math.subset converts matrix to scalar

[DbCrWk]

First, this is a great package!

One small quirk that I've noticed is that if I do something like:

const matrix = [[1, 0], [0, 1]];
const expectedSubMatrix = [[ 1 ]]; // this is a 1-by-1 matrix

const upperElement = math.range(0, 0);
const actualSubMatrix = math.subset(matrix, math.index(upperElement, upperElement));
// actualSubMatrix is the scalar/number 1, rather than a 1-by-1-matrix

Is this an intentional casting of a 1-by-1 matrix into a scalar? If so, is there a way to enforce that pulling the subset of a matrix always returns a matrix?

(For my particular use case, I have a function that accepts an arbitrary index k and returns the k-by-k upper left submatrix. Therefore, to get around this quirk, I have to special case the situation when the passed k value is 1.)

@pdeaville

[josdejong]

Thanks @DbCrWk, great to hear you like the library :)

Changing 1x1 output of a matrix subset is indeed intentional. When retrieving a single value from a matrix, often you want to continue working with it as being a scalar.

There is currently no way to enforce output being kept a matrix. We could consider making this possible for example by configuring predictable: true globally, or passing this as an extra option to subset.

What you could do for now is create a small wrapper around math.subset which checks if the output returned by math.subset is a scalar and if so, turn it into a 1x1 Matrix again and return that.

[DbCrWk]

Good to know. I have a wrapper written around it for now. If you're good with some parameter being passed (or equivalent), I can open a PR on this!

[josdejong]

Thanks for your offer Dev. I'm still a bit in doubt about what would be the best approach here: passing an option along with subset, or using the global option predictable for this. I think I have a slight preference for using a global option for this, since inside the editor you use bracket notations like c = A[1, 2], which under the hood uses subset, but there is no way to pass an option there.

[DbCrWk]

Sounds good! I'll take a crack at it.

[josdejong]

Cool, thanks!

This will be a breaking change so we have to schedule when to release.

@DbCrWk can you please work from the modular_architecture branch instead of develop? That will save us some merge conflicts. We're about to release a large, breaking refactor of mathjs, which for example has a different dependency injection mechanism. See #71.

[DbCrWk]

Cool, will do

[josdejong]

Dev, I've just merged modular_architecture into develop, so please continue any work there again.

[gwhitney]

Closing as duplicate of #2344.