Wwe could easily measure what the cost of testing attribute types at
runtime is and if it's not too bad doing it in the debug build seems
like a good idea.
Phillip J. Eby wrote:
At 10:28 AM 7/20/2005 -0700, Andi Vajda wrote:
Of course, the type check could happen at set time, at the cost of
performance. It's a trade-off I made a long time ago and that could
certainly be revisited. I could also add a 'debug' mode that enforces
type upon setting.
I suppose it would improve performance if an attribute is repeatedly
set to different values before a commit. But it would be interesting
to know how often that occurs along a performance-critical path. It
seems to me that most data-intensive tasks that actually *change* item
attributes are import and synchronization tasks (downloading mail,
sharing, etc.) which are highly unlikely to modify a given item's
value attributes more than once.
So, it may be that if "eager" typechecking were just done for value
attributes, it would have no net performance impact in practice,
because only one check would be occurring whether it's done at
assignment time or commit time.
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/dev
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "Dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/dev
