I realized there was a huge gaping problem in my existing
hash table
code: it was impossible to implement any sort of iteration
interface
at all. A GC could occur at any time and reshuffle the
table
arbitrarily, invalidating any existing iterators. A
'for_each' style
interface like the one for environments is no better. Oops.
I've got a new hash table design, though, that should fix
this, save
memory over the current design, and make it easy to provide
a special
case for hash tables with few entries, if that becomes
desireable.
First, I'm going to implement the interface I've got now
(plus Karl's
suggested improvements), then I'm going to write a test
which catches
the GC-related failure, and then I'm going to re-do the hash
table
implementation in the newer, smaller, iterate-able way.
_______________________________________________
Minor mailing list
Minor red-bean.com
http:/
/www.red-bean.com/mailman/listinfo/minor
|