pkgsrc: since the volk release tarball only provides the required cpu_features
library as a git submodule that has to be manually cloned, we instead depend on
devel/cpu_features from pkgsrc and patch the cmake files to accept the static
library as an external dependency.
Documentation
Update README to be more verbose and to improve usefulness.
Compilers
AppleClang: Treat AppleClang as Clang.
Paired with the cpu_features introduction, this enables us to use architecture
specific kernels on a broader set of platforms.
CMake
Enable to not install volk_modtool.
Remove "find_package_handle_standard_args" warning.
cpu_features
Use cpu_features v0.6.0 as a private submodule to detect available CPU
features.
Fix incorrect feature detection for newer AVX versions.
Circumvent platform specific feature detection.
Enable more architecture specific kernels on more platforms.
Kernels
Disable slow and broken SSE4.1 kernel in volk_32fc_x2_dot_prod_32fc
Adjust min/max for 32f_s32f_convert_8i kernel
Use INT8_* instead of CHAR_*
NetworkX 2.5
Highlights
----------
This release is the result of 10 months of work with over 200 commits by
92 contributors. Highlights include:
- Dropped support for Python 3.5.
- add Pathlib support to work with files.
- improve performance.
- Updated docs and tests.
- Removed code designed to work with Python 2.
New Functions:
- lukes_partitioning
- triadic analysis functions
- functions for trophic levels analysis
- d_separated
- is_regular and other regular graph measures
- graph_hash using Weisfeiler Lehman methods
- common_neighbor_centrality (CCPA link prediction)
- max_weight_clique
- path_weight and is_path
- rescale_layout_dict
- junction_tree
New generators:
- paley_graph
- interval_graph
New layouts:
- multipartite_layout
Improvements
------------
- Add governance documents, developer guide and community structures
- Implement explicit deprecation policy.
- Initiate an NX Enhancement Proposal (NXEP) system
- optimize single_source_shortest_path
- improved consistent "weight" specification in shortest_path routines
- Reduce numpy.matrix usage which is discouraged by numpy.
- improved line color
- better search engine treatment of docs
- lattice and grid_graph and grid_2d_graph can use dim=tuple
- fix initializer of kamada_kawai_layout algorithm
- moral and threshold functions now included in namespace and docs
- scale arrows better when drawing
- more uniform creation of random lobster graphs
- allow editing graph during iteration over connected_components
- better column handling in coversion of pandas DataFrame
- allow simrank_similarity with directed graph input
- ensure VoteRank ability is nonnegative
- speedup kernighan_lin_bisection
- speedup negative weight cycle detection
- tree_isomorphism
- rooted_tree_isomorphism
- Gexf edge attribute "label" is available
API Changes
-----------
- enabled "copy" flag parameter in `contracted_nodes`
- allow partially periodic lattices
- return value for minimum_st_node_cut now always a set
- removed unused "has_numpy" argument from create_py_random_state
- fixed return values when drawing empty nodes and edges
- allow sets and frozensets of edges as input to nx.Graph()
- "weight" can be function for astar, directional_dijksta, all_shortest_path
- allow named key ids for GraphML edge writing
- all keywords are now checked for validity in nx.draw and friends
- EdgeDataView "in" operator checks if nodes are "in nbunch"
- remove completeness condition from minimum weight full matching
- option to sort neighbors in bfs traversal
- draw_networkx accepts numpy array for edgelist
- relabel_nodes with 2 nodes mapped to same node can now create multiedge
- steiner_tree works with MultiGraph
- Add `show` kwarg to view_pygraphviz
- Prepare for turning chordal_graph_cliques into a generator
- GraphML reader keyword force_multigraph creates MultiGraph even w/o multiedges
Changelog:
SymPy 1.7 was released on 29th November 2020.
This version of SymPy has been tested on Python 3.6, 3.7, 3.8, 3.9 and PyPy.
See our Python version support policy for more information on when we plan to
drop support for older Python versions.
Note in particular that SymPy 1.5.1 was the last release to support Python 2.7.
SymPy 1.7 will not work on Python 2.7. Python 2.7 users should install SymPy
1.5.1 and will not be able to install any further SymPy updates without
updating to Python 3.
Install SymPy with
pip install -U sympy
or if you use Anaconda
conda install sympy
Highlights
There are many changes in 1.7 (see below).
Backwards compatibility breaks and deprecations
* core
+ Modules and names that were accidentally importable from sympy are no
longer importable. In sympy 1.6 importing and using these names would
give a deprecation warning. In sympy 1.7 these names are removed. As an
example from sympy import add would need to be from sympy.core import
add or import sympy.core.add as add. (#19554 by @oscarbenjamin)
+ DEPRECATION: Using non-Expr args in Add, Mul or Pow is now deprecated (
#19445 by @oscarbenjamin)
* diffgeom
+ Manifold.patches attribute is deprecated. (#19368 by @JSS95 and
@mcpl-sympy)
+ Patch.coord_systems attribute is deprecated. (#19368 by @JSS95 and
@mcpl-sympy)
+ Class signature names of CoordSystem is deprecated. CoordinateSymbol
class and CoordSystem.symbols attribute are introduced instead. (#19368
by @JSS95 and @mcpl-sympy)
+ CoordSystem.transforms, CoordSystem.connect_to, and
CoordSystem.coord_tuple_transform_to are deprecated. Class signature
relations, CoordSystem.transformation and CoordSystem.transform are
introduced instead. (#19368 by @JSS95 and @mcpl-sympy)
* printing
+ sympy.printing.pretty.pretty_symbology.xstr has been deprecated (#20081
by @eric-wieser)
+ The unicode argument to sympy.printing.stringpict.prettyForm has been
deprecated (#20081 by @eric-wieser)
+ The deprecation of calling preview with viewer="StringIO" is now
expired instead of forgotten (a regression in 1.6.0) (#19905 by
@eric-wieser)
+ The modules sympy.printing.ccode, sympy.printing.fcode, and
