Re: Should we wrap accessors for properties?

[Kevin Ryde <user42 zip com au>]

Torsten Schoenfeld <kaffeetisch gmx de> writes:
> I'm going through the new API in gtk+ 2.14.  As usual, there are quite a few
> accessors added for new properties.  For example,
> gtk_calendar_get_detail_height_rows and
> gtk_calendar_set_detail_height_rows for the "detail-height-rows" property of
> GtkCalendar.  I wonder if we should continue to wrap those.

They're pretty tedious.  They make more sense in C.  Starting from
scratch you'd be very tempted to leave them out ...

> The new properties are automatically available via the generic property
> mechanism ($object->set (foo => ...)

One difference is I think most of the set_foo() don't emit the notify
signal if the value is unchanged, whereas the generic set() emits
always.  A misfeature in that generic no doubt, but one that's probably
unfixable.

(Sloppy code which always does another set() in response to a notify or
toggled or whatnot has good scope for going into an infinite loop unless
there's someone in a cycle checking if the value is unchanged ...)


muppet <scott asofyet org> writes:
> In retrospect, it would've been cleaner to use a separate "properties"
> hash.

I guess a tied hash which was just property get/set could work.
(Somehow ties always seem more attractive in the idea than the actual
use.  I suppose the power comes when you plug it into another library
etc that's demanding a hash/array/whatever.)