Andrew Nikitin wrote:
> What is the reference implementation of dictionary AKA hash?
If your keys are strings, there is always s: (maybe this uses the same
hashing algorithm that supports J's name mechanisms?) .
You still have to do the i. thing but it should be faster than using boxed
strings. Depends on how large your universe is.
Roger wrote:
> Probably best to avoid the phrase (V,a:) or
> (V,nullvalue) in get. . Better to use something like
> V {~ :: (nullvalue"_)
No way! That's scalar programming at its ugliest. One bad key spoils the
bunch. Or you have to apply get at small rank - a no-no.
There has to be a better way. Barring special code for the idiom, maybe use
the special support for append in place [1], inside
put? That is, in the definition of put, insert this line between end. and
i.0 0
V=:V, a: NB. Default invalid keys
> Another is to use J's built in ability to search for a name in locale.
> This one just does not feel right.
I agree that it doesn't feel right, but this is probably the best option (in
terms of convenience and performance). Depending on
the kinds of keys you're dealing with, the "namification" process could have
low overhead. For general keys (i.e. arbitrary
nouns), you could always do something along the lines of tobase64 3!:1 y .
-Dan
[1] http://www.jsoftware.com/help/release/apip.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
