Changelog

\[\newcommand {\lt}[1] {\mathsf{#1}} \newcommand {\mbf}[1] {\mathbf{#1}} \newcommand {\es}[1] {\mathbf{e}_{#1}} \newcommand {\til}[1] {\widetilde{#1}}\]

0.6.1 2026.04.04

[Feature] #580: Mlt can now be constructed from a pre-built sympy component expression by passing the expression as f together with an explicit nargs argument. Previously this path raised NotImplementedError; as a secondary effect, __add__(), __sub__(), and related arithmetic on string-constructed tensors now work correctly. See #578.
[Bug] #569: Lt now accepts an ImmutableDenseMatrix as its matrix argument, fixing a TypeError when constructing linear transformations from SymPy immutable matrices. See #567.
[Bug] #570: Fixed a SymPy 1.13 regression where trigsimp() raises ZeroDivisionError on certain trig expressions (e.g. prolate spheroidal coordinates) during LaTeX printing in _latex(). The broad except Exception was narrowed to except ZeroDivisionError only, falling back to the expanded form for display. See #566.
[Bug] #582: is_blade() now correctly identifies null vectors (e.g. e0 + e1 in G(1,1)) and null blades (e.g. (e0+e1)^e2 in G(1,2)) as blades. Previously the method delegated entirely to is_versor(), which returns False for null multivectors because they have no inverse. Blade-ness is a metric-free concept; the fix checks grade homogeneity first (grade ≤ 1 is always a blade) then falls back to the outer-product squaring test B ^ B == 0 for the null case. See #537.
[Bug] #590: Worked around a performance regression in SymPy 1.13 that caused examples/ipython/LaTeX.ipynb (check('curvi_linear_latex')) to time out after 600 s on SymPy ≥ 1.13. SymPy PR #26390 added an O(N·M) .replace() traversal inside TR3/futrig that is a no-op for galgebra’s symbolic trig arguments but dominated each of the ~70 Simp.apply calls during Ga.build(norm=True) for curvilinear coordinates. The fix uses trigsimp(method='old') via Simp.profile for the affected example, cutting run time from > 600 s to < 6 s. A notebook note documents the two cosmetic output differences from the pre-1.13 form; a proper upstream fix is tracked in #576.
[Support] #565: CI now tests on Python 3.10, 3.11, and 3.12; Python 3.8 and 3.9 are dropped (both reached end-of-life). See #564.
[Support] #570: Refreshed notebook outputs for SymPy 1.13 printing changes (\cdot in Mul expressions and column-format specifiers in \begin{array}). Added scripts/validate_nb_refresh.py — a reusable tool that mechanically confirms a notebook re-execution introduced only cosmetic changes. Added doc/dev/bumping-sympy.md developer guide (version policy, local testing, notebook refresh workflow). Bumped test_requirements.txt pin from sympy == 1.12 to sympy == 1.13.3. See #568, #566.
[Support] #572: Added doc/dev/bumping-python.md developer guide for bumping the Python version in CI and a release-process runbook (doc/dev/releasing.md). Updated README Python prerequisites. See #571, #573.
[Support] #577: Updated the bundled doc/books/Macdonald/GAlgebraPrimer.pdf to the September 15, 2023 revision by Alan Macdonald and corrected the download URL in README and example documentation. See #575.
[Support] #585: Expanded and reorganised is_blade() test coverage: replaced a single ad-hoc test with nine focused methods covering every branch of the fixed implementation, including null vectors, null bivectors, non-homogeneous multivectors, the cached-flag path, and the null-blade guards added to reflect_in_blade() and project_in_blade(). Added a slow-marked test that documents the known grade-≥ 3 limitation (the B ^ B == 0 outer-product test can return a false positive for non-blades in spaces of dimension ≥ 6 without coordinates); the R⁶ DFM counterexample is preserved as an executable regression guard. Registered the slow pytest marker in setup.cfg. See #537.
[Support] #587: Migrated from setup.py to pyproject.toml (PEP 517/621). The build backend is setuptools>=61; _version.py remains the single source of truth for the version. The CI publish job now uses python -m build instead of python setup.py sdist bdist_wheel. setup.py is removed; setup.cfg is retained for flake8 configuration. See #586.
[Support] #589: Added Step 0 to the release-process runbook (doc/dev/release-process.md): open a release issue before preparing the changelog or branching, so that motivation, scope, and unusual pre-release todos are recorded. See #588.

