block_matrix reacts inconsistently with 0

[sagetrac-jbmohler]

Using ZZ(0) as an element of the list passed to block_matrix appears to be a special case somehow and throws an exception rather than creating the matrix seems reasonable to me.

sage: i=MatrixSpace(ZZ,2,2)(1)
sage: i

[1 0]
[0 1]
sage: block_matrix([1,i,1,1])  # this works as I expect

[1 0|1 0]
[0 1|0 1]
[---+---]
[1 0|1 0]
[0 1|0 1]
sage: block_matrix([0,i,1,1])  # this doesn't ... why is 0 special
...
ValueError: Insufficient information to determine dimensions.

This feels to me like a hazardous inconsistency.

Perhaps I should also add that I don't really like that it just blithely assumes I want a square matrix (although I did in my actual usage). Ticket #2429 addresses that issue more wholeheartedly.

Apply:

1. attachment: 4492_block_matrix_rebased.patch
2. attachment: trac_4492-block-matrix-reviewer.patch
3. attachment: 4492_typo.patch
4. attachment: trac_4492-doctest-number-field.patch

CC: @craigcitro @jasongrout @loefflerd

Component: linear algebra

Author: Willem Jan Palenstijn

Reviewer: Aly Deines, Rob Beezer

Merged: sage-4.6.2.alpha3

Issue created by migration from https://trac.sagemath.org/ticket/4492

[sagetrac-sdietzel]

comment:4

So I figured out what is going wrong here. Patch to come shortly.

[wjp]

comment:6

I think that in the doctest you give there actually is insufficient information, because you can't deduce the width of the left blocks, so an exception is in my opinion the right thing to do there.

In the example in this ticket, it seems to break because it tries to deduce the block dimensions in a single pass through the cells. In this pass, it only determines the size of column 2 and rows 1 and 2. It might need multiple passes to find all information.

[wjp]

comment:7

Attachment: 4492_doctest_nonsquare0_block.patch.gz

Robert Bradshaw suggested adding an example that explicitly shows zero blocks may be non-square. I added a patch that adds that to the docstring of block_matrix.