> Le 28 nov. 2018 à 06:03, Paul Eggert a écrit :
>
> Akim Demaille wrote:
>> What do you propose?
>
> A simple, general solution is to use only malloc/calloc/realloc, and to have
> bitset functions fail when they exhaust memory.
Then I guess the current state does what you want. If you
Akim Demaille wrote:
What do you propose?
A simple, general solution is to use only malloc/calloc/realloc, and to have
bitset functions fail when they exhaust memory.
> Le 27 nov. 2018 à 21:31, Paul Eggert a écrit :
>
> On 11/26/18 8:15 PM, Akim Demaille wrote:
>> The current implementation uses xmalloc, but there are also calls to calloc
>> and realloc.
>
> How can it be right to combine xmalloc with calloc/realloc?
I agree.
> Shouldn't the
On 11/26/18 8:15 PM, Akim Demaille wrote:
The current implementation uses xmalloc, but there are also calls to calloc and
realloc.
How can it be right to combine xmalloc with calloc/realloc? Shouldn't
the implementation use xmalloc, xcalloc, and xrealloc for consistency?
Also, what about
Hi Paul!
> Le 25 nov. 2018 à 20:43, Paul Eggert a écrit :
>
> One top-level question is how does memory allocation work? Emacs has its own
> memory allocator, and doesn't want to use plain malloc. It has its own
> xmalloc implementation; will that suffice?
The current implementation uses
Hi Bruno!
> Le 25 nov. 2018 à 21:01, Bruno Haible a écrit :
>
> Hi Akim,
>
>> Or completely change: all the bitsets of a bitsetv have the same dimension,
>> so we could go to bitmatrix.
>
> On the other hand, the module does not implement matrix operations
> (matrix multiplication,
Hi Akim,
> Or completely change: all the bitsets of a bitsetv have the same dimension,
> so we could go to bitmatrix.
On the other hand, the module does not implement matrix operations
(matrix multiplication, transposition, determinant), only transitive
closure (which maybe is not so suitable
Akim Demaille wrote:
I think it should move to gnulib: it's way more powerful than what
we use in Bison only. Paul already reported he would be interested
in using this module in Emacs.
Yes, Emacs already has a bitset implementation for its boolean vectors. My vague
recollection is that the
> Le 25 nov. 2018 à 19:55, Akim Demaille a écrit :
>
>>> Sure. You can push it into gnulib, without having extensive tests.
>>
>> Great! I read this as an ok to push the current state, which takes
>> into account your comments, but will most probably require more changes
>> later.
>>
>>
> Le 25 nov. 2018 à 19:52, Akim Demaille a écrit :
>
>> Le 25 nov. 2018 à 18:54, Bruno Haible a écrit :
>>
>> Hi Akim,
>>
* Regarding tests: There is a large amount of algorithmic code. While
the parts that Bison uses are surely correct, there is a possibility
of bugs in the
> Le 25 nov. 2018 à 18:02, Akim Demaille a écrit :
>
> Hi Bruno!
>
>> * File names are nowadays not limited to 8+3 characters, and names
>> like 'abitset.h', 'bbitset.h', 'ebitset.h', 'lbitset.h', 'vbitset.h'
>> are not really telling. How about using more descriptive names?
>> Note: You can
Hi Akim,
> > * Regarding tests: There is a large amount of algorithmic code. While
> > the parts that Bison uses are surely correct, there is a possibility
> > of bugs in the more rarely used parts. I would find it very useful
> > to have unit tests of all 32 operations.
>
> I definitely
Hi Michael,
I meant to put you in CC, but forgot to do it. This message
is to inform you that we are moving your bitset library, as it
is currently in Bison, to gnulib.
See https://lists.gnu.org/archive/html/bug-gnulib/2018-11/msg00045.html
and following.
Cheers!
> Le 25 nov. 2018 à 09:57,
Hi Bruno!
> Le 25 nov. 2018 à 16:44, Bruno Haible a écrit :
>
> Some remarks:
>
> * I can imagine programs that want bitsets but no bitset-vectors.
> Suggestion: Move the bitsetv* code to a second module, named
> 'bitset-vector'.
That sounds good, indeed.
> * File names are nowadays not
Hi Akim,
This code looks like well-designed, high-quality and very suitable for
gnulib.
I love especially about it that it's adaptable as programs change:
Initially the program might require fixed-size bitsets, and later
move to variable-size bitsets.
> I think it should move to gnulib
Yes,
Second part: doc and tests.
commit 52342213a42a77b639174e3fc7f2c586f6fc2422
Author: Akim Demaille
Date: Sun Nov 25 09:49:09 2018 +0100
bitset: add tests and doc
First stabs at providing a documentation and test for the bitset
module.
* doc/bitset.texi, modules/test-bitset,
16 matches
Mail list logo