John:
I believe that it was designed to return the easiest thing to cast (in C) to
what it really is. I.E. it could be a pointer to an unsigned long -or- another
Pointer,
dependingwhat the user put in that memory location. It turned out (later)
that in C++ this was "awkward" for the Macros JLG() and friends.
Oops, your are talking about the "Judy array pointer". We debated that one
a lot too. I vaguely remember there was a caveat that made it more awkward
(to a type other than a void **). Perhaps C has changed since then.
I think it had to do with one of the tools that did "type" checking (in the
90's).
The Judy API has always been awkward. I have never been happy with it.
I have received a lot of criticism about it over the years. I challenge
anyone to come up with a better one, but so far there have been no takers
(that actually deliver a Judy.h file with the prototypes that can possibly
work).
I spend most of my time with the internals of Judy. The API is just something
that C and C++ make awkward (or I am just not smart enough to make it better).
doug
Doug Baskins <[email protected]>
PS. Much of the design of Judy was to overcome the CPU design (flaws) of
the early 2000's. The Macros were done to avoid an extra subroutine
call -- which were extremely costly before the 2006 CPU designs (core 2 duo).
I now struggle with CPU cache/TLB design flaws (I think) which are very
difficult-to-impossible to predict and/or program around. The symptom is
an apparent TLB miss in the Text cache -- even when the Text working set
is using just 2 pages. So far this "flaw" has been present in all generations
of Intel CPU's since. AMD CPU's have been so slow, it is hard to tell.
>________________________________
>From: john skaller <[email protected]>
>To: judy <[email protected]>
>Sent: Wednesday, November 2, 2011 3:39 AM
>Subject: Judy array types
>
>Instead of a void* why not
>
>struct Judy1Array { void *data};
>struct JudyLArray { void *data};
>
>This has the small advantage that one can't confuse Judy1Array and JudyLArray,
>since they now have distinct types.
>
>--
>john skaller
>[email protected]
>
>
>
>
>
>------------------------------------------------------------------------------
>RSA® Conference 2012
>Save $700 by Nov 18
>Register now!
>http://p.sf.net/sfu/rsa-sfdev2dev1
>_______________________________________________
>Judy-devel mailing list
>[email protected]
>https://lists.sourceforge.net/lists/listinfo/judy-devel
>
>
>------------------------------------------------------------------------------
RSA® Conference 2012
Save $700 by Nov 18
Register now!
http://p.sf.net/sfu/rsa-sfdev2dev1
_______________________________________________
Judy-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/judy-devel
