Symbolica 2.0.0

Symbolica 2.0 has been released! 🎉

Symbolica 2.0 is a major update with a redesigned evaluator stack, richer
Python and Rust APIs, and broader support for special functions and notebook
display.

• Added many transcendental and special functions, including gamma, zeta,
  polylogarithms, Bessel functions, Euler gamma, two-argument atan, and Laurent
  series support around poles.
• Reworked expression evaluation around a unified evaluator API, with builder
  patterns in Rust, a simpler Python evaluator constructor, improved error
  handling, and direct expression evaluation through a single constants map.
• Added custom symbol evaluation and series hooks, including tagged external
  functions, arbitrary-precision constants, C++ external function export, and
  automatic dualization support.
• Improved optimized numerical evaluation with SymJIT 2.18, JIT compilation
  options, SIMD always enabled in the SymJIT backend, save/load support for JIT
  evaluators, and faster batch JIT matrix evaluation.
• Added rich output for notebooks: HTML, LaTeX, and formatted display for
  expressions, series, polynomials, matrices, and graphs.
• Added symbol import/export, ASCII export names, wrapped exported rationals,
  and better function-map merging for exported or optimized evaluators.
• Improved polynomial and rational-polynomial workflows with simpler
  conversion APIs, coefficient-based collection, faster factoring/conversion,
  and Hu-Monagan polynomial GCD enabled when expected to be faster.
• Added a Rust prelude with common imports, plus method-style transcendental
  functions through the new TranscendentalFunctions trait.

Breaking/API changes:

• Removed deprecated methods and shortened Python custom symbol keywords:
  custom_normalization, custom_print, and custom_derivative are now
  normalization, print, and derivative.
• Python Expression.evaluate(constants, functions) and
  evaluate_complex(...) were folded into Expression.evaluate(constants).
• Python optimized evaluator construction changed from
  expr.evaluator(constants, functions, params, ...) to
  expr.evaluator(params, functions=..., ...).
• Rust evaluator construction now returns an EvaluatorBuilder; JIT compilation
  now takes JITCompilationSettings.
• Print options changed: square_brackets_for_function was removed in favor of
  function_brackets, and custom_print_mode now accepts structured data.
• Graph.generate(...) renamed external_nodes to external_edges.
• Minimum Rust version is now 1.89.

Check the migration guide for more details.

Symbolica 1.5.1

Symbolica 1.5.1 has been released!

Notable fixes

• Proper handling of special cases for double-float arithmetic
• Fix normalization of addition of a number and a term
• Drastically improved parsing performance and factorization performance
• Support empty evaluator parameters in Python API

Symbolica 1.5.0

Symbolica 1.5.0 has been released!

Notable changes

• JIT compilation of evaluators using symjit by @Shahriar Iravanian
• A new evaluation pipeline that is much faster and uses less memory
• 2x faster arbitrary precision evaluation
• Recursive match and replace and deepest first replacement
• DoubleFloat arithmetic
• Horner form construction

Symbolica 1.4.0

Symbolica 1.4.0 has been released!

Notable changes

• Improved output featuring bracket colouring, division collection and multi-line formatting
• Much faster and more memory efficient optimized evaluator generation
• Fast arbitrary precision evaluation in Python API
• Typst output mode
• Toggle for partial pattern matching
• Symbol aliases
• Builtin abs and if functions

Symbolica 1.3.0

Symbolica 1.3.0 has been released!

Notable changes

• Overall performance improvement due to faster normalization of terms
• Performance improvements of pattern matching due to proofs of structural mismatches
• Experimental sparse matrix library (written by Gregor Kälin)
• Set parameters as real during evaluation, which allows for faster complex evaluators
• Set dual components of parameters to 0, which allows for faster compiled dual evaluation
• Add replace-once option
• Allow for more flexible input of numpy arrays for evaluators

** Upgrade notice **

• The ordering of terms in the output has changed and may be slightly harder to predict. Use to_canonical_string if you need predictable output.

Symbolica 1.2.0

Symbolica 1.2.0 has been released!

Notable changes

• Support setting namespaces per module in Python

Symbolica 1.1.0

Symbolica 1.1.0 has been released!

Notable changes

• Automatic differentation through duals for expression evaluation
• Attach custom data to symbols
• Improve linear system solving performance
• Joint symbol definition for symbols with interdependent normalization functions

Symbolica 1.0.0

Symbolica 1.0.0 has been released! 🎉 🎉

The first major version is out! Thank you to everyone who contributed to making this happen 🙏 All your feedback and contributions are greatly appreciated.

Notable changes

• Parts of Symbolica are open source. Check out numerica and graphica!
• Support functions in most places that used to accept symbols (collect in function, etc)
• Rework ring traits

Upgrade notice

• Some traits got renamed: NumericalFloatLike -> FloatLike, RealNumberLike -> RealLike

Symbolica 0.20.0

Symbolica 0.20.0 has been released!

Notable changes

• Return underdetermined solutions for linear_system_solve
• Comprehensive parsing of Mathematica (InputForm excluding operators such as . and ->, and FullForm), and improved Mathematica output generation
• Tensor canonization now supports nesting in functions
• Floating point numbers are now supported as coefficients for (rational) polynomials

Upgrade notice

• parse, parse_lit, and try_parse now take the optional default namespace as a named argument, e.g. parse!("x", default_namespace="test")

Symbolica 0.19.0

Symbolica 0.19.0 has been released!

Notable changes

• Branching is now supported for evaluators
• SIMD compiled code generation
• Custom tags for symbols
• New attributes: scalar, real, integer, positive
• Expressions can be tested for realness, integerness, etc
• Complex conjugation of whole expressions (including support for evaluation)
• Wildcards now filter based on their attributes
• Customizable logger
• Support for numpy arrays
• Half-edge notation for graph construction and progress callback

Upgrade notice

• Wildcards now filter based on their attributes (symmetric f_ will not match non-symmetric functions)

The Symbolica Python bindings on PyPi will release as soon as the community packages have been upgraded.