Vectorize Macaulay2 functions

[dkahle]

Last night I ran into this problem with dimension(): I wanted to use primary_decomposition() to compute a primary decomposition of an ideal. In Macaulay2, the result of a call to primaryDecomposition is a list of ideals. Fine. (I actually created a class for this, BTW: m2_ideal_list and m2_ideal_list_pointer; it just seems these things will be pretty common.)

The problem became the dimension() is not vectorized in R or Macaulay2. In other words, if you have an ideal I in Macaulay2, you compute its dimension with dim(I). But if it's a List whose elements are ideals, that won't work, and instead you have to use apply(i, dim), so I baked that into dimension(): it now works on ideals and lists of ideals.

I think the same behavior should apply more broadly. Off the top of my head, we have functions like factor_poly() which wraps the M2 code factor(poly), but if you give it a mpolyList or a character vector of poly's, I expect it to error. I think it would be more sensible to have Macaulay2 vectorize it. So: check if the thing is a vector, and if it is use Macaulay2's apply(x, f) construct to get the job done.

Thoughts?

[coneill-math]

This sounds reasonable. Quite a few of our functions expect a list of parameters as their main argument. How difficult is it to implement vectorization, and would it break any current usage of these functions?

In particular, would it require adding a third calling option? We already have all combinations of "_" and "." versions, adding a third to make 8 different functions would be chaos.