Summary:Review of Gabriels proposal, assoications are reversed
jgarnett just got back from coffee
jgarnett I got to work on some other stuff today
jgarnett but I wanted to walk you though how the "Why we don't want ft to extend at" discussion
jgarnett basically we are all saying the same thing. It was my fault for not renaming AttributeType when we changed its meaning yesterday.
jgarnett Does this chart make sense:
jgarnett Note: The goal here is for FeatureType to be a ComplexType and Feature to be Complex
jgarnett ( I think that is a goal you have in mind as well?)
<--| markhamc has left #geotools
jgarnett groldan ping?
-->| markhamc (email@example.com) has joined #geotools
jgarnett groldan - checking out the picutre ...
jgarnett Note this page was just about complex type support, not validation ... the next page is about schema etc ...
jgarnett But what the heck ...
jgarnett What tool do you use for diagrams
jgarnett Omondo of Viso
jgarnett or something more expensive?
jgarnett Not sure I know that
jgarnett So we have two different "models" going on
jgarnett (I no longer have getNumberOfAttributes for example
jgarnett and we have the association between schema and attribute reversed.
groldan no, we can get rid of it
groldan just want to check the idea
jgarnett Yep it is just going to take me a moment to catch up ...
groldan also I just reversed it, I had it the same as you, but want to discuss
jgarnett Q: Where is Complex Type ?
jgarnett I see Feature and Attribute
jgarnett but not "Complex"
groldan no longer needed?
jgarnett (or List or Map or Array depending on how you use it?)
groldan it would be just a tagging interface if it were there
jgarnett Okay I obviously have to go through this step by step ...
jgarnett ListAttributeType - can only have AtomicAttributeType ?
jgarnett No lists of complex type ?
groldan yes, by definition
jgarnett Okay - just checking.
groldan note ListAttType is NOT the same as SequenceAttType
groldan its the XMLSchema list type
jgarnett Gabriel do we have to explicitly model the difference? From the perspective of the programmer perhaps they should both look the same?
groldan both which?
jgarnett I am going to phrase this badly.
jgarnett When the programer has a list of stuff
jgarnett they don't much mind the difference between it being encoded with AtomicAttirbtueType (and thus ListAttributeType)
jgarnett or if it is done as a ComplexType with a sequence.
groldan ListType is defined to be a list of integers, or dates, etc. something you could do <myList>1 2 3 4</myList>
groldan or <myList values="1 2 3 4 5"/>
jgarnett Defined by GML - not buy us. GML is just one target, I am not sure someone working with a Database needs to care ...
jgarnett That is basically my question. Can we simplify this for the problem of modeling GIS data.
groldan defined by XML Schema, not GML. And yes, we can
jgarnett It is a tough problem, okay I am looking at FeatureAttributeType now ....
jgarnett Note FeatureCollection and Feature can also have a gml:srs attribute ... do you want to capture that ?
groldan finally I completely understood your idea of nodes and schemas (last night I was burned)
jgarnett okay cool, hey does that chart at the top of the page make sense?
jgarnett Thanks for putting all this effort into diagrams.
jgarnett I just turned your picture into a thumbnail (so the page rendered correctly)
groldan np, so Schema captures the validation rules of a "particle". choice/all/seq are particles, but not nodes, NodeSchema takes care of nodes. Nodes can be simple or complex, it doesnt matter
jgarnett I am working on it
jgarnett Should I look at Schema next then, I was looking at Attribute
jgarnett Okay that is making sense.
jgarnett Yeah I don't really know which way the association should go, I kind of like the association from Attribute to NodeSchema
jgarnett Um what about complex attributes ....
jgarnett we can have them right ?
groldan of cuorse
jgarnett Feature with in feature ... perhaps Attribute getSchema(): Schema ?
jgarnett ComplexSchema.getSchema() List<Schema>
jgarnett can we change the name to reflect the fact that it is going to be a list ?
groldan we have a hard time figuring out the direction of the association, kind of think that it could be either way, as long as it is respected for both FType and AttType
jgarnett I had the same structure as you and then changed it to be explicit for each of Choice, List and Set
jgarnett (although I think that was CHoice, Sequence and All
jgarnett Choice List and Set map better to Java types (if they are really what is intended).
groldan I see, but ComplexSchema as superinterface makes sense, since a feature could only have a complexschema, not a nodeschema (ie, it could not be an integer)
jgarnett I agree, on the association.
jgarnett There is a way to figure it out though. Once we have coded up the interfaces we can try coding against them - and whichever way is less of a pain ....
groldan yeah, choice, list and set maps well the concept
jgarnett I am still concerned about Complex being missing ... I would like to capture ISO 19115 Citation with this for example
jgarnett and that is not a Feature.
groldan will send you a xml and xsd doc I was trying against
groldan may help you to figure out that complexes are there
jgarnett Am I correct in thinking that the only thing that makes a Feature special with respect to a Complex is the idea of DefaultGeometry, getBounds, getID ?
groldan that's what makes a Feature different from a normal complex type, yes
groldan a normal complex type cant have id, in gml
groldan nor we can define it, since there not exists the concept of xml element attributes
jgarnett Q: Association between shcema and FeatureType - should that be the other way around ?
jgarnett and Feature pointing directly to FeatrueType ?
groldan just changed it for throwing up discussion
groldan I had it the other way
jgarnett fair enough, I am thinking about it both ways ...
groldan but who validates? the schema, right?
groldan one nice thing is that we no longer have the problem of Filter working against Features, that's what it does!
groldan we can reverse the association, but are in the same situation then for att
jgarnett I think I got why we switch between having these associations each way, it comes down to primary use. I think object s and xpath and you thing xml and validation
jgarnett good thing we are both working on this.
groldan yeah, though we need both worlds
jgarnett getRestriction is supposed to be able to work on individual attributes - including simple attribtues right ?
jgarnett I am thinking of length of a string for example.
groldan Jody: I REALLY have to go home, tomorrow is holiday and need to buy some things
groldan it is
jgarnett Okay sorry I did not know you were on a time limit.
jgarnett I have other work to do as well.
groldan no problem, I'll come tomorrow, but now I have to run away, sorry
jgarnett How would you like to proceed? If I know what your schedule and timeline I can plan.
jgarnett Can you send me an email when you get a chance.
jgarnett Thanks gabriel.
groldan my schedule is a bit constrained already, hopefully this work is so important that worhts the delay
groldan do you think we have enough to try some coding? or have still great concerns?
jgarnett We are still seering in opposite directions, but we are getting closer. I think one more session would give us enough to talk to RobA, Greg, Chris again
jgarnett We need to involve others, if we cannot explain it we should not do it.
jgarnett Do you code in Eclipse?
jgarnett I was thinking we could bang up the interfaces and use Omondo for digrams (it is bidirectional code to diagram)
jgarnett Thanks for the diagram I will keep thinking about it.
groldan yes, np, as long as the community edition works with eclipse 3.1 final
jgarnett yep, although you will have to turn subclipse off (if you use it)
jgarnett cause they disable it for "Team" projects.
groldan ok, I'm using command line svn anyway
jgarnett Okay I will see what we can do, when do you send your report to Rob ?
jgarnett For community buy in I think we will need to present a complete picture of interface, diagram and examples of shapefile and gml document being captured.
jgarnett Enjoy you holiday and thanks for the chat ....
groldan just sent you an example I was using to figure out the schema/type stuff
groldan may be of help
groldan ok, thanks for the chat
groldan good bye
groldan ah, and sorry for putting the diagram in the wrong place, was time constrained