Re: Record support and user-defined fields

[Benjamin Franksen <benjamin.franksen@bessy.de>]

Hi Kay,

I hope you don't mind I post this on the list.

On Thursday 30 June 2005 17:47, you wrote:
> On Jun 30, 2005, at 07:35, Benjamin Franksen wrote:
> > I don't yet see how we specify when 'user-defined fields' get
> > processed.
>
> My proposal allows them to hook into the record's processing
> at two points:
> After getting the value (so the hook can modify the data),
> and at the very end (so the hook can add averaging or FLNKs or ...).
> That handles all the test cases which we had at the SLAC meeting.

Ok, two seems to be a rather modest list. Anyway:

o what if there is no 'value' (see a recent posting on core-talk, I 
believe it was from Andrew)
o what does smoothing do if the value exists but is an array. Or an 
integral type. Or a string?

> > Thus, I argue for abandoning the idea to have per-instance defined
> > fields. Instead, ..... create a new record type.
>
> Sounds great. Can you create a wiki page with more detail
> as to how the dbd & db file would look as well as the API
> for the process()-pieces that each record-snipped needs to provide?

In a nutshell: dbd changes: none (I'll leave type variables out for the 
moment); API: depends on the function block/struct; DA implementation: 
exactly as for the 'fields' view for records. I'll try to present an 
example on the wiki.

Ben