sympy.printing.cxxcode have been renamed to sympy.printing.c,
sympy.printing.fortran, and sympy.printing.cxx, respectively. This was
done to avoid conflicts that occur when a module name is the same as a
function name. The modules are still importable under their old names
but doing so is deprecated and a warning will be given (#19908 by
@asmeurer)
+ from sympy.printing.codeprinter import Assignment no longer works. You
should use from sympy.codegen import Assignment, which is where
Assignment has lived since SymPy 1.1. (#19908 by @asmeurer)
+ Calling preview with viewer="file" no longer looks for a file
executable (a regression in 1.6.0) (#19905 by @eric-wieser)
* solvers
+ solve_linear_system is now a thin wrapper for linsolve. It is
recommended to use linsolve in new code. solve_linear_system maybe
deprecated or removed in future. (#18814 by @oscarbenjamin)
* utilities
+ find_executable is deprecated in favor of the builtin shutil.which. (#
19634 by @eric-wieser)
Changes
* algebras
+ Added DocString for integrate function of quaternion class. (#19410 by
@mohitshah3111999)
* codegen
+ Added support for generating calls to scipy.special.cosm1. (#20011 by
@bjodah)
+ sympy.codegen and sympy.combinatorics are no longer imported as part of
from sympy import *. The codegen and combinatorics modules do not have
any names that are included in the top-level namespace, so their being
imported with the top-level SymPy was a mistake. This improves the
import time of from sympy import *. (#19908 by @asmeurer)
+ Added support for generating calls to numpy.logaddexp &
numpy.logaddexp2. (#19880 by @bjodah)
+ Fix bugs to the parser of matrix expressions in order to build
CodegenArray objects correctly. (#19197 by @Upabjojr)
* combinatorics
+ Fixed a bug in PermutationGroup.minimal_blocks caused by modifying a
list while iterating over its indices. (#19954 by @ilya-pchelintsev)
* concrete
+ Implemented Raabe's Test (#18656 by @sachin-4099)
* core
+ Fixed a few broken cases of expr.is_integer (#20450 by @coproc and
@tbennun)
+ Modules and names that were accidentally importable from sympy are no
longer importable. In sympy 1.6 importing and using these names would
give a deprecation warning. In sympy 1.7 these names are removed. As an
example from sympy import add would need to be from sympy.core import
add or import sympy.core.add as add. (#19554 by @oscarbenjamin)
+ Zero raised to power Negative Infinity gives ComplexInfinity(zoo)
instead of zero (#20212 by @sidhu1012)
+ Extensible add, mul and power functions are introduced to allow sympy
objects to define what classes should be used for them in place of Add,
Mul and Pow (e.g. matrices use MatAdd). This is an experimental
approach aimed at enabling the behaviour of core routines (expand,
collect, etc) to be customised by user-defined types (e.g. MatAdd
rather than Add). This mechanism is still experimental, is not fully
implemented across the core and might be changed or removed in a future
release of sympy. (#19463 by @JSS95 and @mcpl-sympy)
+ divmod on sympy.Float with 0 numerator now results in (0, 0). (#20109
by @eriknw)
+ Dict operations no longer automatically converts strings into SymPy
types. (#19911 by @asmeurer)
+ Dict operations no longer raise SympifyError. (#19911 by @asmeurer)
+ Fixes _eval_nseries() function of power.py (#19508 by @sachin-4099)
+ Remove optional dependency on fastcache (#19492 by @pbrady)
+ DEPRECATION: Using non-Expr args in Add, Mul or Pow is now deprecated (
#19445 by @oscarbenjamin)
+ When creating an Add or Mul with evaluate=False identities (0 or 1) are
no longer removed. (#19450 by @oscarbenjamin)
+ Fixes _eval_nseries() function of mul.py (#19369 by @sachin-4099)
+ Sum is now included in count_ops results (#19261 by @smichr)
+ match has been optimized to improve term-matching in Add (#19261 by
@smichr)
+ match will automatically ignore bound symbols when matching (#19261 by
@smichr)
+ match results will no longer contain extraneous symbols (only those
requested) (#19261 by @smichr)
+ as_dummy should give a canonical result for expressions having bound
symbols (#19261 by @smichr)
+ Lambdas written in terms of different symbols will no longer compare
equal; this was an abuse of the Python == operator (#19261 by @smichr)
+ expr: A new method is_meromorphic is added. (#19306 by @jksuom)
* diffgeom
+ Manifold.patches attribute is deprecated. (#19368 by @JSS95 and
@mcpl-sympy)
+ Patch.coord_systems attribute is deprecated. (#19368 by @JSS95 and
@mcpl-sympy)
+ Class signature names of CoordSystem is deprecated. CoordinateSymbol
class and CoordSystem.symbols attribute are introduced instead. (#19368
by @JSS95 and @mcpl-sympy)
+ CoordSystem.transforms, CoordSystem.connect_to, and
CoordSystem.coord_tuple_transform_to are deprecated. Class signature
relations, CoordSystem.transformation and CoordSystem.transform are
introduced instead. (#19368 by @JSS95 and @mcpl-sympy)
+ CoordSystem.jacobian_determinant method is introduced. (#19368 by
@JSS95 and @mcpl-sympy)
* functions
+ lambdify added a warning when args is a set. (#19792 by @foice,
@RobertoFranceschini, and @sylee957)
+ Fixes _eval_nseries method of log (#20002 by @sachin-4099)
+ Adds _eval_is_meromorphic method and _eval_aseries to class lowergamma
(#19990 by @sachin-4099)
+ Adds _eval_is_meromorphic method and _eval_rewrite_as_tractable method
to class uppergamma (#19990 by @sachin-4099)
+ Rectifies eval method of class besselk (#19990 by @sachin-4099)
+ Adds _eval_is_meromorphic to bessel function. (#19963 by @sachin-4099)
+ Adds _eval_nseries function to sin and cos. (#19916 by @sachin-4099)
+ reduced symbolic multiples of pi in trigonometric functions (#19741 by
@sachin-4099)
+ fixed errors in assumptions when rewriting RisingFactorial /
FallingFactorial as gamma or factorial (#18696 by @sachin-4099)
+ Adds _singularities to LambertW function (#19716 by @sachin-4099)
+ Rectifies _eval_rewrite_as_tractable method of class erf (#19697 by
@sachin-4099)
+ Improved condition checking in piecewise_simplify() (#19596 by
