Rob Atkinson ha scritto:
> Fantastic news.
> 
> I'll look to get the geometryless module into supported status around
> this - my only blocker at the moment is the lack of the functionality
> currently wrapped in sql-datastore  (under 2.4
> ./unsupported/community-schemas/sql-datastore)  in JDBC.  

I did look at the geometryless module some time ago and I was quite
surprised to see it put in the same class geoemtryless support
along with mapping... imho they should be separated.

Most JDBC based datastores nowadays do support tables without a
geometry in (thought probably the new one hasn't been checked for
this specific case, opened a jira to make sure we don't forget
http://jira.codehaus.org/browse/GEOT-2002).

The issue of mapping 4 fields into a bbox or 2 into a point is imho
unrelated to sql and should be handled in a separate wrapper datastore
that only concerns into mapping a geometryless feature type into another
that has geometries. A few examples of geometryless data sources out
of the jdbc land:
* csv/xms/generic flat table files
* a remote geometryless wfs type

I would better see a generic mapping datastore using generic mapping 
strategies, each strategy could be something something like (thinking as
I write):
- set of attributes in input
- set of attributes in output
- invertible flag (is it possible to go from outputs to inputs)?
- forward and backwards data transformation
- forward and backwards filter transformation (when possible)

This could handle everything from simple renames to computations to
lumping togheter a set of fields to build a new one.
Ok, here I'm going way off topic, if you want to discuss this further
please let's open a new mail thread.

> Gabriel did
> this work and I extended it to provide a little bit of dialect-flag
> driven handling of different date literal handling.  My hope is that
> both of these extensions will be rendered superfluous by the new
> prepared statement handling design and dialect handling. Certainly I
> only need the dialect handling for Oracle, so I'd strongly suggest a
> good set of unit tests around date, time, timezones.

The datastore structure provides a set of abstract conformance tests
that specific dialects implement in order to check their conformance
for a specific api or a specific topic. For example, I just recently
added two new base classes for testing geometry type binding and
spatial filters. You're more than welcomed to add date/time handling
tests as well.

> WARNING: There is a flag that Oracle can run under to determine the
> datatype of timestamps returned by JDBC - you need to handle both
> potential settings of this flag. [ -Doracle.jdbc.J2EE13Compliant=true]

What is this all about? A VM flag is scary, if this is useful,
is there any way to set it in the connection string instead?
Hum... yeah, it seems so:
http://www.oracle.com/technology/tech/java/sqlj_jdbc/htdocs/jdbc_faq.htm
thought the document does not make

> There is also a set of extensions under jdbc-regfunc which allow
> specific database functions to be exposed.

Due to vacations and the usual ton of stuff I have to attend to
I did not manage to follow this one. Is this about filters?

> IMHO These capabilities would be of general interest - there's many a
> potential JDBC datastore where you will want the ability to select
> specific columns, or do some joins across lookup tables.  We've been
> put off proposing them as fixes to JDBC pending Jody's view that H2
> work will supercede it.

This is native JDBC functionality and I would indeed like to see this
added to the JDBCDataStore, that is, stuff like:
- way to handle joins
- way to register a random sql query as a new (read only) feature type,
   eventually with parameters (think a dynamic query that extracts
   certain rows based on the current user, we could feed that thru the
   Query hints)

> Gabriel is familiar with the code, so would be the best person to
> comment specifically, but if people want us to try to  roll it in and
> create patches and proposal then we can schedule that - but IMHO it
> would be best if it were considered up front and the capabilities
> supported without the need for a patch.

If we could do planning with lots of time ahead yes, but our reality
is different and does not allow for this kind of planning. The first
cut of the H2 datastore was for the wfs 1.1 + xlink work, and had
certain time limitations.
This one is about adding Oracle support and has to be done within
2-3 weeks tops.
If you're lucky to have the necessary planning time, join, make
proposals, and we'll try to find the time to discuss them.

Cheers
Andrea

-------------------------------------------------------------------------
This SF.Net email is sponsored by the Moblin Your Move Developer's challenge
Build the coolest Linux based applications with Moblin SDK & win great prizes
Grand prize is a trip for two to an Open Source event anywhere in the world
http://moblin-contest.org/redirect.php?banner_id=100&url=/
_______________________________________________
Geotools-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/geotools-devel

Reply via email to