Re: udigits_to_I

[Sam S. <sd...@gn...>]

> * Bruno Haible <oe...@py...t> [2005-01-28 14:44:01 +0100]:
>
> Sam wrote:
>> why is strm_bitbuffer a LISP heap-allocated object and not a malloc'ed
>> memory area?
>
> Because it is stored in the stream, as a cache (for speed), and a pointer
> to malloc'ed memory becomes invalid after savemem + loadmem.

OOPS!
flushing bitbuff on savemem?...
converting bitbuff to a lisp vector on savemem and back on loadmem?...
yuk...

>> disadvantages: lack of modularity (tight coupling with the stream
>> world)
> Yes, this could be improved.

please do.

>> have to save the buffer when allocating bignum in bitbuff_is_I().
> This is normal when we work with Lisp objects.

but in this specific case this is extremely inconvenient:
it prevents bitbuff_to_I() from accepting a pointer to memory as the
argument (see below ***).

>> bitsize and bytesize; which is confusing (one should be enough)
>
> bytesize = ceiling(bitsize,8);
> Indeed just two cheap arithmetic instructions.

let's drop it, then.

>> bitbuff_is_I()&bitbuff_ixu_sub can be rewritten as a general bits<->I API:
>
> Good idea. That could go into a new file, to be included from lisparit.d
> right after intread.d:
> intserial.d  -  serialization of integers as little-endian byte sequences

ok.

>> /* convert the bit buffer of the given length to a (un)signed integer */
>> maygc object bitbuff_to_I (const uintB* buffer, uintL length, bool signed_p);
>
> There probably need to be two variants of this function, one taking an
> "const uintB* buffer", the other one taking an "object buffer" (a
> simple-array of rank 1 and element-type (unsigned-byte 8)).
> I would call these functions LEbytes_to_I and LEbytes_vector_to_I
> instead of bitbuff_to_I.

this need for two functions is the reason why the `maygc' property of
bitbuff_to_I() is so bad (see above ***)

> This intserial.d can be written indepedently of any streams. While
> then changing stream.d to use it, please be careful with the quality
> of the ASSERT_wr_int error message: a stream object with a filename in
> it is very useful for the user when an error occurs.

I hoped that you would take care of this.
I'd rather not touch the bit-pushing code.


-- 
Sam Steingold (http://www.podval.org/~sds) running w2k
<http://www.camera.org> <http://www.iris.org.il> <http://www.memri.org/>
<http://www.mideasttruth.com/> <http://www.honestreporting.com>
I haven't lost my mind -- it's backed up on tape somewhere.