@dhruvmendiratta6 and @smichr)
+ Implemented _eval_rewrite_as_Abs() for sign (#19487 by
@dhruvmendiratta6)
+ Adds cdir parameter to handle series expansions on branch cuts (#19555
by @sachin-4099)
+ Adds _eval_rewrite_as_factorial() and _eval_rewrite_as_gamma()
functionalities for subfactorial (#19515 by @sachin-4099)
+ Fixed cot.as_real_imag (sign error in imaginary part) (#19468 by
@DaveWitteMorris and @gschintgen)
+ Corrects the _eval_as_leading_term() method of tan and sec functions (#
19461 by @sachin-4099)
+ bsplines are more robust in terms of handling symbolic knots (#19272 by
@smichr)
* geometry
+ Ellipses with eccentricity >=1 and for eccentricity < 0 will be handled
as trivial cases (#20045 by @agrawalshubham01)
* integrals
+ upgrades to manualintegrate to support inverse trig functions! (#19993
by @iammosespaulr)
* interactive
+ The _repr_latex_, _repr_svg_, and _repr_png_ methods of subclasses of
sympy objects are no longer discarded when init_printing() is called. (
#19944 by @eric-wieser)
* logic
+ Binary boolean operators on Boolean objects no longer coerce strings (#
20093 by @eric-wieser)
* matrices
+ Adding an array and a matrix now consistently gives TypeError. (#20159
by @sidhu1012)
+ changed subs function to check if input is zip (#19159 by @czgdp1807
and @gfolbe318)
+ added tests in test_commonmatrix.py (#19159 by @czgdp1807 and
@gfolbe318)
+ The ordering of parameters in a parametrised solution from
gauss_jordan_solve was fixed. Previously incorrect results were
returned for some underdetermined systems. (#19885 by @1124816)
+ Prevents infinite recursion with rmul (#19860 by @mloubout)
+ Added class MatrixSet to represent the set of matrices (#19826 by
@Smit-create)
+ Disabled default matrix intermediate product simplification and changed
it to enable via context manager. (#19588 by @oscarbenjamin and
@Pristine-Cat)
+ Use more efficient Cholesky factorization method to check if matrices
are positive semidefinite. (#19573 by @galbwe)
+ row_del and col_del will raise IndexError rather than ValueError when
the index is out of bounds. (#19571 by @sylee957)
+ Fixed a bug in the implementation of Sylvester's criterion for
determining if a matrix is positive semidefinite. (#19556 by @galbwe)
+ Explicit non-matrix are treated as scalar (#19533 by @mloubout)
+ QRdecomposition will return matrices with zero rows and columns for
zero rank matrices. e.g. Matrix([[0, 0, 0], [0, 0, 0], [0, 0, 0]]) will
be decomposed with Matrix(3, 0, []) and Matrix(0, 3, []) (#19319 by
@sylee957)
+ Added Matrix.companion for creating dense companion matrix. (#19339 by
@sylee957)
+ Added CompanionMatrix for creating a symbolic companion matrix. (#19339
by @sylee957)
+ Matrix([]).eigenvals(multiple=True) will give an empty list instead of
an empty dict. (#19355 by @sylee957)
+ Determinant is now considered commutative. (#19354 by @eric-wieser)
* ntheory
+ Integrated Lenstra's Elliptic Curve factorization into factorint (#
19937 by @abh2k)
+ Added quadratic sieve (#19657 by @abh2k)
+ Implemented Lenstra's Elliptic curve factorization and Elliptic curve
primality test (#19436 by @abhinav28071999)
* parsing
+ Added support for while loop in C Parser (#20188 by @smitgajjar)
+ Upgraded Relational Operator support (#19982 by @iammosespaulr)
+ Bra-Ket Notation support (#19982 by @iammosespaulr)
+ Improved Grammar (#19982 by @iammosespaulr)
+ added Latex support for "not equal" and expanded support for less than,
greater than symbols (#19970 by @bhpayne)
+ parse_latex supports \left and \right parenthesis notation. (#19466 by
@sylee957)
+ parse_latex can parse expressions with amsmath spacing (#19475 by
@sylee957)
+ Added parsing of \exp in parse_latex. (#19395 by @sylee957)
+ Add more parsing rules and test examples. (#19177 by @wuyudi)
+ Added support for more data types in C parser: signed char, unsigned
char, short, short int, signed short, signed short int, unsigned short,
unsigned short int, unsigned int, long, long int, signed long, signed
long int, unsigned long, unsigned long int, double, long double. Also,
modified the data type for int and float to intc and float32
respectively, replacing integer and real (#19230 by @smitgajjar)
+ Added support for shorthand operators in C parser: +=, -=, *=, /= and %
= (#19230 by @smitgajjar)
+ Removed the assumption of value of a variable, if it is not initialized
while variable declaration(e.g.; in case of int a;, the value of a was
assumed to be Integer(0), in case of float b;, the value of b was
assumed to be Float(0.0) and in case of bool c;, the value of c was
assumed to be S.false). Also, removed the assumption of default value
of formal parameters while parameter declaration in function definition
or function prototype (e.g.; in case of function definition void func
(int a, float b) { //some code }, default value of formal parametersa
and b were assumed to be Integer(0) and Float(0.0) respectively) (#
19230 by @smitgajjar)
* physics.control
+ Added dc_gain, poles, zeros, and is_stable in TransferFunction class. (
#19896 by @namannimmo10)
+ Adding a new control systems toolbox as physics.control. (#18436 by
@benepla and @namannimmo10)
+ Add TransferFunction, Series, Parallel, and Feedback class for
physics.control submodule (#19390 by @benepla and @namannimmo10)
* physics.quantum
+ Simplification of Dagger() * IdentityOperator() (#19783 by
@dhruvmendiratta6)
+ When printing, Commutator, Anticommutator, InnerProduct, TensorProduct,
and OuterProduct now respect keyword-arguments to sstr. (#19424 by
@eric-wieser)
* physics.units
+ Fixed some dimensional analysis bugs with the addition and
multiplication operators. (#19705 by @mgreminger)
* physics.vector
+ Added user warnings for all possible paths are found in particle.py. (#
20131 by @sidhu1012)
+ Added user warning for cyclic paths in particle.py. (#20131 by
@sidhu1012)
+ Point.vel() now attempts to calculate the velocity using the relative
position from other points if it has not been already set by the user.
