Welcome, Pranav! Here is a good first issue to help get acquainted with the project: stdlib-js/stdlib#519
Looking forward to your first contribution! Let us know if you have any questions or need help. :)
The issue I linked to above is to help get acquainted with project conventions and structure. If you’re interested in lower-level things, depending on how things go, we can explore some of stdlib’s native C/Fortran bindings.
an excellent initiative Pranav, hopefully you will set an example for others to follow!
After cloning, you should follow the development guide: https://github.com/stdlib-js/stdlib/blob/develop/docs/development.md
However, instead of running
make install, just run make install-node-modules; the former builds native add-ons which you do not need to do for the feature you’re working on.
The package development guide provides an overview of package contents: https://github.com/stdlib-js/stdlib/blob/develop/docs/packages.md
Hey! I'm trying to make a contribution to this library, but am getting bogged down in simply writing the documentation and following out how these README are generated. I see that there are for instance, some directives such as <toc> and <toc-links> in the markdown. Is there something I'm missing and there is something that automatically fills in the content between these two tags? What's modifying your markdown?
Are you attempting to contribute a statistical distribution? Otherwise, the only packages which have
<toc> directives are namespace READMEs, which, for individual package contributions, shouldn’t need to be touched.
If it would help, you can use the README snippet found here: https://github.com/stdlib-js/stdlib/blob/develop/tools/snippets/README.md
That snippet should provide a minimal starting point.
Snippets are discussed a bit more in this development guide: https://github.com/stdlib-js/stdlib/blob/develop/docs/packages.md#development
I am trying to contribute a statistical distribution. Metalog (stdlib-js/stdlib#512) Would I need to know how to use the <toc> directives if I was?
Typically statistical distributions are comprised of individual packages for the various properties/methods. Those individual packages are then packaged together as part of a distribution class. So, for your the metalog dist, you might start by creating a package for the PDF, followed by CDF, etc. Once those are created, we’d create a namespace which combines those individual packages into a single class.
A good example of this is
@stdlib/stats/base/dists/normal, which has various sub-packages, such as pdf, cdf, quantile, mgf, etc. Those are then packaged up and exposed via both a class implementation @stdlib/stats/base/dists/normal/ctor and a namespace.
In short, I’d suggest following the precedent, if possible, of the other distributions in the library. If, for whatever reason, those dists are not amenable to what you’re trying to do with Metalog, we can discuss further.
Looking at the RFC, maybe start with
@stdlib/stats/base/dists/metalog/pdf? In which case, you don’t need to know anything about <toc> directives, as those are not applicable for the README you’d create. Instead, you can just use the snippet linked to above or copy and paste a README for a PDF of another distribution and modify accordingly.
Now, not that I imply that
@stdlib is still not ready, but while you are waiting for it, there is a stop-gap solution: https://www.npmjs.com/package/pymportnumpy, pandas and everything else come to Node.js
I am starting to play with NDArrays and have an API question. Wouldn't it make sense to support creating arrays from typed arrays directly? Something like
stdlib.ndarray(new Int32Array(d3.range(10))). Wouldn't that be easier than having to repeat the type and incurring a copy if it doesn't match?
:point_up: Edit: I am starting to play with NDArrays and have an API question. Wouldn't it make sense to support creating arrays from typed arrays directly? Something like
stdlib.ndarray(new Int32Array(d3.range(10))). Wouldn't that be easier than having to repeat the type?
@domoritz:matrix.org You can use the
stdlib.array API: https://github.com/stdlib-js/stdlib/tree/develop/lib/node_modules/%40stdlib/ndarray/array. This infers the dtype from the input buffer. The ndarray constructor is a “lower level” API.
5 replies
Here is where I am playing with the API to understand it: https://observablehq.com/d/2c266ef99216e94f.
@domoritz:matrix.org Here is an edited notebook which fixes a bug in the notebook you posted: https://observablehq.com/d/4a040427e7eb3423
1 reply
Is there a more high-level version of https://stdlib.io/docs/api/latest/@stdlib/math/strided/ops/add? I want to add vectors.
I'm having a hard time getting the arguments right in my vector math.
strided.ops.add(a.length, "int32", a, 1, "int32", a, 1, "int32", out, 1) throws RangeError: invalid arguments. Number of indices must match the number of dimensions. ndims: 2. nargs: 1..
I also wonder whether you have a method to quickly view ndarrays. Scijs has imshow which I found really convenient when I wanted to debug my array math: https://github.com/scijs/ndarray-imshow.
The issue you are encountering is that, for the strided APIs, you need to work directly on the underlying ndarray buffers, rather than on the ndarray object.
Sweet, thank you for all the demos. That's exactly what we need to get rolling.
I wonder whether you plan to have more high level APIs for working with strided arrays at some point similar to scijs? This isn't a blocker for my project but given that I had troubles understanding the APIs, I can imagine others may benefit from more convenient APIs as well.
@domoritz:matrix.org Definitely. Higher level APIs are definitely something we want to add. And some are in progress. We’ve been focusing on C APIs/implementations at the moment, in order to roll out our ufunc functionality. The plan after is to add various slicing and array creation APIs to help with ergonomics.
Hey Harshita! Welcome! You can find a list of good first issues here: https://github.com/stdlib-js/stdlib/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22
E.g., something like stdlib-js/stdlib#811 would be a welcome contribution.
@kgryte: Are you aware of any generic lanczos impementations in JavaScript? I saw that you had some primitives, but nothing that the average layman could use? There are two npm packages but both are limited to 8-bit integers as they do only images, I need to do it for DEM data that could be 16/32 bit integers or even floating point values
(or once again we will have to 'borrow' from Python 😀)
maybe it is the right moment to implement conversions between
numpy arrays and scijs/stdlib arrays for pymport 😀