I have heard that one loud and clear.  I am
sorry I ever considered JudySL with null terminated strings.
As Alan mentions in another post, it is a very difficult problem.
However, I think I have found a solution and trying to find the time
to implement it.  The current proposed API is to just add a string
length parameter to the JudySL API and change the name slightly.
It looks kind of ugly because of the need to return the "length" of
the string in the Next/Prev API's.

Speaking of API's, this is apparently not my forte'.  I would like
some suggestions.

This change would allow all characters (including null) to be in the
index string.

Does that address your questions?


PS.  When Judy was invented, passing parameters was a big extra
overhead to a call.  That is no longer true in a modern processor.
Also, a subroutine call was a big hit on performance, that also is
no longer true.  So, one more advantage of a macro is gone.

----- Original Message ----
From: zooko <zooko@zooko.com>
To: Doug Baskins <dougbaskins@yahoo.com>; judy-devel@lists.sourceforge.net
Sent: Monday, December 10, 2007 2:00:47 PM
Subject: length-associated strings

By the way, one specific detail of extending the API to Judy Trees is 
that I would like is to use keys which are variable-length 
bytestrings associated with a 32-bit length.

I prefer not to use null-termating strings nowadays when explicit-
length strings are safer, faster, and easier for my typical uses. 
It's unfortunate that I can't use these strings as keys in a sorted 
Judy Tree.