On Sun, Aug 17, 2008 at 4:46 PM, Roland Averkamp roland.averkamp@gmx.de wrote:

Is it really necesary to use the Babel and Alexandria libraries in CFFI? Now CFFI depends on two libraries and every other system which uses CFFI needs these two libraries too.

That's an issue that was taken into consideration.

Some comments in Babel mention that it is only usefull for a lisp with unicode suppport. But for example Allegro and Clisp already have functions for character set conversions.

At one point, there was a CFFI branch that implemented foreign encodings using the functionality provided by each Lisp. But the encoding support and error semantics vary wildely and that made it a bit useless. You can find more details about that effort in the mailing list archives.

I think it would be better that strings with a non native character set are handled like (unsigned-byte 8) or (unsigned-byte 16) arrays which are zero terminated. And it is up to the application to convert them, with Babel or some buildin functions.

That would introduce an extra step in the conversion:

foreign memory -> lisp array -> lisp string

A better solution, perhaps, would be to split string support into a different system. Is it worth the trouble though?

Out of curiosity, plain-odbc seems to use strings a lot, I would assume that foreign string encoding support would be useful for your library. Am I wrong?

Regarding Alexandria: I believe only three macros from Alexandria are used.

Maybe ONCE-ONLY is nice, but not much simpler than (let ((a (gensym "a"))) `(let (,a ...) ... ,a)).

The same holds for WITH-UNIQUE-NAMES.

Version 0.10.0 uses ENSURE-LIST, FEATUREP, FORMAT-SYMBOL, IF-LET, MAKE-GENSYM-LIST, ONCE-ONLY, PARSE-BODY, SYMBOLICATE, WHEN-LET and WITH-UNIQUE-NAMES. They are, of course, somewhat simple utilities that could be copied & pasted into Babel and CFFI.

Is that code duplication better or worse than adding a library dependency? Perhaps if you use some of Alexandria's utilities in your own code you'll feel better about it? :-)