0.6.0 2026.04.01

[Feature] #510: Added example notebooks for typical GAs by Greg Grunberg, and they are now part of the tests:
- Standard 2D Model $\mathcal{G}_2$: g2
- Standard 3D Model $\mathcal{G}_3$: g3
- Standard 4D Model $\mathcal{G}_4$: g4
- 3D Homogeneous Coordinates $\mathcal{G}_4$: h3
- 3D Conformal Model, Amsterdam convention $\mathcal{G}_{4,1}$: cm3
- Spacetime algebra: spacetime
- Sphere-related algebras:
  - Geometric algebra for unit sphere in $\mathbb{R}^3$ using spherical coordinates: sp2
  - Unit sphere $\mathbb{R}^3$ as a submanifold of $\mathcal{G}_3$ in cartesian coordinates: sp2g3
  - A geometric algebra for the unit sphere in $\mathbb{R}^3$ as a submanifold of $\mathbb{R}^3$ with spherical coordintes: sp2sp3
  - Spherical Coordinates in $\mathbb{R}^3$: sp3
[Feature] #510: Improved functionality for Mv, Lt by Greg Grunberg, with improved test coverage and minor fixes.

Changes by Greg Grunberg include (see the attached notebooks in #478 for details):
- The following methods of Mv are added, along with their corresponding functions (applied to multivectors A and B):
  - Undualization: A.undual() and undual(A)
  - Grade involution: A.g_invol() and g_invol(A)
  - Clifford conjugation: A.ccon() and ccon(A)
  - Scalar product: A.sp(B) and sp(A,B)
  - Quadratic form: A.qform() and qform(A)
  - Magnitude squared: A.mag2() and mag2(A)
  - Magnitude: A.mag() and mag(A)
- The following methods of Mv are improved, along with their corresponding functions:
  - Norm squared: A.norm2() and norm2(A) now returns the absolute value of the quadratic form of A.
  - Norm: A.norm() and norm(A)
- Lt is significantly improved and fixed, see also Linear Transformations / Outermorphisms in GAlgebra:
  - Lt adopts the contravariant-covariant indexing notation long used in tensor analysis and differential geometry, which is consistent with the indexing notation GAlgebra already uses for the display of a multivector’s basis blade expansion
  - Symbolic_Matrix() allows the creation, for a symbolic transformation $\lt{T}$, of matrices with entries of any of the four forms ${T^i}_j$, $T^{ij}$, $T_{ij}$, or ${T_i}^j$, although GAlgebra will use only the first two.
  - Dictionary_to_Matrix() fixes a bug that incorrectly raises an exception if T maps one or more of the basis vectors to the zero multivector, and improves the readability.
  - Lt fixes a bug that erroneously post multiplies the transformation’s standard matrix by self.Ga.g_inv (the reciprocal metric tensor), resulting in a contravariant-contravariant matrix $[T^{ij}] = [{T^i}_k g^{kj}]$ instead of the standard matrix $[{T^i}_j]$. In the same spirit, matrix() now returns the standard matrix $[{T^i}_j]$ instead of the product matrix $[{T^i}_j][g_{ij}]$.
  - Lt now distinguishes symmetric and antisymmetric transformations from general transformations, Amat will be the standard matrix $[{T^i}_j]$ of the transformation when mode=='g', , but will be $[T_{ij}]$ when mode in ['s','a']. Since $[g^{ij}][T_{ij}] = [{T^i}_j]$, in either case Lt.__init__ will receive the standard matrix as its first parameter.
  - Lt now correctly handles Versor input, and initializes the internal lt_dict for the versor-based linear transformation.
  - Lt adds support for both even and odd versors, after generalization of spinors to versors in Mv.
  - For versor based transformations, the inverse transformation inv() is now based on simply $\til{\mbf{V}}$ instead of $\mbf{V}^{-1} = {\frac 1 {\mbf{V} \til{\mbf{V}}}} {\til{\mbf{V}}}$, as a versor-based transformation is independent of taking nonzero scalar multiples of the versor.
  - Lt adds support for LaTeX printing of versor-based transformations.
  - The determinant method det() for a linear transformation is fixed, it now uses directly the geometric algebra definition of $\lt{L}$’s determinant: $\det(\lt{L}) = \lt{L}(\mbf{E}) \mbf{E}^{-1}$, where $\mbf{E}$ denotes the basis blade $\mbf{E} = \es{1} \wedge \cdots \wedge \es{n}$ for the grade space of pseudoscalars.
Minor fixes includes:
- norm is fixed by using metric.square_root_of_expr instead of sqrt
