Hi all, Yesterday I pushed out the 0.7.0 release of python-flint to PyPI and GitHub.
This release was delayed for some time because I was hoping to work a bit on tidying up some of the newly added interfaces. Recent reported issues show that people were clearly just using the alpha releases though and now that FLINT 3.2.0 is released it seems like a good time just to get the release out with all the new features and improvements. This release contains work from over 100 pull requests from 7 contributors and has some major changes and new features: - finite fields - multivariate polynomials - acb_theta - an experimental interface for generics (providing nf, qqbar, calcium, ...) - new build system: meson instead of setuptools - new platforms: Linux on ARM and free-threaded CPython There are binaries on PyPI for CPython 3.11, 3.12, 3.13 and 3.13t (free-threading) for: Windows x86-64 MacOS x86-64 and arm64 Linux (manylinux) x86-64 and aarch64. For any of those platforms you should just be able to do: $ pip install python-flint. Those binaries bundle the latest FLINT 3.2.0 which provides many improvements in FLINT itself since 3.0.1 which was bundled by python-flint 0.6.0. Possibly it will take a little time for a new conda package to appear because the build system has switched from setuptools to meson which has better handling of detecting compilers, finding C libraries, parallel builds etc. If you have C compilers and pkgconfig can find FLINT then it should be as simple as pip install python-flint to build from source as well. FLINT versions 3.0-3.2 are supported but some features will be disabled if built for older FLINT versions. After install you can run the tests with: $ python -m flint.test Those tests take literally a couple of seconds so it is worth running them and please let us know if you see any failures. The Linux on ARM platform is new. The binaries are built and tested in CI but I don't have access to Linux on ARM hardware so I have not done any local testing and I don't know that anyone else has either. I would be interested to hear from anyone just that the test suite (hopefully) passes. CPython 3.13 has a new free threaded 3.13t build which removes the GIL. Binaries for 3.13t are provided and have been lightly tested (there are no tests that actually exercise concurrency). The free threaded build should make it possible to benefit from thread parallelism at the Python level rather than FLINT's internal use of threads. However python-flint does not currently use any locks to guarantee thread-safety for mutable types like matrices and polynomials. Any objects shared between multiple threads should be treated as immutable in the free threaded build. I have confirmed that setting an element of an fmpz_mat like M[i,j] = 10**100 can cause a segfault if done simultaneously from multiple threads. Provided shared objects are not mutated there are no known thread safety problems. It is possible that in future we could have safer interfaces for multithreading (immutable matrices?) but for now thread safety works on a consenting adults basis. Contributors ---------------- Many thanks to the following people who contributed to this release: Jake Moss Giacomo Pope Joris Roos Edgar Costa Frédéric Chapoton Oscar Benjamin Tom Hubrecht Changes ------------ The full changelog is in the README: https://github.com/flintlib/python-flint?tab=readme-ov-file#070 The documentation is now hosted on readthedocs: https://python-flint.readthedocs.io/en/latest/ There are a lot of changes and new features in this release. There are many small changes like adding new functions, making the methods more consistent across types, improved error handling, preventing crashes and general code improvements. There are also some big new features: Finite fields and univariate polynomials over finite fields have been added in work by Giacomo Pope. This uses the fq_default interface in FLINT and so it wraps all the different finite field types (fq_nmod, fq_zech, etc). Example GF(9): >>> import flint >>> ctx = flint.fq_default_ctx(3, 2) >>> z = ctx.gen() >>> (1 + z)**5 z + 1 Multivariate polynomials over integers, rationals and integers mod n have been added in work by Jake Moss adding new fmpz_mpoly, fmpq_mpoly, nmod_mpoly and fmpz_mod_mpoly types. Example ZZ[x,y]: >>> R = flint.fmpz_mpoly_ctx.get(['x', 'y']) >>> R fmpz_mpoly_ctx(2, '<Ordering.lex: 'lex'>', ('x', 'y')) >>> x, y = R.gens() >>> p = x**2 - y**2 >>> p x^2 - y^2 >>> p.factor() (1, [(x + y, 1), (x - y, 1)]) FLINT's acb_theta module was added by Edgar Costa: >>> from flint.types.acb_theta import acb_theta >>> A = flint.acb_mat([[1+1j]]) >>> acb_theta(A, A) [[-21.1408540317901 +/- 2.32e-14] + [+/- 4.02e-15]j, ... There is a new experimental interface for FLINT's generic types added by myself and subsequently improved by Tom Hubrecht. This exposes many FLINT types that do not otherwise have wrappers in python-flint such as fmpzi, nf, qqbar, calcium etc as well as polynomials, series and multivariate polynomials over all of these types (no matrices yet...): Example calcium and multivariate polynomials over calcium: >>> import flint.types._gr as gr >>> C = gr.gr_complex_ca_ctx.new() >>> C.sqrt(2) 1.41421 {a where a = 1.41421 [a^2-2=0]} >>> C.sqrt(2)**2 2 >>> R = gr.gr_gr_mpoly_ctx.new(C, ['x', 'y']) >>> x, y = R.gens() >>> x + R(C.sqrt(2))*y x + (1.41421 {a where a = 1.41421 [a^2-2=0]})*y The generic interface here is very much an early work in progress. This is also under active development in FLINT itself so make sure to use FLINT 3.2.0 if testing this. Don't be surprised if some things don't work that well at the moment but please do open GitHub issues for any improvements that would make it more immediately usable. Future plans ----------------- Going forwards I think that the big things to do are: - Complete the generics interface and make it well tested. - Make the generic interface more intuitive to use like ZZ.poly_ring() etc. - Restructure the codebase so that most types are based on the generic interfaces. - Add other major non-generic FLINT features like fq_embed etc. - Add better conversions/coercions both within python-flint and also with other types/libraries. - Add type stubs for use with downstream Python type checkers. -- Oscar Benjamin -- You received this message because you are subscribed to the Google Groups "sage-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to sage-devel+unsubscr...@googlegroups.com. To view this discussion visit https://groups.google.com/d/msgid/sage-devel/CAHVvXxRDxgn200SG%3DfRuXoFumhSxtC%3DWPNoxijDfSLQ1Ait5fA%40mail.gmail.com.
