Contents and repositories for JuliaMath

[ararslan]

Let's figure out what we want to adopt from other organizations and what new things we should have.

• Number theory
  • Primes (from Base--see Should primes.jl be moved into its own package? JuliaLang/julia#16357)
  • Elliptic curves
  • Continued fractions (johnmyleswhite/ContinuedFractions.jl)?
• Special functions
  • Special functions (nolta/SpecialFunctions.jl)
• Combinatorics
  • Combinatorics (JuliaLang/Combinatorics.jl)
• Calculus
  • Root finding (JuliaLang/Roots.jl)
  • Differential equations (JuliaLang/ODE.jl)
  • Differentiation/integration (johnmyleswhite/Calculus.jl)?
• Topology
  • Homology, complexes, ...
• Linear algebra
  • LinAlg? (if those functions get removed from Base)
  • Iterative solvers (JuliaLang/IterativeSolvers.jl)
• Arithmetic/numerics
  • Double-double arithmetic (simonbyrne/DoubleDouble.jl)
• Bindings for libraries
  • Yeppp! (JuliaLang/Yeppp.jl)
  • Sundials (JuliaLang/Sundials.jl)
  • VML
  • Apple Accelerate

What do you think of these? What else?

[ararslan]

@jiahao @andreasnoack -- Perhaps relevant to your interests as well?

[mschauer]

Should there be a MathCore.jl library providing basic functionality to break up interdependencies?

[pabloferz]

Also, I don't know if there is interest on the part of @jverzani or @wbhart to migrate their CAS (SymPy.jl and Nemo.jl), for example, to an organization.

[wbhart]

We already have moved Nemo.jl to an organisation on GitHub. We just don't
have a release from the org yet. But one is coming soon enough.

https://github.com/Nemocas/Nemo.jl

Bill.

On 18 May 2016 at 12:11, Pablo Zubieta [email protected] wrote:

Also, I don't know if there is interest on the part of @jverzani
https://github.com/jverzani or @wbhart https://github.com/wbhart to
migrate their CAS (SymPy.jl and NEMO.jl), for example, to an organization.

—
You are receiving this because you were mentioned.
Reply to this email directly or view it on GitHub
#1 (comment)

[jverzani]

There are no reasons SymPy couldn't move over save intertia.

[ararslan]

@mschauer What kind of interdependencies do you imagine would occur and what would you propose to go in a core math repo?

[mschauer]

In the moment of writing I thought about a module Primes including a module SpecialFunctions to call the logarithmic integral function and a reverse dependency for example via quasi montecarlo integration methods but there are likely better examples.

[ChrisRackauckas]

What about number libraries like doubledouble? And what about PetsC?

[ararslan]

@ChrisRackauckas I don't know what those are. Can you gimmeh teh linkz pls?

[ChrisRackauckas]

PetsC.jl is a popular package for sparse solvers and other tools for computational PDEs.

Doubledouble.jl implements quad precision arithmetic.

I'd be interested in migrating my package DifferentialEquations.jl which is for "non-ordinary" differential equations (adding my stochastic differential equations parts next weekend?)

[ararslan]

Hm, JuliaParallel already owns PETSc and since it depends on MPI it probably makes more sense for them to keep it.

Your DifferentialEquations.jl would probably fit here, though the name is a little misleading as it doesn't include ODEs. Org members: What do you think?

@simonbyrne What would you think about migrating your DoubleDouble.jl to JuliaMath? Do you think this would be an appropriate home for it?

[simonbyrne]

Yes, I would be happy to move DoubleDouble.jl.

[ChrisRackauckas]

I think NLSolve.jl would be a good fit here, but it's also in an org.

For symbolic math another interesting development is Sjulia.

[ararslan]

@simonbyrne Awesome! Feel free to transfer at your leisure. 😄

[ararslan]

@ChrisRackauckas NLSolve probably belongs in JuliaOpt more than anywhere else, I would think. As for SJulia, I think we should keep an eye on it as it matures.

[ararslan]

I saw an old julia-dev post where @StefanKarpinski mentioned that an organization like this would be quite broad. I definitely agree with that. As such I think we should think carefully about the scope of the organization and what does and does not belong here. I'd especially be interested in input from others on what does not belong here.

[ChrisRackauckas]

I think to complete my list, Brownian.jl would be a good fit.

But what about linking libraries? Things like MATLAB.jl and Mathematica.jl?

[ararslan]

Would stochastic processes not fall under the umbrella of JuliaStats? I'm not really sure how they've defined their scope.

[ararslan]

Regarding Matlab and Mathematica, I wouldn't consider those on topic here. (Even my proposal of Yeppp! and Sundial is kind of pushing it, I think.) Matlab and Mathematica would be a better fit for the proposed JuliaInterop org, if that comes to fruition. Edit: Looks like it probably won't. But I'd still consider those packages outside of our scope.