- making a “spinor” mv is kept for backward compatibility with existing tests, which is still aliased to making an “even” mv, see discussions 10. An unimplemented suggestion in Changes to module lt.py attached in #478
- fixed linting errors reported by flake8
- fixed some typos of return, raise etc.
[Feature] #510: Added gprint by Alan Bromborsky developed back in 2020, which improved printing of text and LaTeX for multiple GA objects in Jupytper notebooks, and is the recommended way to print GA formulas, see Demo of gprint for a demonstration. Testability and test coverage are improved.
[Feature] #530: Added shirokov_inverse() and hitzer_inverse() (and their module-level counterparts shirokov_inverse(), hitzer_inverse()) for computing the multivector inverse using the Shirokov and Hitzer algorithms.
[Feature] #543: Mv now supports scalar / Mv via __rtruediv__(), so expressions like 1 / spinor work without explicitly calling inv(). See #512.
[Feature] #550: Added galgebra.interop.Cl() and galgebra.interop.kingdon.Cl() as a Cl(p, q, r) interface compatible with kingdon and ganja.js conventions, making it easier to port code between galgebra and other GA libraries. See #524.
[Bug] #511: Mv now correctly handles negative integer power.
[Bug] #513: Mv now correctly handles differentiating by a coordinate symbol.
[Bug] #516: grades no longer incorrectly returns None under some circumstances, as now all initialization branch will correctly call characterise_Mv().
[Bug] #518: Mv now correctly returns Mv instance when raise to power of zero. But in general, if one needs to call Mv methods on a result returned by some GA operations, it would be more prudent to initialize it as an Mv instance first, as sometimes the result becomes a sympy object.
[Bug] #540: Lt() callable constructor now accepts the zero multivector as a return value, so projection maps (e.g. lambda x: (x | e1) * e1) no longer raise ValueError.
[Bug] #536: is_versor() is improved following GSG’s analysis: it now checks that (1) the scalar product with the reverse is a nonzero scalar and (2) the grade-involution sandwich maps every basis vector to a vector. Handles degenerate metrics (e.g. PGA G(2,0,1)) correctly. See #533.
[Bug] #554: norm() and norm() now wrap the result in Abs() for even-grade multivectors, fixing incorrect simplification to negative values. See #522.
[Bug] #556: galgebra.interop.Cl() now resets the global dual mode to galgebra’s default ('I+') after each call, preventing silent contamination when calls to galgebra.interop.Cl and galgebra.interop.kingdon.Cl are interleaved. See #555.
[Support] #510: Testing infrastructure for generating and diffing PDFs is added, see test/test_gprinter.py and .github/workflows/ci.yml for details.
[Support] #535: Fixed CI and linting issues.
[Support] #552: Refactored internals for clarity and correctness:
- er_blade() delegates to Mul instead of hand-rolling the same logic (#140).
- ReciprocalFrame loop rewritten with a cleaner index convention matching the sign formula in the reference text (#249).
[Support] #545: Added a regression test for the spurious \cdot in LaTeX output, confirming the fix documented in #494 holds.
[Support] #558: Added regression tests for matrix() on oblique and non-Euclidean metrics, and for the generic lt() constructor, confirming the fixes from #461 hold.
[Support] #548: Improved README to document the full set of available GA operations. See #523.
[Support] #551: Added examples/ tutorial demonstrating how to extract scalar coordinates from a multivector using get_coefs() and blade_coefs(). See #483.
[Support] #549: Added examples validating galgebra against Russell Goyder’s sundial analysis and geometric algebra cheat sheet (#506):
- examples/Terminal/sundial.py: rotors, reflections, projections, rejections, and the dual in G(3,0) as used in the sundial problem.
- test/test_cheatsheet.py (#557): pytest tests for all 7 sections of the cheat sheet — geometric product decomposition, vector-multivector products, bivector identities, commutator product, pseudoscalar, outermorphism, and adjoint.

0.5.2 2024.05.01

[Support] #517: Add citation info, star history, contributors to README, and fix zenodo citation issue for 0.5.1.

