[ome-users] array types?

[Ilya Goldberg]

So far, we've simply used instances of STs for array elements.  For 
example, the plane/stack statistics calculators declare several STs as 
outputs, and emit a list of them when executed (one ST instance per 
plane/stack in the image).
In this way, each array element becomes a row in the DB, obeying "first 
normal form".

There is of course the possibility for violating this, and we do just 
that when storing a serialized Matlab B-Net object.  This object is 
"atomic" in the sense that its components don't make any sense 
independently.  This is not strictly true, and depends on how you slice 
things.  The disadvantage of doing this is that only specialized 
viewers and analysis modules that recognize this object for what it is 
can do anything intelligent with it because there is no fall-back to 
well-recognized primitive datatypes (i.e. integer, float, date, etc).  
The object is stored as a string because we don't support a binary 
type.

The other obvious violation is the list of image importers in the 
configuration table, though these are not STs.

Strictly speaking, there is no direct support for arrays in Postgres 
either - its merely a convention that relies on a delimiter between 
elements combined with some helper functions to manipulate and query 
these easier.
So, the "standard element delimiter" route can be followed with its 
inherent disadvantages.
In most cases though, first normal form is really the best option in 
the end.  Millions and millions of rows are really not a problem.

-Ilya

On May 5, 2005, at 6:49 PM, kristian kvilekval wrote:

>
> I am in the process of integrating several
> Analysis modules.  However the output the analysis
> is an array of values, sometime quite large.
>
> I noticed that postgres supports array column types,
> however there doesn't seem to be way to express
> this in STD.xsd <element>
>
> Any idea if this would be useful and what it would take
> to actually implement it  or is there a simple work
> around that I am missing.
>
> Thanks
> kris
>
>
>
> -- 
> Kristian Kvilekval
> kris at cs.ucsb.edu  http://www.cs.ucsb.edu/~kris w:805-893-2526 
> h:504-9756
>
> _______________________________________________
> ome-users mailing list
> ome-users at lists.openmicroscopy.org.uk
> http://lists.openmicroscopy.org.uk/mailman/listinfo/ome-users
>