On Wed, Jun 13, 2012 at 8:31 AM, Ronan Lamy wrote:
> Le mercredi 13 juin 2012 à 02:06 -0700, Aleksandar Makelov a écrit :
>> Hi again,
>>
...
>
>> - remove the checks from Perm, and make a *function* PermS (the name
>> is deliberately chosen to be as close as Perm) that performs the
>> checks and then constructs a permutation object using Perm (the name
>> comes from Permutation Safe).
>
> Urgh, no! The design is OK, though not terribly consistent with the rest
> of sympy where users need to directly call class constructors all the
> time, but the naming seems designed to maximise user confusion. Casual
> users shouldn't need to know about the concept of permutation
> "safeness", nor have to remember a cryptic abbreviation for it, and
> functions names (conventionally lowercase) shouldn't make them look like
> classes.
I'm a little confused. Is PermS a user function? I thought it was internal.
If it is internal, and not a function for the casual user, then it
might be renamed _PermS, so it doesn't pop up in tab completion.
Using Perm in Permutation is a bit ugly, but it is endorsed by Knuth
(quote which I found in http://www.xact.es/xPerm/index.html)
''Following a suggestion of Vaughan Pratt, we adopt the convention that
perm = permutation, perhaps thereby saving millions of syllables in
future research.''
I can accept that PermS be used to create safely permutations, along with Perm which creates them without checking, although I would then use only PermS, since I care about the checks; it is a place less to check if I have a bug.
If I need speed I use _new_from_array_form, which is faster than the proposed Perm. I am strongly against _PermS, which is not convenient to type.