0.7.6.1
This is a small bugfix release over SymPy 0.7.6, primarily to fix issues with printing in the Jupyter notebook.
Changes
Fix pretty printing in the Jupyter notebook and Jupyter qtconsole for Jupyter 4.0.
The deprecated linearization method in the mechanics module no longer fails to execute.
0.7.6
Major changes
New module calculus.finite_diff for generating finite difference formulae approximating derivatives of arbitrary order on arbitrarily spaced grids.
New module physics.optics for symbolic computations related to optics.
geometry module now supports 3D geometry.
Support for series expansions at a point other then 0 or oo. See PR #2427.
Rules for the intersection of integer ImageSets were added. See PR #7587. We can now do things like {2⋅m | m ∊ ℤ} ∩ {3⋅n | n ∊ ℤ} = {6⋅t | t ∊ ℤ} and {2⋅m | m ∊ ℤ} ∩ {2⋅n + 1 | n ∊ ℤ} = ∅
dsolve module now supports system of ODEs including linear system of ODEs of 1st order for 2 and 3 equations and of 2nd order for 2 equations. It also supports homogeneous linear system of n equations.
New support for continued fractions, including iterators for partial quotients and convergents, and reducing a continued fraction to a Rational or a quadratic irrational.
Support for Egyptian fraction expansions, using several different algorithms.
Addition of generalized linearization methods to physics.mechanics.
Use an LRU cache by default instead of an unbounded one. See PR #7464. Control cache size by the environment variable SYMPY_CACHE_SIZE (default is 500). SYMPY_CACHE_SIZE=None restores the unbounded cache.
Added fastcache as an optional dependency. Requires v0.4 or newer. Control via SYMPY_CACHE_SIZE. May result in significant speedup. See PR #7737.
New experimental module physics.unitsystems for computation with dimensions, units and quantities gathered into systems. This opens the way to dimensional analysis and better quantity calculus. The old module physics.units will stay available until the new one reaches a mature state. See PR #2628.
New Complement class to represent relative complements of two sets. See Pr #7462.
New trigonometric functions (asec, acsc), many enhancements for other trigonometric functions (PR #7500).
New Contains class to represent the relation "is an element of" (see PR #7989).
The code generation tools (code printers, codegen, autowrap, and ufuncify) have been updated to support a wider variety of constructs, and do so in a more robust way. Major changes include added support for matrices as inputs/outputs, and more robust handling of conditional (Piecewise) statements.
ufuncify now uses a backend that generates actual numpy.ufuncs by default through the use of the numpy C api. This allows broadcasting on all arguments. The previous cython and f2py backends are still accessible through the use of the backend kwarg.
CodeGen now generates code for Octave and Matlab from SymPy expressions. This is supported by a new CodePrinter with interface octave_code. For example octave_code(Matrix([[x**2, sin(pi*x*y), ceiling(x)]])) gives the string [x.^2 sin(pi*x.*y) ceil(x)].
New general 3D vector package at sympy.vector. This package provides a 3D vector object with the Del, gradient, divergence, curl, and operators. It supports arbitrary rotations of Cartesian coordinate systems and arbitrary locations of points.
Backwards compatibility breaks and deprecations
All usage of inequalities (>, >=, <, <=) on SymPy objects will now return SymPy's S.true or S.false singletons instead of Python's True or False singletons. Code that checks for, e.g., (a < b) is True should be changed to (a < b) == True or (a < b) == S.true. Use of is is not recommended here.
The subset() method in sympy.core.sets is marked as being deprecated and will be removed in a future release (issue). Instead, the is_subset() method should be used.
Previously, if you compute the series expansion at a point other than 0, the result was shifted to 0. Now SymPy returns the usual series expansion, see PR #2427.
In physics.mechanics, KanesMethod.linearize has a new interface. Old code should be changed to use this instead. See docstring for information.
physics.gaussopt has been moved to physics.optics.gaussopt. You can still import it from the previous location but it may result in a deprecation warning.
This is the last release with the bundled mpmath library. In the next release you will have to install this library from the official site.
Previously lambdify would convert Matrix to numpy.matrix by default. This behavior is being deprecated, and will be completely phased out with the release of 0.7.7. To use the new behavior now set the modules kwarg to [{'ImmutableMatrix': numpy.array}, 'numpy']. If lambdify will be used frequently it is recommended to wrap it with a partial as so: lambdify = functools.partial(lambdify, modules=[{'ImmutableMatrix': numpy.array}, 'numpy']). For more information see #7853 and the lambdify doc string.
Set.complement doesn't exists as an attribute anymore. Now we have a method Set.complement(<universal_set>) which complements the given universal set.
Removed is_finite assumption (see #7891). Use instead a combination of "is_bounded and is_nonzero" assumptions.
is_bounded and is_unbounded assumptions were renamed to is_finite and is_infinite (see #7947).
Removed is_infinitesimal assumption (see #7995).
Removed is_real property for Sets, use set.is_subset(Reals) instead (see #7996).
For generic symbol x (SymPy's symbols are not bounded by default), inequalities with oo are no longer evaluated as they were before, e.g. x < oo no longer evaluates to True). See #7861.
CodeGen has been refactored to make it easier to add other languages. The main high-level tool is still utilities.codegen.codegen. But if you previously used the Routine class directly, note its __init__ behaviour has changed; the new utilities.codegen.make_routine is recommended instead and by default retains the previous C/Fortran behaviour. If needed, you can still instantiate Routine directly; it only does minimal sanity checking on its inputs. See #8082.
FiniteSet([1, 2, 3, 4]) syntax not supported anymore, use FiniteSet(1, 2, 3, 4) instead See #7622.
