Re: [patch #5629] Make "struct variable" opaque: second half

[John Darrington]

On Sat, Dec 09, 2006 at 07:48:16PM -0800, Ben Pfaff wrote:

     > dictionary.c. */ :  You could enforce this with a couple of #defines and 
a
     > #error directive.
     
     I'd rather not, just because any such solution is only going to
     make things slightly harder to screw up.  I'd prefer to believe
     that programmers can read comments...

OK.

     > 0 I've never been particularly happy with the aux, aux_dtor and obs_vals
     > members.  They could be done, with a hash table keyed on the variable's
     > address.  
     
     That's my preferred solution too.  I just haven't implemented it
     yet.  I want it to properly support caching of data properties
     such as obs_vals.  This requires a mechanism to inform the hash
     table that data has changed, so that the cache must be
     invalidated, and a way for later procedure to discover the
     existence of the cache. 

OK.  This seems to be consistent with the gui's callback requirements
too.

The VAR_STRING / VAR_NUMERIC changes seem OK too (I haven't actually
tried the patch).

J'


-- 
PGP Public key ID: 1024D/2DE827B3 
fingerprint = 8797 A26D 0854 2EAB 0285  A290 8A67 719C 2DE8 27B3
See http://pgp.mit.edu or any PGP keyserver for public key.

pgpFHo0iNALhd.pgp
Description: PGP signature