This behavior mimics the existing behavior in ReferenceFrame.ang_vel_in
(). (#20049 by @moorepants and @sidhu1012)
+ VectorLatexPrinter now respects the symbol_names setting when printing
dynamicsymbols (#19684 by @eric-wieser)
+ VectorLatexPrinter now know about the same trig functions as the
regular printer (#19684 by @eric-wieser)
+ When printing, Vector and Dyadic now respect keyword-arguments to sstr
and latex. As a result, these no longer use vector printing
customizations unless init_vprinting has been called. (#19640 by
@eric-wieser)
* plotting
+ capability to subclass Basebackend and specify it to the plotting
functions, thus creating plots with other plotting libraries (#20463 by
@oscarbenjamin)
+ Fixed plot3d_parametric_line plotting curves out of the box. (#19252 by
@sylee957)
* polys
+ Polys with complex floating point coefficients will now use the CC
domain rather than EX. (#20194 by @oscarbenjamin)
+ Fixed a bug in minimal_polynomial when using Groebner bases (compose=
False) (#19799 by @coproc)
+ Fixed GeneratorsError for creating some elements of FractionField when
its ground domain is FractionField or PolynomialRing. (#19713 by
@sylee957)
+ Fixed Poly initialized with PolyElement mutating the generator of the
coefficient in the form of polynomial. (e.g. Poly(ZZ[x](x+1), y, domain
=ZZ[x]) becomes Poly(y+1, y, domain=ZZ[x])) (#19659 by @sylee957)
+ Fixed Poly initialized with FracElement raising SympifyError. (e.g. dom
= ZZ.frac_field(x); Poly(dom(x+2), y, domain=dom)) (#19659 by @sylee957
)
+ Add division to finite extensions. (#19593 by @gschintgen and @jksuom)
+ Fixed GCD to always return a non-negative constant. (#19569 by
@KaustubhDamania)
+ added new domains for Gaussian integers and rationals (#15396 by
@jksuom and @smichr)
+ Add the Gaussian domains ZZ_I and QQ_I to Poly. Make Poly use the
Gaussian domains automatically when I is present in the input. Add
factorisation over the Gaussian integers (ZZ_I). The polys keyword
argument gaussian=True now results in the domain QQ_I rather than the
extension field QQ<I>. (#15396 by @oscarbenjamin)
+ improvement of dup_zz_mignotte_bound(f, K) by Knuth-Cohen bound (#19254
by @lagamura)
* printing
+ Support frac with pycode printers (for lambdify) (#20115 by @ehren)
+ Round-trip printing of floats (double-precision) to content MathML. (#
19958 by @lcontento)
+ latex now once again requires settings to be passed as keyword and not
positional arguments, and respects printer settings changed with
set_global_settings; restoring the behavior from sympy 1.2 and earlier.
(#20067 by @eric-wieser)
+ sympy.printing.pretty.pretty_symbology.xstr has been deprecated (#20081
by @eric-wieser)
+ The unicode argument to sympy.printing.stringpict.prettyForm has been
deprecated (#20081 by @eric-wieser)
+ The unicode attribute of sympy.printing.stringpict.prettyForm has been
deprecated (#20081 by @eric-wieser)
+ The mpmath code printer now correctly prints the loggamma function. (#
19913 by @abhaydhiman)
+ Use literal suffix for 1.0 in C89CodePrinter's printing of pow when the
exponent is -1 (#19956 by @bjodah and @mmohrhard)
+ The modules sympy.printing.ccode, sympy.printing.fcode, and
sympy.printing.cxxcode have been renamed to sympy.printing.c,
sympy.printing.fortran, and sympy.printing.cxx, respectively. This was
done to avoid conflicts that occur when a module name is the same as a
function name. The modules are still importable under their old names
but doing so is deprecated and a warning will be given (#19908 by
@asmeurer)
+ from sympy.printing.codeprinter import Assignment no longer works. You
should use from sympy.codegen import Assignment, which is where
Assignment has lived since SymPy 1.1. (#19908 by @asmeurer)
+ Calling preview with viewer="file" no longer looks for a file
executable (a regression in 1.6.0) (#19905 by @eric-wieser)
+ The deprecation of calling preview with viewer="StringIO" is now
expired instead of forgotten (a regression in 1.6.0) (#19905 by
@eric-wieser)
+ When printing, NDimArray now respects keyword-arguments to sstr (#19903
by @eric-wieser)
+ preview no longer fails with ValueError if the preamble contains a %
character. (#19858 by @eric-wieser)
+ Types which are not recognized by the LaTeX printer no longer have
their __str__ interpreted as LaTeX, and have their str(...) printed as
pre-formatted text as if they were printed normally. This includes the
builtin str type: latex("hello") now results in the latex \mathtt{\text
{hello}}.
If a custom type intends to be interpreted as latex, it should define
the _latex hook as described in the docs:
def _latex(self, printer):
return str(self) # indicate that the result of __str__ is LaTeX-compatible
To print a string containing LaTeX math using MathJax, use
import IPython
IPython.display.Math(string)
(#19611 by @eric-wieser)
+ The latex printer now shows builtin floats using ...x10^... notation
instead of ...e... notation. (#19611 by @eric-wieser)
+ preview now throws OSError instead of SystemError if viewers cannot be
found. (#19818 by @eric-wieser)
+ Fixed the SciPy printer for sparse matrices (#19633 by @benjaminwolba
and @sylee957)
+ preview(..., output='pdf') now uses dvipdfmx instead of dvipdf if
available. As a result, it now works with a MiKTeX installation on
windows. (#19635 by @eric-wieser)
+ LatexPrinter._print is no longer called on strings which have already
been converted to Latex. (#19614 by @eric-wieser)
+ A new sympy.printing.defaults.Printable base class was added, which is
a mixin providing _repr_latex_ in terms of LatexPrinter().doprint().
