ANN: Our Julia paper in SIAM Review


#1

Hello all,

For many of you this may already be old news. But, since many people are still citing the arxiv paper, it is worth mentioning that our paper - Julia: A fresh approach to numerical computing is now published in the SIAM review, and I put up a short blog about it.

http://julialang.org/blog/2017/03/julia-fresh-paper

We wrote it for the applied mathematicians rather than for computer scientists (for whom we have other publications).

Not sure if Community is the right forum to post in, and if we should have one for papers, given that there are now a whole bunch.


Citations in docs
#2

R has a citation() function that will print the recommended citation(s) for the language to the console. You can also pass it a package name and get the recommended citation for that package. Maybe worth considering? It certainly makes it easy for users to cite the right source.


#3

Yes.
If Base.citation() was defined,

Then could also define:

Base.citation(m::Module) = citation(Val{modulename(m)}())

And then package authors could define:

Base.citation(::Val{:MyPackagesMainModuleName}) = "A paper about my package, by package author..."

Maybe citation() would go through each currently loaded Module and look for a citation for all of them and make an overall report. (With the citation for the language itself at the top.)

And citation(Core) would give the citation for the language itself, alone.


#4

The citation function is actually a pretty cool idea, especially with modules being able to define them. Perhaps a PR is in order! Would one just spit out the BibTeX? I guess there could be extra arguments for the format.

-viral


#5

Very nice and Julian! The R way (which is not necessarily better) would be to save an e.g. BibTex formatted CITATION file in the package dir, and then just something like readlines(joinpath(Pkg.dir("MyPackage"), "CITATION")), parse it into a Dict, and then have a keyword argument for human-readable format or all the big citation formats. It would have the advantage of working without loading a package. But I think I prefer @oxinabox solution as it feels more “Julian”.


#6

I feel that there are many situations where it would be nice to get information (“metadata”) about packages without loading them. Maybe something like this can be introduced in Pkg3.


#7

I am trying Julia as a numerical analyst and estimate it superior to MATLAB etc.
When reading the SIAM paper, I was led to new ideas, which I immediately wanted to try out.
In have Julia 0.5.1 installed on Fedora25 and wanted to try out the Escher-Package v0.3.3, propagated in the paper as a simple means for creating web contents. Unfortunately the server could not load any example file.
After a trivial code change in cli/serve.jl ( “ui == Function” was replaced by “isa(ui, Function)” ) the server started up but only 6 of the 11 package examples worked.
Most errors were related to Gadfly. After fixing the build.jl of Cairo, 8 of 11 were ok, the rest probably due to errors in Escher itself.
I propose to fix the Escher package to make it usable with the current version of Julia in order to avoid too much frustration to the willing Julia aspirant.