Dedalus

What's Changed
* Fixed bug in setting scales with arrays (e.g. when loading data from hdf5) (https://github.com/DedalusProject/dedalus/commit/d2532390b6164e537b9dc41bea359860abac6c83)
* Updated sparse matrix routines for compatibility with scipy v1.12 (https://github.com/DedalusProject/dedalus/commit/1885f707936f058fea3d4ab80ee6c567182843ce)
* Added MPI explicitly to linking arguments during build, to help when not explicitly compiling with mpicc (https://github.com/DedalusProject/dedalus/commit/bbb4f93e204fac08d442ad61c8263732dd033bc5)


**Full Changelog**: https://github.com/DedalusProject/dedalus/compare/v3.0.1...v3.0.2

What's Changed
* Reuse the LU decomposition when solving for left eigenvectors using the sparse solver by csskene in https://github.com/DedalusProject/dedalus/pull/277
* Patch a few distributor methods that weren't updated to v3.0 API and rerun tutorials (c5ae610)

**Full Changelog**: https://github.com/DedalusProject/dedalus/compare/v3.0.0...v3.0.1

This is the first official release of v3 of Dedalus, which entails substantial restructuring from v2 to support:
- curvilinear bases
- tensorial equation entry (div, grad, curl)
- higher-order problem formulations
- problems involving fields with different dimensionalities

Supported bases now include:

- Fourier (real and complex)
- Jacobi (including Chebyshev and Legendre)
- Annulus (using rational Chebyshev functions)
- Disk (using one-sided Jacobi polynomials)
- Sphere (using spin-weighted spherical harmonics)
- SphericalShell (using rational Chebyshev functions)
- Ball (using one-sided Jacobi polynomials)

Other changes are documented in the [Changes from Dedalus v2](https://dedalus-project.readthedocs.io/en/latest/pages/changes_from_v2.html) docs.

**Note**: This is expected to be the last major v2 release. While we plan to continue issuing bugfixes and backporting major performance improvements and minor API additions from v3, we encourage new users to begin using v3.

Feature additions:

- Transposed LU factorizations for the direct solution of bordered problems without Dirichlet preconditioning. This enables the efficient use of e.g. integral boundary conditions (e1541b9ce1b968d8886889f68b513b0b00d2ba0c).
- Interpolation can now be triggered by using a field/operator's `__call__` method, like `u(y='left')` (248922118c7bbbea77e5677de75efe588f258ea3).
- Left eigenvectors now available with the sparse eigenvalue solver (ad264a476dd14909ff32661e536206a7aa229441).
- Simplified stat logging has be backported from v3 (73981d803ce681f746b1a4562e6b82bd0d5c1c61).

Performance changes:

- Improved sparse matrix applications and other performance improvements have been backported from v3 (36c8e14a40aa82f124097144b3670821fe0367f6).

Feature additions:
- New `Legendre` polynomial basis (77d92e5253f2706ea0ebf0e9b99461bbfcc63dd4).
- Problems now support variables marked as constant in the coupled dimension (see 1D Lane Emden example) (77d92e5253f2706ea0ebf0e9b99461bbfcc63dd4).
- New methods for computing CFL times based on diffusive terms in conservative and non-conservative form (6e2cec29999db36e3cb5dfc9573a8c1db290b7ca).
- Grid spacings are now available in the parsing namespace as e.g. `'sx'` for the `'x'` dimension (71219d1a1a67f51feb799aa8179d46765fdbfd38).
- Number of tau rows removed per equation can now be specified with the `'tau'` keyword in `Problem.add_equation` (ce8c1317b13487012e4c518650de48b202fc8fa9).
- New command-line options `get_config` and `get_examples` for copying the configuration file and example scripts (cc0373e742dcaa56c88db4d29a0b841a48b1d5c0).

Behavior modifications:
- Default value of `'ncc_cutoff'` changed from `1e-10` to `1e-6` to produce more banded NCC matrices (7efb88446b806a0dd0ffaec0d5d08fbd7689fb9c)
- Default value of `'entry_cutoff`' changed from `0` to `1e-12` to prevent rounding errors from reducing matrix sparsity (7efb88446b806a0dd0ffaec0d5d08fbd7689fb9c).
- Hermitian symmetry for real problems is now explicitly enforced every 100 iterations by default. This can be modified via the `'enforce_real_cadence'` keyword for IVP solvers (5b3be571679a73ae7e28d77af093c1ba61c0f4e3).

Performance changes:
- Improved strategy for reducing the number of transforms when evaluating RHS terms (0ce1482b3516bcb4052b9babd029c191e04f2da1).