Message72749
The intention is that different interpreters have distinct, separate
instances of the type objects. This is desirable in case of class
variables, and standard for Python-defined types. Clearly, 3.0 won't
provide that separation for many of the builtin types (and never may);
for out-of-core types, it is really the choice of the module designer.
I forgot to implement PyType_Copy; it is meant to create a new type
object which is a field-by-field copy. It's now too late to add it to
3.0; extension authors requiring it now should include it's source code
(which is currently not even written - contributions are welcome).
Reconsidering it: what should it do to the base types?
a) just incref tp_base and clone tp_bases,
b) offer an optional parameter to specify an alternative base object,
c) offer an optional parameter to specify an alternative bases tuple. |
|
Date |
User |
Action |
Args |
2008-09-07 19:28:35 | loewis | set | recipients:
+ loewis, _doublep |
2008-09-07 19:28:35 | loewis | set | messageid: <1220815715.73.0.472406288146.issue3760@psf.upfronthosting.co.za> |
2008-09-07 19:28:35 | loewis | link | issue3760 messages |
2008-09-07 19:28:34 | loewis | create | |
|