0.5.1 2024.03.31

[Feature] #436: The Ga constructor now allows algebras with only a single basis vector, via a trailing comma in the list of bases. This enables algebras like the complex (Ga('i,', g=[-1])) and dual (Ga('delta,', g=[0])) numbers to be used.
[Feature] #450: ANSI printing can now be turned off with Eprint(deriv=None, fct=None, base=None).
[Feature] #454: galgebra.lt.Lt objects can now be used with Gas that are not constructed with a coords argument.
[Feature] #492: (also #473) Support Python 3.8-3.11, drop support for earlier Python versions.
[Feature] #493: SymPy 1.9-1.12 are tested and supported.
[Bug] #448: Calling galgebra.printer.Eprint() no longer causes ANSI escape codes to appear in the names of the coefficients of ga.mv('A', 'vector').
[Bug] #467: Basis vectors are not normalized in Metric with norm=True when derivatives of basis vectors aren’t available.
[Bug] #468: Hyperbolic functions now requires explicit collect and trigsimp to simplify unlike SymPy 1.6.1 and before.
[Bug] #494: There is an extra \cdot in LaTeX output for multiplication between a numeric power and a polynomial with a leading numeric coefficient since SymPy 1.10. This has to be fixed by SymPy, before that, be cautious when seeing unexpected \cdot in LaTeX output.
[Bug] #495: MatrixFunction is broken since SymPy 1.11, which is required by initializing Ga with gsym and coords, this is now fixed with a workaround (#507).
[Support] #450: Many undocumented parts of galgebra.printer.Eprint() have been removed:
- The debug argument to Eprint, which just printed out the other arguments.
- The on argument to Eprint, which if False just made the call do nothing at all. Use if cond: Eprint(...) instead of Eprint(..., on=cond).
- The overrideable default settings Eprint.defaults.
- The constant dictionaries Eprint.ColorCode and Eprint.InvColorCode.
- The class methods Eprint.Base, Eprint.Fct, Eprint.Deriv, and Eprint.Strip.
- The class attributes Eprint.base, Eprint.fct, Eprint.deriv, and Eprint.normal.
If you were relying on these details to implement your own ANSI printing, it is recommended that you use a package like colorama instead.
[Support] #454: The l.coords and l.X attributes of galgebra.lt.Lt objects are deprecated in favor of using l.Ga.coords and l.Ga.coord_vec.
[Support] #455: The static method format and the attribute Lt.mat_fmt of galgebra.lt.Lt are removed as they were never used.
[Support] #457: Line-wrapping of the plaintext printing are disabled by default, as it makes reviewing the output easier, to re-enable it, call sympy.init_printing with wrap_line=True.
[Support] #458: The output of Lt objects now use $\mapsto$ instead of naming the transformation.
[Support] #459: Various cleanups and fixes to Lt:
- Construction of Lt from a linear function is fixed
- The broken construction of Lt from a tuple is removed
- The code of Dictionary_to_Matrix and Matrix_to_dictionary is improved
- Properties of Lt are tidied, documented, and better tested (#460)
[Support] #476: Make tests more robust by directly testing the printer, rather than __str__.
[Support] #487: CI has been changed to Github Actions.
[Support] #497: The documentation is now built with the latest Sphinx (7.2.6) and MathJax 3.
[Support] #503: The README now supports Github dark mode, and is included by sphinx_mdinclude in Sphinx with improved support for TOC and math formulas.

0.5.0 2020.06.05

[Feature] #326: The galgebra.ga.Ga.make_grad() function now accepts multivectors, not just vectors.
[Feature] #243: galgebra.mv.Mv.subs() now accepts all the same arguments as sympy.subs().
[Feature] #252: (also #310) Many attributes of Ga instances are now _always_ present, rather than being conditionally present depending on the type of algebra. These include:
- r_basis
- r_basis_mv
- r_basis_dict
- g_inv
- g_adj
Other properties that are still only meaningful for some algebras now exist but raise clearer error messages:
- coord_vec
- bases
- basic_mul
These lists are not exhaustive.
[Feature] #336: The scalar product $a * b$ is now available via galgebra.ga.Ga.scalar_product(). Note that there is no operator overload on multivectors yet.

[Feature] #364: The galgebra print customizations Dmode, Fmode, fmt, and Mlt.lcnt can now be set via:

sympy.init_printing(
    omit_function_args=True,  # Fmode
    omit_partial_derivative_fraction=True,  # Dmode
    galgebra_mv_fmt=1,  # fmt
    galgebra_mlt_lcnt=1,  # lcnt
)

These names are provisional, and may change in future.

