Re: Datasheet internals

[Ben Pfaff]

John Darrington <address@hidden> writes:

> On Mon, Jun 11, 2007 at 08:14:59PM -0700, Ben Pfaff wrote:
>      John Darrington <address@hidden> writes:
>      
>      > Doesn't this mean that casereaders need to be reference counted?  I
>      > keep getting seg faults which turn out to be due to datasheet.c
>      > referencing casereaders which have been destroyed.
>      
>      There's a missing comment on datasheet_create.  Here's a fix.
>      Does it clarify the issue?
>
> I see.  So I can't pass the reader to proc_set_active_file_data.  I
> should use casereader_clone?

I think you and I might have different mental models of what's
going on.  Let me explain what I had in mind, and then you can
explain how your model is different, and then we can come to some
kind of meeting of the minds.

The model I was envisioning was this.  The GUI would interact
primarily with a datasheet and a dictionary.  The GUI-datasheet
shows what's in the struct datasheet, the dialog boxes work
mostly with the dictionary, and so on.  Then, when the user told
the GUI to do something that needed a procedure or other syntax
to be run, the datasheet would be destructively transformed into
a casereader with datasheet_make_reader, the casereader would be
put into a dataset, the syntax would be run, and then the
casereader would be extracted from the dataset and destructively
transformed into a struct datasheet, which becomes what the GUI
works with until the next bit of syntax needs to get run.

This model was born from some lack of knowledge of what the GUI
really wants.  Tell me, how can it be made more GUI-friendly?
-- 
Ben Pfaff 
http://benpfaff.org