Feature idea: Infer implicit dimension value in 'reshape'

[malmaud]

In numpy, I can write

x=rand(10)
and then to reshape x to a (5,2) array, write either
reshape(x, (5, -1))
or
reshape(x, (-1, 2))
instead of
reshape(x, (5,2))

It's convenient since one of the dimensions to reshape is always implicitly defined by the other, and numpy lets you use -1 to signal the dimension whose value should be inferred from the others.
Any interest in having julia's reshape have the same functionality? I could implement it.

[StefanKarpinski]

We've intentionally steered away from using negative values to mean "indexing from the end", instead allowing end in indexing expressions as a special identifier that means the size of the dimension being indexed. Unfortunately, the end thing doesn't work in non-indexing contexts. The trouble with negative indexing is that it can be introduced non-syntactically: it's easy to accidentally write an expression that results in a negative values unintentionally, which then silently does indexing from the end instead of giving a bounds error as it should.

[kmsquire]

How about

reshape(x, (5,:))

as a syntax?

[johnmyleswhite]

I'd be pretty happy using colon that way.

[JeffBezanson]

I think this is too featurey.