[storm] proposal for improvement

[Thomas K.]

Hi Gustavo,

> What about having a "schema" attribute which takes an arbitrary annotation?
> 
> Something like Int("name", primary=True, schema=...)
> 
> Would that be useful at all?
> 

Why not? It brings a connection between necessary definitions for 
querying tables by storm and attributes, which are necessary for 
creating and modifying the defined table. But that's the first step.

The next step is now to define, what's the content of this attribut 
"schema". Because it's a named attribute it should have a "defined 
type", means that it would be a good idea to define a class, which holds 
informations for "schema". (to hold this "type" open, means it's a 
"personal" flavour what's assigned, is not a good idea. You will get 
many different solutions with different interfaces, which are not 
compatible together) Informations could mean something like zope.schema 
(hi Stephen! ;-) ) for the connection to an middleware provider. (that's 
what zope.schema provides in my opinion) But for creating or modifying 
tables it's necessary to define for example IsNULL, size, defaults and 
many other values. The problem here is to give a possibility to store 
informations for __any specialized value__, which is maybe used only by 
one database adapter. (which is possible now or in future)

So we need ideas to solve this: On one hand to store informations, which 
are needed by creating tables or modifying tables (maybe later, maybe 
never, this is a really complex area) and on the other hand to give a 
connection to, for example, middleware. For me, the first part is 
necessary, the second could be a solution for others and so also necessary.

And a last question is to answer: what, if a column dosn't define this 
schema attribute: create (for example) table with defaults or do not 
create table at all because of missing schema attribute.

So, this are my "2 cents" for the moment.

cu, Thomas