Script 'mail_helper' called by obssrc Hello community, here is the log from the commit of package python-numexpr for openSUSE:Factory checked in at 2023-01-03 15:04:27 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/python-numexpr (Old) and /work/SRC/openSUSE:Factory/.python-numexpr.new.1563 (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Package is "python-numexpr" Tue Jan 3 15:04:27 2023 rev:18 rq:1046267 version:2.8.4 Changes: -------- --- /work/SRC/openSUSE:Factory/python-numexpr/python-numexpr.changes 2022-08-04 13:23:19.472475012 +0200 +++ /work/SRC/openSUSE:Factory/.python-numexpr.new.1563/python-numexpr.changes 2023-01-03 15:04:34.794234799 +0100 @@ -1,0 +2,13 @@ +Mon Jan 2 18:44:01 UTC 2023 - Dirk Müller <dmuel...@suse.com> + +- update to 2.8.4: + * Support for Python 3.11 has been added. + * Thanks to Tobias Hangleiter for an improved accuracy complex `expm1` function. + While it is 25 % slower, it is significantly more accurate for the real component + over a range of values and matches NumPy outputs much more closely. + * Thanks to Kirill Kouzoubov for a range of fixes to constants parsing that was + resulting in duplicated constants of the same value. + * Thanks to Mark Harfouche for noticing that we no longer need `numpy` version + checks. `packaging` is no longer a requirement as a result. + +------------------------------------------------------------------- Old: ---- numexpr-2.8.3.tar.gz New: ---- numexpr-2.8.4.tar.gz ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ python-numexpr.spec ++++++ --- /var/tmp/diff_new_pack.fu4Y0e/_old 2023-01-03 15:04:35.278237624 +0100 +++ /var/tmp/diff_new_pack.fu4Y0e/_new 2023-01-03 15:04:35.282237647 +0100 @@ -1,7 +1,7 @@ # # spec file for package python-numexpr # -# Copyright (c) 2022 SUSE LLC +# Copyright (c) 2023 SUSE LLC # # All modifications and additions to the file contributed by third parties # remain the property of their copyright owners, unless otherwise agreed @@ -17,7 +17,7 @@ Name: python-numexpr -Version: 2.8.3 +Version: 2.8.4 Release: 0 Summary: Numerical expression evaluator for NumPy License: MIT @@ -26,13 +26,11 @@ Source: https://files.pythonhosted.org/packages/source/n/numexpr/numexpr-%{version}.tar.gz BuildRequires: %{python_module devel >= 3.7} BuildRequires: %{python_module numpy-devel >= 1.13.3} -BuildRequires: %{python_module packaging} BuildRequires: %{python_module setuptools} BuildRequires: fdupes BuildRequires: gcc-c++ BuildRequires: python-rpm-macros Requires: python-numpy >= 1.13.3 -Requires: python-packaging %python_subpackages %description ++++++ numexpr-2.8.3.tar.gz -> numexpr-2.8.4.tar.gz ++++++ diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/ANNOUNCE.rst new/numexpr-2.8.4/ANNOUNCE.rst --- old/numexpr-2.8.3/ANNOUNCE.rst 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/ANNOUNCE.rst 2022-10-26 04:43:06.000000000 +0200 @@ -1,42 +1,29 @@ ======================== -Announcing NumExpr 2.8.3 +Announcing NumExpr 2.8.4 ======================== Hi everyone, -Please find here another maintenance release of NumExpr. Support for Python 3.6 -has been dropped to enable support for NumPy 1.23 (and by extension Python 3.11 -when it is released). Wheels for ARM64 multilinux should be available again after -troubles with GitHub Actions and Apple Silicon wheels are also now available on -PyPi for download. +This