Any user type which subclasses this will opt into init_printing. (#
19425 by @eric-wieser)
+ Printing unevaluated Muls with multiple Number factors will show all
factors strictly in order. Identities will also be shown explicitly. (#
19450 by @oscarbenjamin)
+ user types that implement _latex are now printed as LaTeX when embedded
within builtin collections like list or dict. (#19389 by @eric-wieser)
+ fixed a bug where srepr function would not print dictionary and set
properly (#19346 by @rational-kunal)
+ tuples of one element now include the usual trailing comma (#19348 by
@eric-wieser)
+ mat_symbol_style='bold' no longer applies to subscripts of matrix
symbols (#19301 by @eric-wieser)
* series
+ Adds e.is_Pow heuristic to limits.py to improve the limit evaluations
of Pow objects (#19680 by @sachin-4099)
+ Changes in mrv() function of gruntz.py and cancel() function of
polytools.py resolves RecursionError and Timeout in limit evaluations (
#19646 by @sachin-4099)
+ modified order.py to better work with Add objects. (#19546 by
@maurogaravello)
+ Adds a functionality to the doit() method of limits.py which uses
is_meromorphic() for limit evaluations (#19432 by @sachin-4099)
+ Fixed wrong computations of fourier_series for even or odd functions
with limits that are specified non-central. (#19382 by @sylee957)
+ Replaces xreplace() with subs() in rewrite() function of gruntz.py
resolving incorrect limit evaluations (#19297 by @sachin-4099)
+ Adds a condition to limitinf() function of gruntz.py resolving
incorrect limit evaluations (#19292 by @sachin-4099)
* sets
+ Earlier expr and sets were treated equal which gave incorrect output
for some set functions(mainly : - in, is_subset), made sets and expr
not to be equal (#20208 by @sidhu1012)
+ nested multi-symbol first arg for ConditionSet now handled with subs/
as_dummy (#19512 by @smichr)
+ the bound symbols cannot be replaced with subs (#19512 by @smichr)
+ the error checking for mismatched signatures for sym and the base set
has been improved (#19512 by @smichr)
+ ConditionSet still tries to unify symbols and denest a base set that is
given as a ConditionSet but will no longer introduce new symbols (and
will leave the base set a a ConditionSet) when this cannot be done (#
19512 by @smichr)
+ Fixed ConditionSet.dummy_eq() and ConditionSet.as_dummy(). (#19502 by
@gschintgen)
+ Fixed ConditionSet.subs() in the case where the substitution targets a
free symbol. (#19495 by @gschintgen)
+ Fixed _infimum_key sorting key. (#19437 by @gschintgen)
+ ImageSets with Lambdas having different symbols will no longer compare
equal and should be compared like a.dummy_eq(b) (#19261 by @smichr)
* simplify
+ Fixed sqrtdenest giving wrong results for some forms of sqrt(a + b*sqrt
(r)) (#19932 by @sylee957)
+ Fixes a bug in collect when collection keys are not independent. (#
19720 by @mloubout)
+ Improved simplification for sign() (#19596 by @dhruvmendiratta6 and
@smichr)
+ Improved trigsimp for hyperbolic functions nested in non-trig
functions. (#19548 by @gschintgen)
+ Fixes a bug in collect when collected symbols appear in the facorized
expression. (#19431 by @mloubout)
* solvers
+ Added simplification strategies for simplifying the solutions of
systems of ODEs (#19998 by @mijo2 and @oscarbenjamin)
+ Fix bug in nonlinsolve leading to key-error (#19189 by @nsfinkelstein
and @oscarbenjamin)
+ improved functionality for solveset in case of inequalities (#19978 by
@maurogaravello)
+ Handling of inequalities involving Piecewise is improved in solveset. (
#19986 by @maurogaravello)
+ Solving higher-order system of ODEs by reducing/transforming it into
the first-order system of ODEs by numerous methods. (#19838 by @mijo2)
+ Added component division technique to divide the system of ODEs into
logical sub-systems and solving each of these separately. (#19762 by
@mijo2)
+ Extending checkodesol and constants_renumber to handle system of ODEs (
#19733 by @mijo2)
+ Added dsolve_system that can solve a system of ODEs (#19695 by @mijo2)
+ API for the _linear_neq_order1_type1-4 was changed to make it easier
for anyone to access the solver (#19653 by @mijo2)
+ Solving linear systems particularly involving polynomial coefficients
is much faster (#18844 by @oscarbenjamin)
+ Added the new n equations linear first-order non-constant coefficient
non-homogeneous solver where the coefficient matrix of the system of
ODEs is commutative with its antiderivative. (#19594 by @mijo2)
+ solveset will always use a symbol that has only either the real or
complex attribute and no other attribute; when a ConditionSet is
returned, the original symbol for which the solution is being sought
will be used if it does not cause evaluation of the result. (#19512 by
@smichr)
+ In dsolve there is no a general solver that can solve systems of
constant coefficient non-homogeneous first order ODEs of any size in
terms of integrals. (#19341 by @mijo2)
+ Fixed exception handling in solveset's secondary trigonometric solver.
(#19566 by @gschintgen)
+ equations with an additive generator are now solved more quickly (#
19524 by @smichr)
+ Improved solveset capabilities for solving trigonometric equations,
notably rational and symbolic coefficients are now supported. (#19507
by @gschintgen)
+ solve_linear_system is now a thin wrapper for linsolve. It is
recommended to use linsolve in new code. solve_linear_system maybe
deprecated or removed in future. (#18814 by @oscarbenjamin)
+ Addition of a solver that handles a subclass of linear first order
non-constant coefficient homogeneous systems of ODEs with any number of
equations. (#19185 by @mijo2)
* stats
+ The state space of ContinuousMarkovChain is now by default a Range
object instead of S.Reals. (#20150 by @czgdp1807 and @naveensaigit)
+ The generator matrix of ContinuousMarkovChain is now by default a
MatrixSymbol object instead of None. (#20150 by @czgdp1807 and
@naveensaigit)
+ RandomMatrixSymbol.doit will be invariant upon call. (#20136 by
@sylee957)
+ The state space of DiscreteMarkovChain is now by default a Range object
instead of S.Reals. (#20042 by @Maelstrom6)
+ The transition probability matrix of DiscreteMarkovChain is now by
default a MatrixSymbol object instead of None. (#20042 by @Maelstrom6)
+ Fixed xreplace infinitely expanding random matrix ensemble classes. (#
20018 by @sylee957)
+ All the ensemble classes (RandomMatrixEnsemble, GaussianEnsemble,
GaussianOrthogonalEnsemble, ...) now becomes stub function
constructors. If you want to use them as classes, you should import
them as RandomMatrixEnsembleModel, GaussianEnsembleModel, ... (#20018
by @sylee957)
+ Added sampling of Joint RVs from external libraries (#19848 by
@Smit-create)
+ Added sampling for Matrix Distributions (#19857 by @Smit-create)
+ Allowing CompoundDistribution to handle more than one random variables
(#19808 by @Smit-create)
+ Added Wishart and MatrixNormal Distributions (#19795 by @Smit-create)