[Feature] #364: The galgebra printer now respects the global inv_trig_style sympy print setting.
[Feature]: Generally improved documentation, including:
- The re-inclusion of missing code examples that were already referenced from the text (#293).
- Fixed footnote and citation references (#287, #290, #291)
- Faster and smaller pages with less MathJax loading time (#292, #296).
- An adapted version of a notebook tutorial from Alan Bromborsky is now present in the docs, Introduction to using GAlgebra (#375).
- Type annotations on many functions in the API docs (#305).
- Correction of curly quotes that led to invalid python syntax (#294).
[Feature] #212: Python 2.7 is no longer supported, allowing more features from Python 3.x to be used.
[Feature] #232: Python 3.8 is now supported and tested.
[Bug] #344: galgebra.metric.collect() no longer discards terms that were not requested.
[Bug] #53: Calling sympy.sympify() (or any other sympy function) on a Mv instance no longer raises RecursionError, and instead raises TypeError with a helpful message.
[Bug] #258: The result of simplifying sympy expressions is no longer dependent on whether galgebra.printer.Format() has been called.
[Bug] #264: Dop no longer emits latex strings when printed in non-latex mode.
[Bug] #319: galgebra.ga.Ga.make_grad() no longer crashes when called on an algebra with no coords.
[Bug] #319: galgebra.ga.Ga.make_grad() no longer has a broken cache that ignores cmpflg if both the left and right gradient operator are requested for the same vector
[Bug] #319: galgebra.mv.Mv.get_coefs() now returns 0 in the place of empty coefficients.
[Bug] #320: galgebra.lt.Mlt no longer crashes at construction, arithmetic, or multiplication.
[Bug] #323: (also #377, #340) Many non-Mv objects now render better in the default sympy printer. This bug manifested most often when using Mv.obj, and would result in misrenderings like $e^e_{xy}$ when $e_x \wedge e_y$ was intended, or $e_{x.y}$ when $e_x \cdot e_y$ was intended.
[Bug] #345: e_sq no longer contains a floating point term.
[Bug] #348: galgebra.printer.Print_Function() no longer emits code after the function body. Previously, examples had a def dummy(): pass function to work around this.
[Bug] #354: galgebra.printer.oprint() no longer strips the last ) from sympy.Matrix objects within lists.
[Bug] #369: The fmt argument fo the Fmt method of Mv, Dop, Lt, Mlt no longer has side effects on subsequent print statements.
[Bug] #369: (also #380) The Fmt method of Mv, Dop, Lt, Mv now works properly in both IPython (giving plaintext output) and Jupyter (giving LaTeX output).
[Bug] #372: For scalar multivectors, the printed result is now mathematically equivalent in plaintext and latex mode.
[Bug] #378: (also #379) The post-processing performed by galgebra.printer.tex() is now better-behaved:
- It no longer translates grad into $\nabla$ if it is part of a longer word like gradual.
- It no longer mangles occurrences of @@.
- It no longer throws away # and % symbols that appear in the middle of lines.
- It no longer injects stale information from previous lines into subsequent lines.
- It now strips all leading and trailing whitespace before wrapping in equation or align.
[Bug] #382: The result of calling galgebra.printer.latex() on a multi-vector when the global galgebra_mv_fmt setting is not 1 is now valid to use within math mode.
[Bug] #387: galgebra.printer.Print_Function() no longer emits invalid LaTeX with an unwanted \\
[Bug] #391: Subtracting a scalar from a Mv instance no longer fails.
[Bug] #431: Left and right contraction are no longer swapped on scalar Expr instances.
[Support] #64: The following attributes have been deprecated in favor of using the new flat attribute:
- galgebra.ga.Ga.indexes_lst → galgebra.ga.Ga.indexes .flat
- galgebra.ga.Ga.bases_lst → galgebra.ga.Ga.bases .flat
- galgebra.ga.Ga.blades_lst → galgebra.ga.Ga.blades .flat
- galgebra.ga.Ga.mv_blades_lst → galgebra.ga.Ga.mv_blades .flat
Since the replacement properties also include the scalar element, ga.blades_lst is equivalent to ga.blades.flat[1:] (and likewise for the other properties).
[Support] #245: (also #248, #334) The following attributes have been deprecated to reduce the number of similar members in Ga.
- Unified into the table_dict attribute of ProductFunction instances:
  - galgebra.ga.Ga.wedge_table_dict → wedge.table_dict
  - galgebra.ga.Ga.left_contract_table_dict → left_contract.table_dict
  - galgebra.ga.Ga.right_contract_table_dict → right_contract.table_dict
  - galgebra.ga.Ga.dot_table_dict → hestenes_dot.table_dict
  - galgebra.ga.Ga.mul_table_dict → mul.table_dict