is a maintenance and bug-fix release for NumExpr. In particular, now we have +added Python 3.11 support. Project documentation is available at: http://numexpr.readthedocs.io/ -Changes from 2.8.1 to 2.8.2 +Changes from 2.8.3 to 2.8.4 --------------------------- -* Support for Python 3.6 has been dropped due to the need to substitute the flag - `NPY_ARRAY_WRITEBACKIFCOPY` for `NPY_ARRAY_UPDATEIFCOPY`. This flag change was - initiated in NumPy 1.14 and finalized in 1.23. The only changes were made to - cases where an unaligned constant was passed in with a pre-allocated output - variable: - -``` - x = np.empty(5, dtype=np.uint8)[1:].view(np.int32) - ne.evaluate('3', out=x) -``` - - We think the risk of issues is very low, but if you are using NumExpr as a - expression evaluation tool you may want to write a test for this edge case. -* Thanks to Matt Einhorn (@matham) for improvements to the GitHub Actions build process to - add support for Apple Silicon and aarch64. -* Thanks to Biswapriyo Nath (@biswa96) for a fix to allow `mingw` builds on Windows. -* There have been some changes made to not import `platform.machine()` on `sparc` - but it is highly advised to upgrade to Python 3.9+ to avoid this issue with - the Python core package `platform`. +* Support for Python 3.11 has been added. +* Thanks to Tobias Hangleiter for an improved accuracy complex `expm1` function. + While it is 25 % slower, it is significantly more accurate for the real component + over a range of values and matches NumPy outputs much more closely. +* Thanks to Kirill Kouzoubov for a range of fixes to constants parsing that was + resulting in duplicated constants of the same value. +* Thanks to Mark Harfouche for noticing that we no longer need `numpy` version + checks. `packaging` is no longer a requirement as a result. + What's Numexpr? --------------- diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/PKG-INFO new/numexpr-2.8.4/PKG-INFO --- old/numexpr-2.8.3/PKG-INFO 2022-06-25 22:35:21.620025000 +0200 +++ new/numexpr-2.8.4/PKG-INFO 2022-10-26 04:59:27.071168200 +0200 @@ -1,6 +1,6 @@ Metadata-Version: 2.1 Name: numexpr -Version: 2.8.3 +Version: 2.8.4 Summary: Fast numerical expression evaluator for NumPy Home-page: https://github.com/pydata/numexpr Author: David M. Cooke, Francesc Alted, and others @@ -16,6 +16,7 @@ Classifier: Programming Language :: Python :: 3.8 Classifier: Programming Language :: Python :: 3.9 Classifier: Programming Language :: Python :: 3.10 +Classifier: Programming Language :: Python :: 3.11 Classifier: Operating System :: Microsoft :: Windows Classifier: Operating System :: POSIX Classifier: Operating System :: MacOS @@ -86,9 +87,9 @@ The result is that NumExpr can get the most of your machine computing capabilities for array-wise computations. Common speed-ups with regard to NumPy are usually between 0.95x (for very simple expressions like -:code:`'a + 1'`) and 4x (for relatively complex ones like :code:`'a*b-4.1*a > -2.5*b'`), although much higher speed-ups can be achieved for some functions -and complex math operations (up to 15x in some cases). +:code:`'a + 1'`) and 4x (for relatively complex ones like :code:`'a*b-4.1*a > 2.5*b'`), +although much higher speed-ups can be achieved for some functions and complex +math operations (up to 15x in some cases). NumExpr performs best on matrices that