+ Change in return type of P and E with evaluate=False. With evaluate=
False, P and E are made to return Probability and Expectation object
respectively. (#19819 by @Smit-create)
+ Added MatrixGamma Distribution (#19734 by @Smit-create)
+ Added symbolic classes of Moment and CentralMoment (#19724 by
@Smit-create)
+ Added support for Compound Distributions (#19648 by @Smit-create)
+ Added doit in class Probability (#19696 by @Smit-create)
+ Added MutlivariateNormal and MutlivariateLaplace function (#19631 by
@Smit-create)
+ Added Poisson, Wiener and Gamma Processes (#19387 by @Smit-create)
+ Added Expectation Matrix, Variance Matrix and CrossCovariance Matrix (#
19529 by @Smit-create)
+ Added sample_stochastic for sampling from stochastic processes. (#19500
by @Smit-create)
+ FIxes free_symbols method of RandomIndexedSymbol (#19459 by
@Smit-create)
+ Added Sampling from external libraries for all the random variables of
sympy (#19342 by @Smit-create)
+ Added doit method in Expectation and made E to call Expectation (#19290
by @Smit-create)
+ Added is_random in sympy.stats.rv to check if expression contains
random variables (#19304 by @Smit-create)
+ expand added in sympy.stats.symbolic_rv API. (#19295 by @Smit-create)
+ Added Lomax and Bounded pareto distribution (#19273 by @Smit-create)
+ Added sampling methods for continuous variables (#18754 by @Smit-create
)
+ Added library option in sample (#18754 by @Smit-create)
+ sample returns an iterator object since version 1.7 (#18754 by
@Smit-create)
* tensor
+ Adding ArrayDerivative class as subclass of Derivative. This new
handles derivatives involving non-scalar expressions. (#20072 by
@Upabjojr)
* utilities
+ sympy.utilities.iterables.partitions() no longer reuses the same
dictionary for each yielded output. (#20154 by @asmeurer)
+ Add support for Integral with lambdify (using scipy or mpmath) (#20134
by @ehren)
+ Lambdifying an expression with loggamma using mpmath no longer raises
ImportError. (#19913 by @abhaydhiman)
+ find_executable is deprecated in favor of the builtin shutil.which. (#
19634 by @eric-wieser)
* vector
+ integral of parametric region depends on the order of limits at the
time of initialization. (#20044 by @friyaz)
+ vector_integrate can integrate over ImplicitRegion objects. (#19883 by
@friyaz)
+ Added a function to find a rational point on conic (#19807 by @friyaz
and @Upabjojr)
+ Added support to create implictly defined regions. (#19681 by @friyaz)
+ Added support to integrate scalar/vector fields over objects of
geometry module. (#19650 by @friyaz)
+ added class to represent integral of scalar/vector field over a
parametric surface. (#19539 by @friyaz)
+ Modified API of ParametricIntegral class (#19580 by @friyaz)
+ Added class to represent a parametric region in space. (#19472 by
@friyaz)
* other
+ The LaTeX docs now need to built with xelatex rather than pdflatex. (#
20309 by @oscarbenjamin)
+ Support for Python 3.5 has been dropped. SymPy now requires Python 3.6
or newer. (#20145 by @oscarbenjamin)
+ more function raise TypeError when passed invalid keyword-arguments,
rather than ignoring them silently (#20086 by @eric-wieser)
+ assert sympy.testing.pytest.raises(Exception, func) no longer always
asserts when pytest is not present (#20012 by @eric-wieser)
+ Make SymPy do less at import time so that import sympy is faster. (#
19910 by @asmeurer)
+ Improve code quality by cleaning the doctests up (#19406 by
@InCogNiTo124 and @oscarbenjamin)
+ intersphinx links to :mod:`sympy` no longer point to the documentation
for Immutable Matrices (#19430 by @eric-wieser)
1.19.4:
MAINT: Add check for Windows 10 version 2004 bug.
REV: Revert OpenBLAS to 1.19.2 version for 1.19.4
1.19.3:
BLD: set upper versions for build dependencies
BUG: Set deprecated fields to null in PyArray_InitArrFuncs
ENH: Warn on unsupported Python 3.10+
MAINT: Update test_requirements.txt.
ENH: Support for the NVIDIA HPC SDK nvfortran compiler
BUG: Cygwin Workaround for #14787 on affected platforms
BUG: Fix memory leak of buffer-info cache due to relaxed strides
MAINT: Backport openblas_support from master.
TST: Add Python 3.9 to the CI testing on Windows, Mac.
TST: Simplify source path names in test_extending.
gfortran 10 rejects certain argument mismatches, causing build
errors. This issue is fixed upstream[0], but the commit modifies
40 files, so until the next release use -fallow-argument-mismatch
to downgrade these to warnings, as suggested by the gcc 10 porting
guide[1].
[0] 9418632214
[1] https://gcc.gnu.org/gcc-10/porting_to.html#argument-mismatch
OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
OpenBLAS is an open source project supported by
Lab of Parallel Software and Computational Science, ISCAS.
This package builds the parallel library using pthreads.
OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
OpenBLAS is an open source project supported by
Lab of Parallel Software and Computational Science, ISCAS.
This package builds the parallel library using OpenMP.
OpenBLAS is an optimized BLAS library based on GotoBLAS2 1.13 BSD version.
OpenBLAS is an open source project supported by
Lab of Parallel Software and Computational Science, ISCAS.
This package builds the serial library.
Needs patching and testing on platforms besides Linux
Version 2.41 released on July 29, 2020 (some bug fixes of version 2.40).
Version 2.40 released on July 22, 2020.
A new solver: dual coordinate descent method for linear one-class SVM; see the paper
The Newton solver is updated to have faster training speed; see the release note
A new option -R to allow users not to regularize bias (when -B 1 is used)
Upstream changes:
Formula 1.2-4
model.part() method tries to avoid calling has_dot() by checking the data attributes first. This can greatly improve speed when there are lots of variables in a model part that is actually not of interest.
Upstream changes:
Version 3.0-4
o Added "long"-format version of the O'Brien-Kaiser repeated-measures data set.
o Fix some typos in doc for CES11 data set (fixes courtesy of Andi Fugard).
Version 3.0-3
o Update references in docs and DESCRIPTION.
Convert to use the new mk/blas.buildlink3.mk rather than a specific blas
package
These dependents were overlooked in the initial upgrade and rev bump because
they listed blas in DEPENDS rather than including the bl3.
Install the new interchangeable BLAS system created by Thomas Orgis,
currently supporting Netlib BLAS/LAPACK, OpenBLAS, cblas, lapacke, and
Apple's Accelerate.framework. This system allows the user to select any
BLAS implementation without modifying packages or using package options, by
setting PKGSRC_BLAS_TYPES in mk.conf. See mk/blas.buildlink3.mk for details.
This commit should not alter behavior of existing packages as the system
defaults to Netlib BLAS/LAPACK, which until now has been the only supported
implementation.
