Hacker News Reader

Comment by susi22 | Hacker News Reader

Stephan, sorry for using this thread to ask you a question:

In the last ~5-10 year everything has moved more an more to the web. HTML5+JS have pretty much replaced Flash,Java and other proprietary stuff. ~10 years ago it was almost all PHP, ASP. Then came Python etc.

If you look at the speed that libraries were developed for node.js (and now Go), it seems like the web-dev community can really push forward a technology if it can be used to serve web content.

Q: Is there anything done in Julia, to actively attract web-devs for it? I think that if Julia were "promoted" to also be perfect use as a web-server it would attract much more developer and speed up library development.

Don't get me wrong, I'm not a web-dev, but would use Julia to replace MATLAB, but I think it might be a good strategy for Julia to speed up the development (especially for libraries).

View on HN

I have an entirely different take on the speed with which libraries have developed for Node and Go – I don't believe that it's because they are "web languages", but rather because both JavaScript and Go are languages where libraries are written in the high-level user language rather than being written in C via an API. JavaScript gestated in the browser where writing extensions in some other language simply wasn't possible; with the advent of V8, it suddenly became a fast language and all that JavaScript code became the core of an ecosystem of pure JavaScript libraries. Go is new and fast and therefore it has never made sense to write Go libraries in a lower level language.

That being said, writing web apps in Julia is great and very much supported and encouraged. Like Node, we use libuv for I/O and networking. Others have mentioned Morsel, Julia's Sinatra-like web framework, which unsurprisingly – since it's based on the same underlying event framework – gets similar performance to Node.js.

Not every language has to have webdev functionality shoehorned into it. I hardly think that having more web developers involved would speed up the development of the high quality scientific packages that Julia contributors are already currently working on. The people the Julia community should focus on attracting are data scientists and researchers in quantitative fields. That being said, the guys at Hackerschool are working on a web framework for Julia https://github.com/hackerschool/Morsel.jl.

Oh and major kudos to the Julia team. This is a seriously awesome piece of software. The decision to write most of the runtime in Julia itself and rely on a solid JIT compiler and simple FFI to get the necessary performance was very well thought out.

I agree that web developers may not help building a scientific computation language. But...

> Not every language has to have webdev functionality shoehorned into it.

I'm still in doubt about this. Web is the basic human I/O nowadays.

Web is the low-hanging fruit. Take a nice-looking language, add a web framework, and developers can instantly do something useful in a language they like. And as you say, it's a widespread need, so even a percentage of that attention is quite a few people.

I think if a language really nailed a niche, especially a lucrative one like scientific/numeric computing, it could be successful on those merits alone, because it would get a large percentage of that niche.

I like where you're going with this. My question is similar: instead of thinking about Julia serving web content, what are the possibilities for getting it to interact with JS in-browser? If Julia could drive something like D3, or even D3 itself, that would make me really think about moving away from the IPython/Numpy/Pandas setup. (IPython itself already works with Julia, I think.)

John Myles White's Vega.jl package [1] already supports inline D3 plots in IJulia notebooks using Vega [2]. I think it's still a bit beta (we only wrote the IJulia kernel a few weeks ago), but it's already quite capable. This may soon become my favorite way of doing data exploration in Julia (for other things I still love my terminal-based repl).

[1] https://github.com/johnmyleswhite/Vega.jl

[2] https://github.com/trifacta/vega

> If you look at the speed that libraries were developed for node.js (and now Go), it seems like the web-dev community can really push forward a technology

Libraries aren't really being 'developed' for new languages. The best ones are being copied and ported from existing open source code bases, thus the time to deliver libraries has fallen dramatically over the last 10-15 years.

Reply on news.ycombinator.com