  - galgebra.ga.Ga.basic_mul_table_dict → basic_mul.table_dict
- Unified into methods of ProductFunction instances:
  - galgebra.ga.Ga.non_orthogonal_bases_products → basic_mul.of_basis_bases
  - galgebra.ga.Ga.wedge_product_basis_blades → wedge.of_basis_blades
  - galgebra.ga.Ga.geometric_product_basis_blades → mul.of_basis_blades
  - galgebra.ga.Ga.dot_product_basis_blades and galgebra.ga.Ga.non_orthogonal_dot_product_basis_blades → <some_dot_type>.of_basis_blades. In future it will no longer be possible to call the non-ortho method for an orthogonal ga, or vice-versa.
- Unified into galgebra.ga.Ga.blade_expansion_dict:
  - galgebra.ga.Ga.blade_expansion → blade_expansion_dict.items()
- Unified into galgebra.ga.Ga.base_expansion_dict:
  - galgebra.ga.Ga.base_expansion → base_expansion_dict.items()
- Unified into galgebra.ga.Ga.basic_mul_table_dict:
  - galgebra.ga.Ga.basic_mul_table → basic_mul_table_dict.items()
  - galgebra.ga.Ga.basic_mul_keys → basic_mul_table_dict.keys()
  - galgebra.ga.Ga.basic_mul_values → basic_mul_table_dict.values()
- Unified into galgebra.ga.Ga.indexes_to_bases_dict:
  - galgebra.ga.Ga.indexes_to_bases → indexes_to_bases_dict.items()
  - galgebra.ga.Ga.bases_to_indexes → indexes_to_bases_dict.inverse.items()
  - galgebra.ga.Ga.bases_to_indexes_dict → indexes_to_bases_dict.inverse
- Unified into galgebra.ga.Ga.indexes_to_bases_dict:
  - galgebra.ga.Ga.indexes_to_blades → indexes_to_blades_dict.items()
  - galgebra.ga.Ga.blades_to_indexes → indexes_to_blades_dict.inverse.items()
  - galgebra.ga.Ga.blades_to_indexes_dict → indexes_to_blades_dict.inverse
[Support] #280: The galgebra.utils module, which provided Python 2 compatibility helpers, has been deprecated.
[Support] #335: The undocumented method galgebra.ga.Ga.X() (the same as coords_vec) has been deprecated.
[Support] #338: The undocumented and mispelt static method galgebra.lt.Mlt.extact_basis_indexes() (which just computed a value equivalent to galgebra.ga.Ga.basis_super_scripts) has been deprecated.
[Support] #202: Pdop and Sdop instance are no longer associated with a Ga. As a result, their .Ga attribute no longer exists, and the pdop() and sdop() methods of Ga are deprecated in favor of calling the constructors directly. For Ga-aware operators, continue to use Dop.
[Support] #216: galgebra.metric.test_init_slots has been removed. The functionality this provided is superseded by the language feature of keyword-only arguments.
[Support] #252: (also #310) The inverse_metric() and derivatives_of_g() methods of Metric are deprecated, as the properties they computed (g_inv and dg) are now computed automatically.
[Support] #259: galgebra.printer.oprint now aligns results in columns
[Support] #284: (also #317) galgebra.lt.Lt.setup() has been deprecated, along with the attributes it populated galgebra.ga.Ga.lt_x (the same as coords_vec) and galgebra.ga.Ga.lt_coords (the same as coords).
[Support] #286: The galgebra.deprecated module, which provides old-style APIs, now emits DeprecationWarnings.
[Support] #320: The following attributes of galgebra.ga.Ga have been removed:
- a
[Support] #349: galgebra.printer.Get_Program() is deprecated. This function used to have to be called before galgebra.printer.Print_Function(), but now all it does is act as an enable/disable flag for the latter, such that Get_Program(True); Print_Function() was a no-op.
[Support] #358: galgebra.printer.find_executable has been removed, as the functionality is provided by shutil.which().
[Support] #367: (also #364, #369) The following properties of GaLatexPrinter and GaPrinter have been removed:
- fmt
- prev_fmt
- dop_fmt
- prev_dop_fmt
- lt_fmt
- prev_lt_fmt
- Dmode
- Fmode
- inv_trig_style
- dop
[Support] #367: The fmt_dop argument to and dop_fmt attribute of galgebra.mv.Dop have been removed, as they had no effect.
[Support] #376: (also #258, #371) The repr() and str() of both galgebra and sympy objects will no longer ever return latex strings, instead always returning plaintext strings. If you want latex strings, use galgebra.printer.latex(). Note that this does not affect support for print(sympy_object), which continues to print latex representations provided galgebra.printer.Format() has been called.
[Support] #384: The dop argument to Format(), along with the corresponding static member of LatexPrinter, has been removed, as had no effect.
[Support] #385: galgebra.printer.print_replace has been removed.
[Support] #386: The title attribute of Dop has been removed. as it was always None.
[Support] #393: galgebra.dop.Sdop.str_mode has been removed, as it was always False.

