Opened 9 years ago

Last modified 9 years ago

#13678 closed enhancement

Allow tab completion of matrix constructor — at Version 13

Reported by: robertwb Owned by: jason, was
Priority: major Milestone: sage-5.7
Component: linear algebra Keywords:
Cc: Bouillaguet Merged in:
Authors: Robert Bradshaw, Volker Braun Reviewers: Volker Braun
Report Upstream: N/A Work issues:
Branch: Commit:
Dependencies: Stopgaps:

Status badges

Description (last modified by robertwb)

E.g. matrix.identity(...), matrix.random(...), matrix.load(...) etc. all discoverable by tab completion.

Apply 13678-matrix-methods.v3.patch and 13678-doctests.patch to the Sage library and 13678-sagenb.patch to sagenb.

Change History (18)

Changed 9 years ago by robertwb

comment:1 Changed 9 years ago by robertwb

  • Status changed from new to needs_review

comment:2 Changed 9 years ago by rbeezer

Very nice! I discovered a couple constructors I didn't know about (just reading the patch). I'll add it to my queue of things to review, but anybody else should feel free to beat me to it.

comment:3 Changed 9 years ago by rbeezer

I like it! Works as advertised. I'll run tests shortly. In the meantime, a suggestion and a question.

(1) Stripping out "matrix" from the name of the method is a nice touch, but then the docstring appears (to the novice) to be talking about a slightly diffferent command than the one queried. Having ready access to constructors is the first step in experimenting with Sage, then tab-completion takes over for methods on that object. So tab-completion to discover constructors is fabulous. But it should be as totally straightforward as possible, IMHO.

What do you think of putting "matrix" back in the method names of the matrix object? Yes, it is verbose and redundant. It does not seem to complicate tab-completion (ie, you do not need to use tab any more in either scenario). And in the Sage library, authors can use the "old" method names without involving the matrix object (and I would think this would be preferable).

(2) Is there anyway to make the func_* methods on this object invisible on tab-completion? I have no real good idea if they are useful, or just detritus. It'd be great if matrix.<tab> only showed the constructors.

Rob

comment:4 Changed 9 years ago by robertwb

Both valid points of feedback.

(1) I'd lean towards not putting matrix back in the name; what if we modified some of the examples to use the new format as well? Would that make things clear that we're talking about the "same function." We could also (automatically) add a line in the docstring that "This function is available as matrix.identity or identity_matrix"

(2) Short of hiding them for all functions (-1 to that) another option is to matrix() a __call__ method on a class. If you want I can post a new patch doing this.

Changed 9 years ago by vbraun

Improved patch

comment:5 follow-up: Changed 9 years ago by vbraun

  • Authors set to Robert Bradshaw, Volker Braun
  • Description modified (diff)
  • Reviewers set to Volker Braun

Great idea to use a decorator! I've implemented the __call__ version. Also, I gave the decorator an optional name= argument if you want to override the automatic name generation. But I'm happy with the matrix-removed names. I agree that it would be better if matrix.foo were then mentioned in the docstring of foo_matrix, feel free to fix this. But its not really necessary. I'm giving positive review to Robert's patch ;-)

comment:6 in reply to: ↑ 5 Changed 9 years ago by rbeezer

Replying to vbraun:

I agree that it would be better if matrix.foo were then mentioned in the docstring of foo_matrix, feel free to fix this.

If it is easy to inject something like "matrix.foo() is equivalent to foo_matrix()," could this be added?

It drives me nuts when we define an alias for some method and it is not included in the docstring of the original. I'd put this in the same category, so if a one-time hunk of code takes care of it, then I'd love to see that happen. Just trying to keep an eye out for the newcomers and minimize unnecessary confusion when there are two ways to do things.

Changed 9 years ago by robertwb

More improvements.

comment:7 Changed 9 years ago by robertwb

  • Description modified (diff)

OK, I added automatic mention of the two forms in the docstring, as well as slight simplification to Volkers patch and showing the actual constructor code for ?? (which I've actually used before).

comment:8 Changed 9 years ago by robertwb

APply only 13678-matrix-methods.v3.patch

comment:9 Changed 9 years ago by jason

A few more examples of special matrices are up at #13703

comment:10 Changed 9 years ago by jason

Gee, it seems like it would be nice to generalize the decorator one more level, so I could do something like:

@namespace('matrix')
def vandermonde(R, v):
    return matrix(R, len(v), lambda i,j: v[i]^j)

I could see this being useful for other situations like this too.

@namespace('groups')
def alternating(...):
    ....

comment:11 Changed 9 years ago by vbraun

  • Status changed from needs_review to needs_work

There are a few failing doctests; sage_getsource and sage_getdoc get confused and interact(matrix) needs fixing.

comment:12 Changed 9 years ago by Bouillaguet

  • Cc Bouillaguet added

Changed 9 years ago by robertwb

Changed 9 years ago by robertwb

Apply to sagenb repo.

comment:13 Changed 9 years ago by robertwb

  • Description modified (diff)
  • Status changed from needs_work to needs_review
Note: See TracTickets for help on using tickets.