Minor changes
Updated the parsing module to allow sympification of lambda statements to their SymPy equivalent.
Lambdify can now use numexpr by specifying modules='numexpr'
Use with evaluate(False) context manager to control automatic evaluation. E.g. with evaluate(False): x + x is actually x + x, not 2*x
IndexedBase and Indexed are changed to be commutative by default
sympy.core.sets moved to sympy.sets
Changes in sympy.sets:
Infinite Range is now allowed. See PR #7741
is_subset(): The is_subset() method deprecates the subset() method. self.is_subset(other) checks if self is a subset of other. This is different from self.subset(other), which checked if other is a subset of self.
is_superset(): A new method is_superset() method is now available. self.is_superset(other) checks if self is a superset of other.
is_proper_subset and is_proper_superset: Two new methods allow checking if one set is the proper subset and proper superset of another respectively. For eg. self.is_proper_subset(other) and self.is_proper_superset(other) checks if self is the proper subset of other and if self is the proper superset of other respectively.
is_disjoint(): A new method for checking if two sets are disjoint.
powerset(): A new method powerset() has been added to find the power set of a set.
The cardinality of a ProductSet can be found using the len() function.
Changes in sympy.plot.plot_implicit:
The plot_implicit function now also allows explicitly specifying the symbols to plot on the X and Y axes. If not specified, the symbols will be assigned in the order they are sorted.
The plot_implicit function also allows axes labels for the plot to be specified.
rules for simplification of ImageSet were added PR#7625. As a result {x | x ∊ ℤ} now simplifies to ℤ and {sin(n) | n ∊ {tan(m) | m ∊ ℤ}} automatically simplifies to {sin(tan(m)) | m ∊ ℤ}
coth(0) now returns Complex Infinity. See #7634
dioptre is added to physics.units #7782
replace now respects commutativity #7752
The CCodePrinter gracefully handles Symbols which have string representations that match C reserved words. #8199
limit function now returns an unevaluated Limit instance if it can't compute given limit, see #8213
0.7.5
Major changes
The version of mpmath included in SymPy has been updated to 0.18.
New routines for efficiently compute the dispersion of a polynomial or a pair thereof.
Fancy indexing of matrices is now provided, e.g. A[:, [1, 2, 5]] selects all rows and only 3 columns.
Enumeration of multiset partitions is now based on an implementation of Algorithm 7.1.2.5M from Knuth's The Art of Computer Programming. The new version is much faster, and includes fast methods for enumerating only those partitions with a restricted range of sizes, and counting multiset partitions. (See the new file sympy.utilities.enumerative.py.)
distance methods were added to Line and Ray to compute the shortest distance to them from a point.
The normal_lines method was added to Ellipse to compute the lines from a point that strike the Ellipse at a normal angle.
inv_quick and det_quick were added as functions in solvers.py to facilitate fast solution of small symbolic matrices; their use in solve has reduced greatly the time needed to solve such systems.
solve_univariate_inequality has been added to sympy.solvers.inequalities.py.
as_set attribute for Relationals and Booleans has been added.
Several classes and functions strictly associated with vector calculus were moved from sympy.physics.mechanics to a new package sympy.physics.vector. (PRs #2732#2862#2894)
New implementation of the Airy functions Ai and Bi and their derivatives Ai' and Bi' (called airyai, airybi, airyaiprime and airybiprime, respectively). Most of the usual features of SymPy special function are present. Notable exceptions are Gruntz limit computation helpers and meijerg special functions integration code.
Euler-Lagrange equations (function euler_equations) in a new package sympy.calculus (PR #2431).
Minor changes
Some improvements to the gamma function.
generate_bell now generates correct permutations for any number of elements.
It is no longer necessary to provide nargs to objects subclassed from Function unless an eval class method is not defined. (If eval is defined, the number of arguments will be inferred from its signature.)
geometric Point creation will be faster since simplification is done only on Floats
Some improvements to the intersection method of the Ellipse.
solutions from solve of equations involving multiple log terms are more robust
idiff can now return higher order derivatives
Added to_matrix() method to sympy.physics.vector.Vector and sympy.physics.dyadic.Dyadic. (PR #2686).
Printing improvements for sympy.physics.vector objects and mechanics printing. (PRs #2687, #2728, #2772, #2862, #2894)
Functions with LaTeX symbols now print correct LaTeX. (PR #2772)
init_printing has several new options, including a flag print_builtin to prevent SymPy printing of basic Python types (PR #2683), and flags to let you supply custom printers (PR #2894).
improvements in evaluation of imageset for Intervals (PR #2723).
Set properties to determine boundary and interior (PR #2744).
input to a function created by lambdify no longer needs to be flattened.
Backwards compatibility breaks and deprecations
the submatrix method of matrices was removed; access the functionality by providing slices or list of rows/columns to matrix directly, e.g. A[:, [1, 2]].
Matrix([]) and Matrix([[]]) now both return a 0x0 matrix
terms_gcd no longer removes a -1.0 from expressions
extract_multiplicatively will not remove a negative Number from a positive one, so (4*x*y).extract_multiplicatively(-2*x) will return None.
the shape of the result from M.cross(B) now has the same shape as matrix M.
The factorial of negative numbers is now zoo instead of 0. This is consistent with the definition factorial(n) = gamma(n + 1).
1/0 returns zoo, not oo (PR #2813).
zoo.is_number is True (PR #2823).
oo < I raises TypeError, just as for finite numbers (PR #2734).
1**oo == nan instead of 1, better documentation for Pow class (PR #2606).
20dec34d0a