0.4.5 2019.12.31

[Feature] #50: (also #51, #56) Improve documentation formatting:
- LaTeX and code samples are now appropriately formatted
- Attributes of classes now have permalinks
[Feature] #66: The mul_table_dict table, and the equivalent tables for the other products, are now computed lazily when indexed. These are now all documented too.
[Feature] #66: Remove unused or unusable code in the public API:
- Ga.mul_table, Ga.wedge_table, Ga.dot_table, Ga.left_contract_table, and Ga.right_contract_table, all of which were the empty list, [].
- galgebra.mv.modules, a string which served no purpose (#71).
- __add_ab__, __sub_ab__, __mul_ab__, and __div_ab__, none of are real magic method names (#67). No code should be calling these directly anyway.
- Dop.flatten_one_level which is better spelt itertools.chain.from_iterable (#175).
- Dop.basic, which was a non-working version of grads() (#185).
- Pdop.setGa, Sdop.setGa, Dop.setGa, all of which rely on dangerous global state - use the ga keyword argument of the constructors of these types instead (#163).
[Feature] #78: grads() now raises a better error when it fails, and is faster.
[Feature] #164: (and #169, #170) Sympy 1.5 is officially supported and tested.
[Feature] #172: galgebra.__version__ has been added, which contains the version string.
[Feature] #195: Deprecate confusing parts of the public API:
- galgebra.ga.Ga.mv_I(), which was a bad way to spell galgebra.ga.Ga.E().
- galgebra.ga.Ga.mv_x(), which was a bad way to create a vector with a weird name.
- Pdop(None), which is better spelt Pdop({}), the latter of which has always worked (#187).
- galgebra.ga.Ga.Pdop_identity(), which is better spelt ga.pdop({}) (#194).
- galgebra.ga.Ga.Pdiffs(), which is better spelt ga.pdop(x) (#194).
- galgebra.ga.Ga.sPds(), which is better spelt ga.sdop(x) (#194).
[Bug] #60: (also #141) Make the following operations on galgebra.mv.Mv non-mutating:
- blade_rep()
- base_rep()
- diff()
- simplify()
- expand()
- collect()
- print(mv)
Any code relying on this behavior will need to change from x.method() to x = x.method(). Note that the latter syntax was always supported even before this change.
[Bug] #61: Make contraction and Hestenes dot products thread-safe. Previously these relied on the dot_mode setting not being changed mid-operation. The dot() method still respects this setting, but is no longer used internally.
[Bug] #80: (also #57, #58, #97) The galgebra.mv.Mv constructor no longer silently accepts illegal arguments, and produces better error messages.
[Bug] #81: (also #180) Passing coefficients as Mv(coefs, 'odd', ga=ga) is forbidden.
[Bug] #90: galgebra.ga.Ga.blades, galgebra.ga.Ga.bases, and galgebra.ga.Ga.indices now reference the scalar S(0) as the single grade-0 object. Previously they listed no grade 0 objects.
[Bug] #134: dot_table_dict now contains correct values (zero) for scalar keys
[Bug] #151: (also #150) Dop, Sdop, and Pdop no longer have mutating methods. This fixed issues where not only would the laplacian be sometimes calculated incorrectly, but its correctness would vary depending on whether it had been printed!
[Bug] #165: galgebra.metric.linear_expand() no longer accepts a mode argument, as this did not work properly. For the old behavior of linear_expand(x, mode=True), use linear_expand_terms(x) instead.
[Bug] #177: Dop objects that evaluate to 0 no longer raise cryptic ValueErrors when operated on.
[Bug] #151: Dop([], ga=ga) and Sdop([], ga=ga) now evaluate to multiplication by zero, not by one. Multiplication by one can as always be spelt Dop([(S(1), Pdop({}, ga=ga}))], ga=ga).
[Bug] #157: Adding an Sdop instance to a scalar now works as intended, rather than crashing.
[Bug] #183: op1 == op2 now works correctly for Sdop and Dop instances.
[Bug] #187: Pdop no longer silently accepts extra arguments.
[Bug] #187: Sdop(x) now functions as intended as an alias of Sdop([(S(1), Pdop({x: 1}))]), which previously crashed.
[Bug] #188: Passing a ga keyword-argument to sm(), mv(), and lt() is now an error, previously it was silently ignored.
[Bug] #188: pdop() and sdop() are now transparent aliases to Pdop and Sdop, respectively. Previously, they would crash.
[Support] #46: (also #69) Remove unnecessary executable bit from importable python files, and the corresponding no-op code that would be run.
[Support] #55: Rename *kargs to *args internally, to match convention. This has no effect on callers, but makes the docs and source easier to read.
[Support] #59: (also #65) Make internal attributes and helper functions private. All of the following have been made private by being renamed:
- Mv.make_grad (#59).
- Mv.make_scalar (#59).
- Mv.make_vector (#59).
- Mv.make_bivector (#59).
- Mv.make_pseudo_scalar (#59).
- Mv.make_multivector (#59).
- Mv.make_spinor (#59).
- Mv.make_odd (#59).
- Ga.build_bases() (#65).
- Ga.basis_product_tables() (#65).
- Ga.build_reciprocal_basis() (#65).
- Ga.build_connection() (#65).
- Ga.non_orthogonal_mul_table() (#65).
- Ga.base_blade_conversions() (#65).
- Ga.init_connect_flg() (#65).
- Ga.derivatives_of_basis() (#65).
- Ga.lt_flg (#65).
- Ga.agrads (#65).
- Ga.dbases (#65).
- Ga.XOX (#195).
[Support] #72: Other internal cleanup.
[Support] #167: Python 3.4 is no longer supported.
[Support] #83: This is the last release to support Python 2.7.