Details:
Add new mk/blas.buildlink3.mk for inclusion in dependent packages
Install compatible Netlib math/blas and math/lapack packages
Update math/blas and math/lapack MAINTAINER approved by adam@
OpenBLAS, cblas, and lapacke will follow in separate commits
Update direct dependents to use mk/blas.buildlink3.mk
Perform recursive revbump
CHANGES IN R 4.0.3:
NEW FEATURES:
* On platforms using configure option --with-internal-tzcode,
additional values "internal" and (on macOS only) "macOS" are
accepted for the environment variable TZDIR. (See ?TZDIR.)
On macOS, "macOS" is used by default if the system timezone
database is a newer version than that in the R installation.
* When install.packages(type = "source") fails to find a package in
a repository it mentions package versions which are excluded by
their R version requirement and links to hints on why a package
might not be found.
* The default value for options("timeout") can be set from
enviromnent variable R_DEFAULT_INTERNET_TIMEOUT, still defaulting
to 60 (seconds) if that is not set or invalid.
This may be needed when child R processes are doing downloads,
for example during the installation of source packages which
download jars or other forms of data.
LINK-TIME OPTIMIZATION on a UNIX-ALIKE:
* There is now support for parallelized Link-Time Optimization
(LTO) with GCC and for 'thin' LTO with clang _via_ setting the
LTO macro.
* There is support for setting a different LTO flag for the Fortran
compiler, including to empty when mixing clang and gfortran (as
on macOS). See file config.site.
* There is a new LTO_LD macro to set linker options for LTO
compilation, for example to select an alternative linker or to
parallelize thin LTO.
DEPRECATED AND DEFUNCT:
* The LINPACK argument to chol.default(), chol2inv(),
solve.default() and svd() has been defunct since R 3.1.0. Using
it now gives a warning which will become an error in R 4.1.0.
BUG FIXES:
* The code mitigating stack overflow with PCRE regexps on very long
strings is enabled for PCRE2 < 10.30 also when JIT is enabled,
since stack overflows have been seen in that case.
* Fix to correctly show the group labels in dotchart() (which where
lost in the ylab improvement for R 4.0.0).
* addmargins(*, ..) now also works when fn() is a local function,
thanks to bug report and patch PR#17124 from Alex Bertram.
* rank(x) and hence sort(x) now work when x is an object (as per
is.object(x)) of type "raw" _and_ provides a valid `[` method,
e.g., for gmp::as.bigz(.) numbers.
* chisq.test(*, simulate.p.value=TRUE) and r2dtable() now work
correctly for large table entries (in the millions). Reported by
Sebastian Meyer and investigated by more helpers in PR#16184.
* Low-level socket read/write operations have been fixed to
correctly signal communication errors. Previously, such errors
could lead to a segfault due to invalid memory access. Reported
and debugged by Dmitriy Selivanov in PR#17850.
* quantile(x, pr) works more consistently for pr values slightly
outside [0,1], thanks to Suharto Anggono's PR#17891.
Further, quantile(x, prN, names=FALSE) now works even when prN
contains NAs, thanks to Anggono's PR#17892. Ditto for ordered
factors or Date objects when type = 1 or 3, thanks to PR#17899.
* Libcurl-based internet access, including curlGetHeaders(), was
not respecting the "timeout" option. If this causes
unanticipated timeouts, consider increasing the default by
setting R_DEFAULT_INTERNET_TIMEOUT.
* as.Date(<char>) now also works with an initial "", thanks to
Michael Chirico's PR#17909.
* isS3stdGeneric(f) now detects an S3 generic also when it it is
trace()d, thanks to Gabe Becker's PR#17917.
* R_allocLD() has been fixed to return memory aligned for long
double type PR#16534.
* fisher.test() no longer segfaults when called again after its
internal stack has been exceeded PR#17904.
* Accessing a long vector represented by a compact integer sequence
no longer segfaults (reported and debugged by Hugh Parsonage).
* duplicated() now works also for strings with multiple encodings
inside a single vector PR#17809.
* phyper(11, 15, 0, 12, log.p=TRUE) no longer gives NaN; reported
as PR#17271 by Alexey Stukalov.
* Fix incorrect calculation in logLik.nls() PR#16100, patch from
Sebastian Meyer.
* A very old bug could cause a segfault in model.matrix() when
terms involved logical variables. Part of PR#17879.
* model.frame.default() allowed data = 1, leading to involuntary
variable capture (rest of PR#17879).
* tar() no longer skips non-directory files, thanks to a patch by
Sebastian Meyer, fixing the remaining part of PR#16716.
# ellipsis 0.3.1
* Fixed an incompatibility with R devel.
* New `?dots_used` topic from which you can inherit documentation for
`...` documentation when the dots are passed to methods.
Upstream changes:
Changes to Version 3.0-10
o Fixed as error when using parameter name "(Intercept)" with deltaMethod().
o Several changes to Boot() to correct bugs, support the plinear algorithm, to correct use of weights for lm, and allow use of weights for nls.
o Fix quantregLine() to work with development version 5.69 of quantreg package (following heads-up from Roger Koenker).
Version 0.51.2
This is a bugfix release for 0.51.1. It fixes a critical performance bug in the CFG back edge computation algorithm that leads to exponential time complexity arising in compilation for use cases with certain pathological properties.
Continue. Don’t visit already checked successors
Version 0.51.1
This is a bugfix release for 0.51.0, it fixes a critical bug in caching, another critical bug in the CUDA target initialisation sequence and also fixes some compile time performance regressions:
Fix objmode cache segfault
Fix compilation slowdown due to controlflow analysis
CUDA: Don’t make a runtime call on import
Fix: Make UnicodeCharSeq into str for comparison.
Fix: Failure in test_cuda_submodules
1.19.2:
TST: Change aarch64 to arm64 in travis.yml.
MAINT: Configure hypothesis in ``np.test()`` for determinism,...