are too large to fit in L1 CPU cache. In order to get a better idea on the different speed-ups that can be achieved @@ -114,17 +115,17 @@ From Source ^^^^^^^^^^^ -On most `Nix systems your compilers will already be present. However if you +On most \*nix systems your compilers will already be present. However if you are using a virtual environment with a substantially newer version of Python than your system Python you may be prompted to install a new version of `gcc` or `clang`. For Windows, you will need to install the Microsoft Visual C++ Build Tools -(which are free) first.The version depends on which version of Python you have +(which are free) first. The version depends on which version of Python you have installed: https://wiki.python.org/moin/WindowsCompilers -For Python 3.6+ simply installating the latest version of MSVC build tools should +For Python 3.6+ simply installing the latest version of MSVC build tools should be sufficient. Note that wheels found via pip do not include MKL support. Wheels available via `conda` will have MKL, if the MKL backend is used for NumPy. @@ -145,7 +146,7 @@ NumExpr includes support for Intel's MKL library. This may provide better performance on Intel architectures, mainly when evaluating transcendental -functions (trigonometrical, exponential...). +functions (trigonometrical, exponential, ...). If you have Intel's MKL, copy the `site.cfg.example` that comes with the distribution to `site.cfg` and edit the latter file to provide correct paths to diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/README.rst new/numexpr-2.8.4/README.rst --- old/numexpr-2.8.3/README.rst 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/README.rst 2022-10-26 04:43:06.000000000 +0200 @@ -60,9 +60,9 @@ The result is that NumExpr can get the most of your machine computing capabilities for array-wise computations. Common speed-ups with regard to NumPy are usually between 0.95x (for very simple expressions like -:code:`'a + 1'`) and 4x (for relatively complex ones like :code:`'a*b-4.1*a > -2.5*b'`), although much higher speed-ups can be achieved for some functions -and complex math operations (up to 15x in some cases). +:code:`'a + 1'`) and 4x (for relatively complex ones like :code:`'a*b-4.1*a > 2.5*b'`), +although much higher speed-ups can be achieved for some functions and complex +math operations (up to 15x in some cases). NumExpr performs best on matrices that are too large to fit in L1 CPU cache. In order to get a better idea on the different speed-ups that can be achieved @@ -88,17 +88,17 @@ From Source ^^^^^^^^^^^ -On most `Nix systems your compilers will already be present. However if you +On most \*nix systems your compilers will already be present. However if you are using a virtual environment with a substantially newer version of Python than your system Python you may be prompted to install a new version of `gcc` or `clang`. For Windows, you will need to install the Microsoft Visual C++ Build Tools -(which are free) first.The version depends on which version of Python you have +(which are free) first. The version depends on which version of Python you have installed: https://wiki.python.org/moin/WindowsCompilers -For Python 3.6+ simply installating the latest version of MSVC build tools should +For Python 3.6+ simply installing the latest version of MSVC build tools