0.4.4 2019.09.30

[Feature] #9: Documentation now available at https://galgebra.readthedocs.io/
- Convert doc to Sphinx with the help of pandoc, notedown and nbsphinx
- Add Getting Started guide to README
- Update installation instructions in README
- Add migration guide from sympy.galgebra and brombo/galgebra
- Add Changelog
- Add doc for examples, tests and bundled resources
[Feature] #17: Fix examples under both Python 2 & 3
- Fix examples/* and verify them in CI using nbval
- Test coverage increased from 48.89% to 66.52%
- Add README for test and examples
[Bug] #31: Freeze the depended version of SymPy to 1.3
[Bug] #32: Fix calculation of reciprocal basis for non-orthogonal basis
[Bug] #30: Fix bugs of using LaTeX as symbol name
[Bug] #29: Fix that sometimes plain text is printed with or instead of LaTeX in Jupyter Notebooks
[Bug] #27: Fix broken class MV
[Bug] #26: Fix calculation of the Christoffel symbols
[Bug] #18: Fix TypeError of unicode string, use BytesIO instead of StringIO
[Bug] #15: Fix printing of some products and inverses of multivectors
[Support] #17: Setup Circle CI to build more Python versions faster
- TravisCI build for PRs is now removed

0.4.3 2018.02.18

[Feature] #2: Support Python 3
- Convert code to be Python 3 compatible
- Pass tests under both Python 2 & 3
- Support installing from PyPI instead of setting pth
- Support importing with from galgebra.<package name> import *
[Bug] #2: Fixes brombo/galgebra#20: Incorrect LaTeX output in test_derivatives_in_spherical_coordinates
[Bug] #2: Fixes brombo/galgebra#19: Failures in test_noneuclidian_distance_calculation
[Support] #2: Clean up obsolete code in setup.py and make it simple
[Support] #2: Remove .pyc files and add standard .gitignore for python
[Support] #8: Remove NumPy dependency
[Support] #8: Validate existing Jupyter notebooks using nbval
[Support] #8: Add test coverage in CI using using pytest and CodeCov
[Support] #2: Setup Travis CI