<<O>>  Difference Topic AbstractModel (r1.2 - 13 Apr 2005 - Main.apache)

The Abstract Data Model

I'd like to use this topic to discuss the abstract data model, as suggested by JohnCaron. I'll try to get some UML here for what I think the ADM should be.

Line: 34 to 34

-- JohnCaron - 16 Oct 2003

Deleted:
<
<

Well, this is a bit annoying; I'm testing out RCS on the new host.

JamesGallagher

Added more

JamesGallagher


META FILEATTACHMENT DAP_4_ADM.png attr="" comment="" date="1065217793" path="DAP_4_ADM.png" size="6395" user="JamesGallagher" version="1.2"
META FILEATTACHMENT N1.png attr="" comment="NetCDF object model" date="1065222835" path="C:\dev\netcdf-java-new\doc\N1.png" size="14353" user="JohnCaron" version="1.1"
META FILEATTACHMENT N4.png attr="" comment="NetCDF-4 (ver2) UML" date="1066167761" path="C:\dev\netcdf-java-new\doc\N4.png" size="20924" user="JohnCaron" version="1.1"

 <<O>>  Difference Topic AbstractModel (r1.1 - 13 Apr 2005 - Main.apache)
Line: 1 to 1
Added:
>
>

The Abstract Data Model

I'd like to use this topic to discuss the abstract data model, as suggested by JohnCaron. I'll try to get some UML here for what I think the ADM should be.

John (and anyone else), can you add UML for other data models like netCDF and HDF5? What I'm concerened about is making sure that our model has a way to represent the concepts present in HDF4/5 and netCDF resonably well. I'd still like to pursue the goal of having HDF4 and HDF5 client libraries...

-- JamesGallagher - 03 Oct 2003

  • Here is one view of the DAP Abstract Data Model:
    DAP_4_ADM.png

  • Comments:
    • I used aggregation for Value because it seems those have a lifetime that is, in the abstract sense, different from the lifetime of a variable in the DAP. I used 0..* because there are times when a variable has no value.
    • I wonder about making Attribute use Type...
    • I'm not totally comfortable with UML, so make any changes you want. I used Poseidon to draw the figures; I can put the stuff in CVS if you'd like.


Ok, Im attaching a recent UML diagram of both the current model and a draft of a new model as we merge/munge/mangle with HDF5.

NetCDF current model

N1.png

Draft "NetCDF-4" model

N4.png

-- JohnCaron - 03 Oct 2003

Sequences and Selection

I see sequences as a nice addition that will allow variable length arrays to be effiently represented. Allowing them to be selected in all cases seems to me to be problematic. If I have a large sequence, i cannot efficiently implement the equivilent of a SQL WHERE clause, unless i have built an index. It seems to me that this ability is an addditional capability similar to a function, which should only be allowable if its advertised in the capability object.

-- JohnCaron - 16 Oct 2003

Well, this is a bit annoying; I'm testing out RCS on the new host.

JamesGallagher

Added more

JamesGallagher

META FILEATTACHMENT DAP_4_ADM.png attr="" comment="" date="1065217793" path="DAP_4_ADM.png" size="6395" user="JamesGallagher" version="1.2"
META FILEATTACHMENT N1.png attr="" comment="NetCDF object model" date="1065222835" path="C:\dev\netcdf-java-new\doc\N1.png" size="14353" user="JohnCaron" version="1.1"
META FILEATTACHMENT N4.png attr="" comment="NetCDF-4 (ver2) UML" date="1066167761" path="C:\dev\netcdf-java-new\doc\N4.png" size="20924" user="JohnCaron" version="1.1"