should be sufficient. Note that wheels found via pip do not include MKL support. Wheels available via `conda` will have MKL, if the MKL backend is used for NumPy. @@ -119,7 +119,7 @@ NumExpr includes support for Intel's MKL library. This may provide better performance on Intel architectures, mainly when evaluating transcendental -functions (trigonometrical, exponential...). +functions (trigonometrical, exponential, ...). If you have Intel's MKL, copy the `site.cfg.example` that comes with the distribution to `site.cfg` and edit the latter file to provide correct paths to diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/RELEASE_NOTES.rst new/numexpr-2.8.4/RELEASE_NOTES.rst --- old/numexpr-2.8.3/RELEASE_NOTES.rst 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/RELEASE_NOTES.rst 2022-10-26 04:43:06.000000000 +0200 @@ -2,6 +2,18 @@ Release notes for NumExpr 2.8 series ==================================== +Changes from 2.8.3 to 2.8.4 +--------------------------- + +* Support for Python 3.11 has been added. +* Thanks to Tobias Hangleiter for an improved accuracy complex `expm1` function. + While it is 25 % slower, it is significantly more accurate for the real component + over a range of values and matches NumPy outputs much more closely. +* Thanks to Kirill Kouzoubov for a range of fixes to constants parsing that was + resulting in duplicated constants of the same value. +* Thanks to Mark Harfouche for noticing that we no longer need `numpy` version + checks. `packaging` is no longer a requirement as a result. + Changes from 2.8.1 to 2.8.3 --------------------------- diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/complex_functions.hpp new/numexpr-2.8.4/numexpr/complex_functions.hpp --- old/numexpr-2.8.3/numexpr/complex_functions.hpp 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/complex_functions.hpp 2022-10-26 04:43:06.000000000 +0200 @@ -158,9 +158,10 @@ static void nc_expm1(npy_cdouble *x, npy_cdouble *r) { - double a = exp(x->real); - r->real = a*cos(x->imag) - 1.0; - r->imag = a*sin(x->imag); + double a = sin(x->imag / 2); + double b = exp(x->real); + r->real = expm1(x->real) * cos(x->imag) - 2 * a * a; + r->imag = b * sin(x->imag); return; } diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/expressions.py new/numexpr-2.8.4/numexpr/expressions.py --- old/numexpr-2.8.3/numexpr/expressions.py 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/expressions.py 2022-10-26 04:43:06.000000000 +0200 @@ -15,10 +15,6 @@ import threading import numpy -# NumPy's behavior sometimes changes with versioning, especially in regard as -# to when ints are cast to floats. -from packaging.version import parse, Version -_np_version_forbids_neg_powint = parse(numpy.__version__) > Version('1.12.0b1') # Declare a double type that does not exist in Python space double = numpy.double @@ -142,6 +138,8 @@ # a float (32-bit) array with a double (64-bit) constant. if isinstance(x, double): return double + if isinstance(x, numpy.float32): + return float if isinstance(x, (int, numpy.integer)): # Constants needing more than 32 bits are always # considered ``long``, *regardless of the platform*, so we @@ -156,7 +154,7 @@ y = converter(x) except Exception as err: continue - if y == x: + if y == x or numpy.isnan(y): return converter @@ -273,8 +271,7 @@ @ophelper def pow_op(a, b): - if (_np_version_forbids_neg_powint and - b.astKind in ('int', 'long') and + if (b.astKind in ('int', 'long') and a.astKind in ('int', 'long') and numpy.any(b.value < 0)): @@ -333,6 +330,7 @@ return OpNode('pow', [a, b]) # The functions and the minimum and maximum types accepted +numpy.expm1x = numpy.expm1 functions = { 'copy': func(numpy.copy), 'ones_like': func(numpy.ones_like), @@ -504,7 +502,7 @@ def __init__(self, value=None, children=None): kind = getKind(value) # Python float constants are double precision by default - if kind == 'float': + if kind == 'float' and isinstance(value, float): kind = 'double' LeafNode.__init__(self, value=value, kind=kind) diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/functions.hpp new/numexpr-2.8.4/numexpr/functions.hpp --- old/numexpr-2.8.3/numexpr/functions.hpp 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/functions.hpp 2022-10-26 04:43:06.000000000 +0200 @@ -102,27 +102,27 @@ #define ELIDE_FUNC_CC #define FUNC_CC(...) #endif -FUNC_CC(FUNC_SQRT_CC, "sqrt_cc", nc_sqrt, vzSqrt) -FUNC_CC(FUNC_SIN_CC, "sin_cc", nc_sin, vzSin) -FUNC_CC(FUNC_COS_CC, "cos_cc", nc_cos, vzCos) -FUNC_CC(FUNC_TAN_CC, "tan_cc", nc_tan, vzTan) -FUNC_CC(FUNC_ARCSIN_CC, "arcsin_cc", nc_asin, vzAsin) -FUNC_CC(FUNC_ARCCOS_CC, "arccos_cc", nc_acos, vzAcos) -FUNC_CC(FUNC_ARCTAN_CC, "arctan_cc", nc_atan, vzAtan) -FUNC_CC(FUNC_SINH_CC, "sinh_cc", nc_sinh, vzSinh) -FUNC_CC(FUNC_COSH_CC, "cosh_cc", nc_cosh, vzCosh) -FUNC_CC(FUNC_TANH_CC, "tanh_cc", nc_tanh, vzTanh) -FUNC_CC(FUNC_ARCSINH_CC, "arcsinh_cc", nc_asinh, vzAsinh) -FUNC_CC(FUNC_ARCCOSH_CC, "arccosh_cc", nc_acosh, vzAcosh) -FUNC_CC(FUNC_ARCTANH_CC, "arctanh_cc", nc_atanh, vzAtanh) -FUNC_CC(FUNC_LOG_CC, "log_cc", nc_log, vzLn) -FUNC_CC(FUNC_LOG1P_CC, "log1p_cc", nc_log1p, vzLog1p) -FUNC_CC(FUNC_LOG10_CC, "log10_cc", nc_log10, vzLog10) -FUNC_CC(FUNC_EXP_CC, "exp_cc", nc_exp, vzExp) -FUNC_CC(FUNC_EXPM1_CC, "expm1_cc", nc_expm1, vzExpm1) -FUNC_CC(FUNC_ABS_CC, "absolute_cc", nc_abs, vzAbs_) -FUNC_CC(FUNC_CONJ_CC, "conjugate_cc",nc_conj, vzConj) -FUNC_CC(FUNC_CC_LAST, NULL, NULL, NULL) +FUNC_CC(FUNC_SQRT_CC, "sqrt_cc", nc_sqrt, vzSqrt) +FUNC_CC(FUNC_SIN_CC, "sin_cc", nc_sin, vzSin) +FUNC_CC(FUNC_COS_CC, "cos_cc", nc_cos, vzCos) +FUNC_CC(FUNC_TAN_CC, "tan_cc", nc_tan, vzTan) +FUNC_CC(FUNC_ARCSIN_CC, "arcsin_cc", nc_asin, vzAsin) +FUNC_CC(FUNC_ARCCOS_CC, "arccos_cc", nc_acos, vzAcos) +FUNC_CC(FUNC_ARCTAN_CC, "arctan_cc", nc_atan, vzAtan) +FUNC_CC(FUNC_SINH_CC, "sinh_cc", nc_sinh, vzSinh) +FUNC_CC(FUNC_COSH_CC, "cosh_cc", nc_cosh, vzCosh) +FUNC_CC(FUNC_TANH_CC, "tanh_cc", nc_tanh, vzTanh) +FUNC_CC(FUNC_ARCSINH_CC, "arcsinh_cc", nc_asinh, vzAsinh) +FUNC_CC(FUNC_ARCCOSH_CC, "arccosh_cc", nc_acosh, vzAcosh) +FUNC_CC(FUNC_ARCTANH_CC, "arctanh_cc", nc_atanh, vzAtanh) +FUNC_CC(FUNC_LOG_CC, "log_cc", nc_log, vzLn) +FUNC_CC(FUNC_LOG1P_CC, "log1p_cc", nc_log1p, vzLog1p) +FUNC_CC(FUNC_LOG10_CC, "log10_cc", nc_log10, vzLog10) +FUNC_CC(FUNC_EXP_CC, "exp_cc", nc_exp, vzExp) +FUNC_CC(FUNC_EXPM1_CC, "expm1_cc", nc_expm1, vzExpm1) +FUNC_CC(FUNC_ABS_CC, "absolute_cc", nc_abs, vzAbs_) +FUNC_CC(FUNC_CONJ_CC, "conjugate_cc",nc_conj, vzConj) +FUNC_CC(FUNC_CC_LAST, NULL, NULL, NULL) #ifdef ELIDE_FUNC_CC #undef ELIDE_FUNC_CC #undef FUNC_CC diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/interpreter.cpp new/numexpr-2.8.4/numexpr/interpreter.cpp --- old/numexpr-2.8.3/numexpr/interpreter.cpp 