CoCoA » CoCoALib

Here is an alternative suggestion (affecting the names of all PPMonoids).

The underlying principle is that an object which imposes limitations should have a name which indicates that it is not fully general (thereby presumably encouraging a potential user to check the documentation...)

Currently all PPMonoids (except PPMonoidEvZZ) impose limitations on the maximum degree of a PP. It would make more sense if the degree limited PPMonoidEv had a name something like PPMonoidEv_SmallExp. Then we could also have PPMonoidEv_BigExp.

The length of the class names is not important to a normal CoCoALib user who sees only the pseudo-ctors NewPPMonoidEv. We could merge NewPPMonoidEv and NewPPMonoidEvZZ into a single pseudo-ctor which takes an optional arg saying whether we want "big" or "small" exponents -- strictly following my reasoning above the default ought to be "big" exponents (but here I see a potential clash between ease of use and "correctness").

Inside the .C file we could create two (virtually identical?) classes choosing whatever names we like. It might make sense to put these classes inside a namespace, so we'd actually have PPMonoidEv::SmallExpImpl and PPMonoidEv::BigExpImpl (for instance). If the impls really are virtually identical we ought to consider a template class implementation; at the moment I do not expect the impls to be sufficiently similar (because one packs the exponents into integers, the other will probably not use any such packing).

We could also consider a "very small exponent" case too. This might give some slight space advantage, but probably no significant variation in computation time. Note that currently the integer type used for exponent values is determined by a typedef in the file config.H. Personally I doubt that a "very small exp" version will be useful because new computers seem to have lots of RAM (8Gbytes in a portable is already available); in any case a perceptible gain would arise only if the coeffs are in a small finite field and there are lots of indets.

We have not yet decided about internal impl names.
JAA still likes his suggestion made two posts ago.

A "very small exp" version may be slightly faster because more PPs fit in the cache, but I doubt the slight gain would be worth the extra complication in the source.