search

scijs / ndarray

📈 Multidimensional arrays for JavaScript
MIT License
1.21k stars 61 forks source link

Docs should say that hi/lo accept nulls #23

Open letmaik opened 9 years ago

letmaik commented 9 years ago

Since lo()/hi() accept nulls this should be added to the Readme as it is quite convenient.

rreusser commented 8 years ago

Can you elaborate on your use case? I've used it to just just slice an array in one particular dimension without specifying the bounds of the others.

letmaik commented 8 years ago

Yes exactly, that's what I do as well, slicing across one or more dimensions and leaving others untouched (where I use null then instead of giving explicit full bounds).

rreusser commented 8 years ago

See: https://github.com/scijs/ndarray-concat-rows/blob/master/index.js#L72

I'm updating the docs for another ticket so will add a note on this now.

letmaik commented 8 years ago

What about the example? I don't see any null's.

Am 07.01.2016 um 21:49 schrieb Ricky Reusser:

See: https://github.com/scijs/ndarray-concat-rows/blob/master/index.js#L72

I'm updating the docs for another ticket so will add a note on this now.

— Reply to this email directly or view it on GitHub https://github.com/scijs/ndarray/issues/23#issuecomment-169816677.

rreusser commented 8 years ago

Ah, sorry. Yeah, seemed (unless I'm wrong) that as long as it's undefined or null (omitted as an argument = undefined), lo doesn't increase the lower bound and hi doesn't decrease the upper bound in the unspecified dimensions.

rreusser commented 8 years ago

In other words, I used that to grab a section of the ndarray along the first dimension only by leaving the others implicitly unspecified. Concatenating cols was a little more complicated because I had to construct the array of indices explicitly.

(Side note: I'm realizing as I clean some of this up that really more and more of the code should use dynamic code generation to keep this generic n-dimensional programming from actually having to operate in an abstract number of dimensions. With code generation, you just generate the code for the actual number of dimensions for the given input…)

But anyway, AFAIK, either null or implicitly undefined both works.

letmaik commented 8 years ago

Right, ok.

About dynamic code generation, as long as it actually has a measureable performance benefit and the code is used enough times that it matters I'm all for it.