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/interpreter.cpp 2022-10-26 04:43:06.000000000 +0200 @@ -261,7 +261,7 @@ MKL_INT j; vzExp(n, x1, dest); for (j=0; j<n; j++) { - dest[j].real -= 1.0; + dest[j].real -= 1.0; }; }; @@ -269,8 +269,8 @@ { MKL_INT j; for (j=0; j<n; j++) { - dest[j].real = x1[j].real + 1; - dest[j].imag = x1[j].imag; + dest[j].real = x1[j].real + 1; + dest[j].imag = x1[j].imag; }; vzLn(n, dest, dest); }; diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/necompiler.py new/numexpr-2.8.4/numexpr/necompiler.py --- old/numexpr-2.8.3/numexpr/necompiler.py 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/necompiler.py 2022-10-26 04:43:06.000000000 +0200 @@ -342,7 +342,8 @@ a = 1 + 3j; b = 5.0 ne.evaluate('a*2 + 15j - b') """ - constants_order = sorted(ast.allOf('constant')) + constant_registers = set([node.reg for node in ast.allOf("constant")]) + constants_order = sorted([r.node for r in constant_registers]) constants = [convertConstantToKind(a.value, a.astKind) for a in constants_order] return constants_order, constants @@ -635,11 +636,15 @@ r_constants = 1 + len(nex.signature) r_temps = r_constants + len(nex.constants) + def parseOp(op): + name, sig = [*op.rsplit(b'_', 1), ''][:2] + return name, sig + def getArg(pc, offset): - arg = nex.program[pc + offset] - op = rev_opcodes.get(nex.program[pc]) + arg = nex.program[pc + (offset if offset < 4 else offset+1)] + _, sig = parseOp(rev_opcodes.get(nex.program[pc])) try: - code = op.split(b'_')[1][offset - 1] + code = sig[offset - 1] except IndexError: return None @@ -662,10 +667,13 @@ source = [] for pc in range(0, len(nex.program), 4): op = rev_opcodes.get(nex.program[pc]) - dest = getArg(pc, 1) - arg1 = getArg(pc, 2) - arg2 = getArg(pc, 3) - source.append((op, dest, arg1, arg2)) + _, sig = parseOp(op) + parsed = [op] + for i in range(len(sig)): + parsed.append(getArg(pc, 1 + i)) + while len(parsed) < 4: + parsed.append(None) + source.append(parsed) return source diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/tests/test_numexpr.py new/numexpr-2.8.4/numexpr/tests/test_numexpr.py --- old/numexpr-2.8.3/numexpr/tests/test_numexpr.py 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/tests/test_numexpr.py 2022-10-26 04:43:06.000000000 +0200 @@ -25,12 +25,14 @@ sin, cos, tan, arcsin, arccos, arctan, arctan2, sinh, cosh, tanh, arcsinh, arccosh, arctanh, log, log1p, log10, exp, expm1, conj) +import numpy from numpy.testing import (assert_equal, assert_array_equal, assert_array_almost_equal, assert_allclose) from numpy import shape, allclose, array_equal, ravel, isnan, isinf import numexpr from numexpr import E, NumExpr, evaluate, re_evaluate, disassemble, use_vml +from numexpr.expressions import ConstantNode import unittest @@ -39,10 +41,6 @@ double = np.double long = int -# Recommended minimum versions -from packaging.version import Version -minimum_numpy_version = Version('1.7.0') -present_numpy_version = Version(np.