cholmes I think we tried to set it up to send commits to the list, but it didn't work.
jgarnett dzwiers went to get the sys admin.
jgarnett thanks
jgarnett Shall we move on ...
5) CoordinateSequence for Query
jeichar and Martin this gets into the getting Factories where you need them.
The nested constructor idea woudl involve Query providing a FeatureTypeFactory for use by the datastore (allowing the Query to mandiate everything from CRS to GeometryFactory)
But the Query.setFactory approach may be more use friendly.
If possible I would like to focus on GeometryFactory/CoordinateSequenceFactory rather then CRS FactoryFinder - because I think it is a simple case.
polio Chris, Dave will be done soon
dave you should be added today
dblasby thank-you
polio Chris, he will contact you
cholmes We really want that kind of information in query? It seems to me more how the datastore is working, instead of what you are requesting from it...
jeichar Lite Renderer2 would really like the ability to set the type of coordinate sequence factory.
jgarnett And andrea wanted the ability to "force" the CRS of a FeatureStore for about a year now.
jeichar Mainly so that, during reprojections, new objects do not need to be created.
cholmes Yeah, I actually disagree with that in Query, myself.
jgarnett We have the design (with all the factories) it is the application that is proving difficult.
cholmes But I need to write up my thoughts as to why, and to propose an alternative.
jgarnett It could be a Transaction hint chris.
That even makes sense.
jeichar - does that make sense for LiteRenderer?
Oh wait you may have more then one thread using the same transaction...
cholmes Yeah, as a hint makes more sense to me.
jgarnett Query hint then?
cholmes Yeah, I was thinking of something along those lines - do we have a Query hint yet?
jeichar I don't like a hint in transaction really. more in query
I don't mind using hints though.
jgarnett It is odd because the factories are design with these uses in mind, we just never buttoned down a formal way to get them where they are needed.
Instead we work against a bunch of FacotryX.getDefault() most of the time.
cholmes A query hint would also be nice to add new functionality that all datastores don't support right away - I think anything in Query itself should always work, it just may require us to do it all in java, and thus a bit slower.
(yes, I know we want coordinate stuff in all datastores, I'm just saying)
jgarnett I like the idea of hints being a bit informal (maybe not always supported).
dblasby java isnt always slow)
jgarnett The factory idea is core to geotools - I would not mind the support of them being made a bit stronger then just a hint.
But yeah lets start with a hint and see how things work out.
cholmes well, it is when the choices are between loading everything from the db into java across jdbc vs. having the database itself figure out what to return.
but no, java isn't always slow.
+1 on hint
jeichar +1 to hint.
jgarnett +1 hint
jmacgill +1 hint
dblasby okay - I know what I mean, chris.
jgarnett (although I think we may revisit the issue later) Thanks for putting up with my long email everyone.
6) fill me in on postgis fid details - I realized I have patches waiting, on somewhat related stuff, that I want to integrate today
me = cholmes
Well jgarnett != cholmes but me in the subject = cholmes
jeichar I'm just tracking down the change I made to make it work.
jgarnett Last week was so long ago.
cholmes Yeah, just curious what you guys did, as I realized I have patches to apply, that may be similar. Sorry I forgot about them when you guys were working on it.
jgarnett In that case we may have just duplicated work - Postgis was not generating its FIDs in any sane manner.
Their was a class their that did FID generation but it was not being used.
And the create loop did not bother to return the FID even when it was known.
A simple mistake really .. but one that cost brain cells to debug.
jeichar That's more or less it.
CIA-11 dzwiers * r11395 geotools/gt/ (4 files in 4 dirs): JTS update ... new env
jeichar A FID was created but not set in the feature so it was lost immediately after creation.
cholmes Do you remember what class the fixes are in?
jeichar I'm trying to find it.
I'd love for you to check it.
cholmes (actually it looks like the fixes are related to createSchema, so it may be a different problem).
jgarnett While we wait dblasby I almost have the "empty" hsql datastore ready to commit.
cholmes Cool, I'd be happy to review.
polio ah, that was me
jesse and I added crs support to postgis
cholmes Successfully?
dblasby thanks jody
cholmes Where did you do the changes?
polio he made a crs factory thingy and I hooked it into postgis
jmacgill sorry, gotta run
polio that was the createschema changes ... it was just a couple lines that called the PostGisAuthorityFactory
that allowed the featureType to know it's projection
CIA-11 jgarnett * r11396 geotools/gt/ext/hsql/tests/: Removed file/folder
jgarnett * r11397 geotools/gt/ext/hsql/ (5 files in 5 dirs): Removed file/folder
cholmes Ok, cool.
jgarnett dblasby are you happy getting geotools and setting up a build env?
dblasby I was going to tackle that after the meeting
I'm in linux now and want to do that in windows
jgarnett I think that was the last meeting item ...
Are we done guys, back to normal development ...

  • No labels