[Shootout-list] A few unrelated points
Bengt Kleberg
bengt.kleberg@ericsson.com
Mon, 27 Sep 2004 10:14:26 +0200
Isaac Gouy wrote:
...deleted
> 2) We seem to be intent on removing the hash test because more time is
> spent creating hash-keys than doing insert/lookup.
>
> Does that mean much more than hash is a test of both number-to-string
> conversion and hashtable insert/lookup - just as spell checker combines
> line-oriented I/O and hashtable insert/lookup?
since i was the person (please correct me if i am wrong here) to suggest
that we should go from the ''this test is meaningless'' to actually
remove it, i would like to answer.
the hash test as it is today is supposed to be a hash test only. we do
not need to remove it. instead we could just change the name of the test
and i would be happy.
i prefer tests that test one thing at a time. call it low level
benchmarking. therefor i would like to question the need for a test that
does both number-to-string conversion and hashtable insert/lookup. we
can have two tests if we are interested.
the spell checker on the other hand is actually doing something. it is
(to me) a high level benchmark. it has a context. to do what it is doing
it needs to combine line-oriented I/O and hashtable insert/lookup.
> I'm a little puzzled that no one has suggested that there should be a
> "map" rather than "hash" test.
one thing at a time, please :-) in the todo section there is already an
entry for adding map/fold/etc.
bengt