BLD: pin setuptools < 49.2.0
ENH: Add NumPy declarations to be used by Cython 3.0+
BUG: Remove non-threadsafe sigint handling from fft calculation
BUG: core: fix ilp64 blas dot/vdot/... for strides > int32 max
DOC: Use SPDX license expressions with correct license
DOC: Fix the link to the quick-start in the old API functions
BUG: fix pickling of arrays larger than 2GiB
Changes in version 2.7-0
o New dialogs for repeated-measures ANOVA and ANCOVA and plots with one or
two repeated-measures factors.
o New dialogs to reshape data sets from wide to long and from long to wide.
o Better handling of discrete numeric variables (after suggestions by Ethan Harris).
o Fix clipboard data set input to work on macOS.
o Improved pie charts and bar graphs via the piechart() and Barplot()
functions in RcmdrMisc.
o New File > Exit > Restart Commander menu item (suggestion of Vilmantas Gegzna).
o Fix Tk error produced by closing the dataset viewer and then repopening
the same data set (reported by Jose G. Conde).
o Fixes to handling panes with console.output and log.commands options set
to various values (thanks to Vilmantas Gegzna).
o Better handling of character data following change to stringsAsFactor
option in R 4.0.0+.
o Check for presence of Tktable and adapt to its absence by suppressing the
Rcmdr dataset editor and delta method menu item; Tktable eventually to be
replaced with tablelist.
o Small improvements and bug fixes.
o Updated translations (with thanks to): Basque (Joserra Rueda), French
(Milan Bouchet-Valat), Hungarian (thanks to Tamas Ferenci), Indonesian (I
Made Tirta), Japanese (Takaharu Araki), Korean (Jong-Hwa Shin), Russian
(Alexey Shipunov), Slovenian (Matjaz Jeran), Spanish (Manuel Munoz
Marquez).
Changes in version 2.6-3
o Fix bug in processing resubmitted multi-line commands (reported by
Katarina Kosmelj).
o Fix bug in processing multiline comments in onSubmit() in Commander.
o Updated Slovenian translation (thanks to Matjaz Jeran).
Changes in version 2.6-2
o New Hungarian translation (thanks to Tamas Ferenci and Andrea Harnos).
o Fixed bug in Ctrl-S for script window (reported by Mayeul Kauffmann).
Changes in version 2.6-1
o Restore older version of splitCmd().
o For GLMs and GLMMs, provide confidence intervals for exponentiated
coefficients for models using log or logit link (suggestion of Tan Chuen
Seng).
o Make listTwoLevelFactors() work properly for logical and character
variables (treated as factors) with NAs (problem pointed out by Vilmantas
Gegzna).
o Insure that data set columns are usable in the Rcmdr, e.g., excluding
list columns (problem pointed out by Vilmantas Gegzna).
o Change instances of class(x) == "y" to inherits(x, "y").
o Use apalike rather than apa BiBTeX package in Getting Started vignette
(latter has been removed from TeXLive, reported by Brian Ripley).
Changes in version 2.6-0
o Linear and generalized linear mixed models dialogs added.
o Made editDataset() with data.frame method (suggestion of Ulrike Gromping).
o New R Commander hex sticker.
o Small fixes and improvements.
o Updated translations (with thanks to): Chinese (Tsung-wu Ho), Indonesian
(I Made Tirta), Japanese (Takaharu Araki), Polish (Lukasz Daniel), Romanian
(Adrian Dusa), Russian (Alexey Shipunov), Slovenian (Matjaz Jeran), Spanish
(Manuel Munoz Marquez).
Changes to Version 3.0-9
o Declare linearHypothesis.survreg() method, previously omitted from
NAMESPACE (following question by Robert Dinterman); tweaks to survreg
methods for linearHypothesis(), Anova(), and deltaMethod() to make them
more robust.
o Fix bug in hccm() when model matrix includes just one column (reported by
Justin Yap).
o Fix buglet in scatterplotMatrix.formula() when groups specified.
o Several functions are now generic: avPlots(), ceresPlots(), crPlots(),
mcPlots().
o Enable spread smoothers for crPlot(), ceresPlot(), and residualPlot()
(suggestion of Dave Armstrong).
o Small changes to docs.
Changes to Version 3.0-8
o Import() now defaults to stringsAsFactors=FALSE to match the default
behavior of read.table() as of R 4.0.0.
o New strings2factors() function to convert character variables in a data
frame to factors.
o Added data.frame method for S().
o Fixed a bug in estimateTransform.lm() that returned the correct answer
with an unneeded warning.
o Fixed Anova.coxph() to handle weights properly (problem reported by
Daniel Morillo Cuadrado).
Changes to Version 3.0-7
o Fixed boxCox(), powerTransform() to work when 'car' package is not attached
o Fixed Anova() to work with "lme" models without an intercept (to fix bug
reported by Benjamin Tyner).
Changes to Version 3.0-6
o Explicitly import utils::globalVariables() (suggestion of Vince Carey).
Changes to Version 3.0-5
o Change class(x) == "y" to inherits(x, "y") throughout code (thanks to Martin Maechler).
Changes to Version 3.0-4
o Scoping improvements to deltaMethod(), which now also takes an optional
hypothesized value; contribution and suggestion by Pavel Krivitsky.
o Make Anova() work for coxph() models with strata (following problem
reported by Susan Galloway Hilsenbeck).
o Modify carPalette() to provide a colorblind-friendly palette as an option
(suggestion of Sercan Kahveci).
o Small improvements.
Changes to Version 2.7-1
o Added repeatedMeasuresPlot().
o Added reshapeL2W() and reshapeW2L().
o The stringsAsFactors argument to readSAS(), readSPSSS(),
readStata(), and readXL() now defaults to FALSE.
o Barplot(), discretePlot(), Dotplot(), Hist(), plotMeans()
accommodate logical and character variables as grouping factor(s).
Changes to Version 2.7-0
o Added discreteCounts() and piechart() (after suggestions of Ethan Harris).
o Added counts/percents to bars in Barplot() (suggestion of Ethan Harris).
o Fixed DeltaMethod() so that it doesn't ignore the level argument
(reported by on de Haan).
o Small improvements.
Version 2.4-2, 2020-09-12
* Support reading/writing special values (e.g. NA, Inf) without substitution,
mainly in cases where type conversion between R and NetCDF is not required.
* Fix selection of na.mode values 0,1,2
* Fix range checks for numeric conversions with packing
* Significantly increase test coverage of type conversion routines
* Support parallel NetCDF when running with MPI
* Allow creation of datasets in "data64" (CDF5) format
* Support diskless files and in-memory manipulation of persistent files
Version 2.3-1, 2020-04-27
* Fix problem when defining scalar variables with netcdf-4.7.4
Version 2.1-1, 2019-10-18
* Remove automatic handling of fill values in user-defined types
* Link rwinlib libraries by pathname to fix r-hub Windows builds
* Correct quoting of external software names in DESCRIPTION
Version 2.0-4, 2019-10-13
* Fix OSX packages by linking expat library
Version 2.0-3, 2019-10-05
* Support NetCDF-4 features for reading and writing of datasets.
* Suggest bit64 package for NC_INT64 and NC_UINT64 types.
* Drop support for netcdf-3.x library, require netcdf-4.x.
* Drop support for udunits-1 library.
* Disable calendar functions if udunits-2.x is not found during build.
