ivirshup

Most adobe applications have quite extensive APIs. I think you can even do some linear algebra via Illustrator.

I type it.

It’s typically short short, and it’s a very destructive action. It’s also very apparent to me why I would want to type it out.

That said, someone is going to make a mistake at some point. I’d expect better from github support here.

Logging in does not completley solve this with google. A number of pages languages are still IP based.

You can do both. Check out the difference between `import` and `using`: https://docs.julialang.org/en/v1/manual/modules/

The RFID tags are on a sticker, you can see it on the back when you take it off.

Pre-commit has it's own CI that does this for you.

But if you're in a private repository, you can just have a job that runs pre-commit. The versions of the pre-commit tests are hard-coded into the config.

There's also a bot that keeps your pre-commit config up to date with the newest released versions of the checks.

The guy who coined a new term every time he had a new product to sell?

Don’t be too scared by all the TODO’s in the docs, most of that stuff actually works – the docs just haven’t been written yet. Admittedly those docs have had TODOs on them for over a year.

To see how it works, you’ve got to look at the repo or any of the talks. The developers are also quite active on the tracker, so do just ask if something is unclear.

It’s a ragged array that you can work with in python code, but also Jax and numba code.

JAX/ XLA seem quite popular, especially since you can use them in Colab on TPUs. IIRC, there are people who have managed get jax compiled with ROCm support.

Tensorflow lets you deploy to quite a few backends.

oneAPI support in GPUArrays.jl seems to be coming along.

They also hire software engineers to integrate CUDA into open source projects. On the one hand, those projects can now take advantage of advances in hardware. On the other, users will have to buy NVIDIA to use it, so how open is it? Also NVIDIA will not provide any help with CI, so those costs are borne by the project – or the code isn't tested.

I was, but mostly through XLA or Vulkan layers. Also, I’d define workable as a pretty low bar.

"just create a new instruction set"?

It's taken the better part of a decade for workable alternatives to CUDA to emerge.

How so?

Facebook buying Instagram and WhatsApp is essentially them buying out their competition.

Google running a cloud service is more like renting out excess capacity – which is quite common across industries. Plus there's a lot of competition in the cloud space.

But our package is for scientific data analysis. It’s a python library.

If you’re distributing an application, shouldn’t you just ship an environment with our package bundled?

Can someone make the case for distributing python packages at the distro level to me? Especially scientific software for data analysis?

Our project gets a few issues opened by distro maintainers who have trouble with some part of their build process. Is it really worth our project maintainers time to help troubleshoot esoteric build processes when we already provide source, wheel, and conda distributions?

Big enough to attract developers. As a bioinformatic tool developer, I'm probably gonna stick with Python and Julia since I have access to wider ecosystem of general purpose mathematical libraries and expertise.

Do you really want to have your cocaine purchase in a permanent distributed ledger though? Seems like it could come up again in court.

I think this counts as confirmed (it was kayak/ orbitz)[1]. Pretty sure I heard one of the engineers talking about it in an interview too?

[1]: https://www.wsj.com/articles/SB10001424052702304458604577488...

Presumably this extends beyond Australia though, right? As Square paying $1,000 per Australian would be a bit high.

I would spend $5 on a font, but would not spend $100. I would buy the font the author of this article likes for even $10, but it costs $200 (for just the monospaced variant).

The price point for fonts is set for people who intend to use the font to make money – like graphic designers. One could imagine a lower price point for private use, similar to media.

For sure. I particularly took issue with:

> In e.g. Python, you are not going to run into types you want to subclass, but can't.

In python, you can only really subclass things that were designed to be subclassed. Subclassing arrays, dataframes, or anything from xarray or dask is explicitly recommended against by those packages. In practice I only subclass the base scikit-learn transformers, Abstract Base Classes, or classes I've written.

You can set optimization per module with `Base.Experimental.@optlevel`, though I'm not finding any documentation for it. Could swear it was in a release note.

Can data.table play well with the tidyverse yet?

And as a side note, I believe it's moving down the rankings in the h2o benchmarks [1].

[1]: https://h2oai.github.io/db-benchmark/

But they claim:

> Because these numbers are based on the least expensive plan, they are best case scenarios.

Which is... a weird statement. Best case scenario for what? The cheapest plan is the best case if you wanted to access the site once per billing period. But then price per visit is the cost of plan? I don't know what they're trying to say here.

PRs which add tests and docs, but not new features to maintain? From new contributors? Your professor sounds like a true hero.

As an open source maintainer, I'm happy to have contributions from first time contributors, even if they aren't completley familiar with the workflow. It's how I learned, and I'd love be able to pass this experience on.

That said, I would love if the professor would reach out so I could recommend specific issues and let them know whether I'll be particularly available for more in depth PR mentoring this semester.

It does seem like they're accounting for the burden put on maintainers as well:

> Getting a contribution accepted is not a prerequisite for passing the assignment

> a small-scale contribution is the only realistic goal. The key to making the course's assignment work, is to have what are, on first sight, very low ambitions for the students' contributions.

I initially hadn't realized these were meant to be abstracts (as the site doesn't say this). Knowing this makes hard mode much easier.

I'd been having trouble with ones which had a reasonable logical flow, but didn't communicate a complete idea.

Of course, pretty small N so YMMV

I was thinking something like matplotlib or gnuplot, but I think we may exist in very different ecosystems.

But also, thinking about this a bit more, doesn't excel have the ability to plot geographic data?

> We plot single cell experiment data (lots of points) and we find vegalite, very performant.

I'd be interested in hearing more about your experience here. While I love the grammar of vega-lite (largely via altair), I've definitely run into trouble with many points. Is your work in this area available somewhere? Also, is this for Vitessce?

I think a lot of the performance issues I've come across is based on from copying data between python and the browser. For example, with a million cells you've got at least 5 floats per cell + json structure shipped to the browser. Plus this data tends to stick around in the browser. I'm bullish on approaches like moving data more quickly (possibly even sharing memory, if it isn't a real browser) via Arrow and doing aggregations on the server side with tools like altair-transform. But it's still early days for these.