[ChrisRackauckas]

I agree with the Interop. As for stochastic processes, I think that would go with math instead of stats. Stats uses stochastic processes and many statistical tests/algorithms build on ideas from stochastic processes, but the core of stochastic processes is still over to math. I think that's why you'd find people researching stochastic processes more centered in mathematics departments (though they have a large spread through other fields as well).

I think we need to define a bit about what is JuliaMath. I would be under the impression that it deals with at least everything from "traditional applied math", such as computational PDEs, numerical linear algebra, dynamical systems, and numerical analysis. These days the departments tend to include stochastic differential equations and computational algebra.

Maybe there are two different orgs that should be made: one based on numerics (numerical analysis and numerical linear algebra), and one based on dynamics (ODEs, PDEs, SDEs, control systems, bifurcation analysis, etc.). But if the key idea is to pull them together for cross-contributions, numerical linear algebra should really be kept with PDEs since they are really closely related in practice.

[ararslan]

I agree with your definition for org, though I think numerics and dynamics can coexist here just fine.

With the "computational" aspect in mind, I think that symbolic manipulations and other CAS stuff should not live here. That would include things like SymPy, Nemo, and SJulia (arguably Calculus.jl as well). Hopefully that wouldn't be too controversial for @pabloferz, who specifically mentioned CAS. 😄

[simonbyrne]

Honestly, I would be careful about moving too many repos too early, otherwise orgs have a danger of ending up full of abandonware. I would start with a few (or create the new ones), and go from there.

[pabloferz]

Following @simonbyrne, I think, for the time being, we should focus in accommodate here those parts from Base that fit here.

[simonbyrne]

I would nominate Primes.jl and SpecialFunctions.jl as good places to start.

[musm]

Same with Simon, I feel like this should start of as a small set of curated packages that are essential and of high quality (a la matlab toolboxes) and the maybe go from there.

[mschauer]

then maybe go from there.

Let's not overshoot in the other direction though.

[simonbyrne]

We could also start by moving the relevant ones from the JuliaLang org as well.

[ararslan]

Yes, all of that sounds great to me. When I was naming packages before, I was thinking more long term; I absolutely agree that we should start small.

Perhaps once @mschauer's and @pabloferz's outstanding PRs for primes.jl (JuliaLang/julia#16333 and JuliaLang/julia#16349, respectively) are merged, we can branch out a Primes repo.

[simonbyrne]

Primes.jl has now been created!

[ararslan]

@simonbyrne Awesome, thanks! How will that work with Pablo and M.'s PRs on base Julia?

[pabloferz]

I guess we can close them over there and bring them here.

[simonbyrne]

I've changed the permissions so that all members have commit access, so @pabloferz and @mschauer you should feel free to move your PRs over (and make any other changes you feel deserving).

[tkelman]

Why is Sundials.jl crossed out? I'm about to raise an issue there, either JuliaMath or JuliaODE (which currently only has a fork of ODE.jl) would be good places for that to move to.

[ChrisRackauckas]

It's crossed out because it was being put as an "Interop package". ODEInterface would be in the same boat.

[tkelman]

Well ODEInterface isn't in JuliaLang so it doesn't necessarily need to move except for the sake of making it easier for multiple people to help with maintenance.

[ChrisRackauckas]

Oh, I didn't know Sundials was in JuliaLang.

[ararslan]

@tkelman What do you mean it's crossed out... >_>

[ChrisRackauckas]

ODE.jl should be crossed out because of the new JuliaDiffEq org.

[ararslan]

If there's an organization specifically for differential equations, wouldn't it make more sense for them to have Sundials?

Looks like it was already moved.

[ChrisRackauckas]

@simonbyrne what has happened with the GSoC projects by @sunoru? It seems like VSL.jl and RNG.jl are very good starts, but seem to be abandoned since GSoC ended (1 month of no commits). Would these be good candidates to move here so we could continue to develop them? (RNG.jl in particular looks like a very promising result)

[sunoru]

Hi Chris. I'm so sorry that I have been kind of busy in this new semester, but I haven't abandoned the project. There are many TODOs in RNG.jl.

[ararslan]

@sunoru You're doing great work on that project. I hope you enjoyed GSoC and I'm glad you're still interested in maintaining those packages. Keep up the good work! 👍

[ChrisRackauckas]

Good to hear. I put in some PRs to fix the testing and requirements. I noticed these were odd (not including v0.5, and while requiring only v0.5-) which is why I was wondering if it was still under development.

[sunoru]

@ararslan Thank you very much! I'll keep doing my best.

There are some existing issues and todos. You are welcomed to join the discussion if you are interested.

[ChrisRackauckas]

Closing this since the broad changes have already occurred. Other migrations should be new issues.