__version__) class test_numexpr(TestCase): """Testing with 1 thread""" @@ -481,6 +479,31 @@ else: self.fail() + def test_disassemble(self): + assert_equal(disassemble(NumExpr( + "where(m, a, -1)", [('m', bool), ('a', float)])), + [[b'where_fbff', b'r0', b'r1[m]', b'r2[a]', b'c3[-1.0]'], + [b'noop', None, None, None]]) + + def test_constant_deduplication(self): + assert_equal(NumExpr("(a + 1)*(a - 1)", [('a', np.int32)]).constants, (1,)) + + def test_nan_constant(self): + assert_equal(str(ConstantNode(float("nan")).value), 'nan') + + # check de-duplication works for nan + _nan = ConstantNode(float("nan")) + expr = (E.a + _nan)*(E.b + _nan) + assert_equal(NumExpr(expr, [('a', double), ('b', double)]).constants, (float("nan"),)) + + + def test_f32_constant(self): + assert_equal(ConstantNode(numpy.float32(1)).astKind, "float") + assert_equal(ConstantNode(numpy.float32("nan")).astKind, "float") + assert_equal(ConstantNode(numpy.float32(3)).value.dtype, numpy.dtype("float32")) + assert_array_equal(NumExpr(ConstantNode(numpy.float32(1))).run(), + numpy.array(1, dtype="float32")) + def test_unaligned_singleton(self): # Test for issue #397 whether singletons outputs assigned to consts must be # aligned or not. @@ -801,14 +824,9 @@ str_list = [ b'\0\0\0', b'\0\0foo\0', b'\0\0foo\0b', b'\0\0foo\0b\0', b'foo\0', b'foo\0b', b'foo\0b\0', b'foo\0bar\0baz\0\0'] - min_tobytes_version = Version('1.9.0') for s in str_list: r = evaluate('s') - if present_numpy_version >= min_tobytes_version: - self.assertEqual(s, r.tobytes()) # check *all* stored data - else: - # ndarray.tostring() is deprecated as of NumPy 1.19 - self.assertEqual(s, r.tostring()) # check *all* stored data + self.assertEqual(s, r.tobytes()) # check *all* stored data def test_compare_copy(self): sarr = self.str_array1 @@ -1075,10 +1093,6 @@ from numexpr.cpuinfo import cpu import platform - np_version = Version(np.__version__) - - if np_version < minimum_numpy_version: - print('*Warning*: NumPy version is lower than recommended: %s < %s' % (np_version, minimum_numpy_version)) print('-=' * 38) print('Numexpr version: %s' % numexpr.__version__) print('NumPy version: %s' % np.__version__) diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr/version.py new/numexpr-2.8.4/numexpr/version.py --- old/numexpr-2.8.3/numexpr/version.py 2022-06-25 22:35:20.000000000 +0200 +++ new/numexpr-2.8.4/numexpr/version.py 2022-10-26 04:59:26.000000000 +0200 @@ -1,4 +1,4 @@ # THIS FILE IS GENERATED BY `SETUP.PY` -version = '2.8.3' +version = '2.8.4' numpy_build_version = '1.21.6' platform_machine = 'AMD64' diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr.egg-info/PKG-INFO new/numexpr-2.8.4/numexpr.egg-info/PKG-INFO --- old/numexpr-2.8.3/numexpr.egg-info/PKG-INFO 2022-06-25 22:35:20.000000000 +0200 +++ new/numexpr-2.8.4/numexpr.egg-info/PKG-INFO 2022-10-26 04:59:26.000000000 +0200 @@ -1,6 +1,6 @@ Metadata-Version: 2.1 Name: numexpr -Version: 2.8.3 +Version: 2.8.4 Summary: Fast numerical expression evaluator for NumPy Home-page: https://github.com/pydata/numexpr Author: David M. Cooke, Francesc Alted, and others @@ -16,6 +16,7 @@ Classifier: Programming Language :: Python :: 3.8 Classifier: Programming Language :: Python :: 3.9 Classifier: Programming Language :: Python :: 3.10 +Classifier: Programming Language :: Python :: 3.11 Classifier: Operating System :: Microsoft :: Windows Classifier: Operating System :: POSIX Classifier: Operating System :: MacOS @@ -86,9 +87,9 @@ The result is that NumExpr can get the most of your machine computing capabilities for array-wise computations. Common speed-ups with regard to NumPy are usually between 0.95x (for very simple expressions like -:code:`'a + 1'`) and 4x (for relatively complex ones like :code:`'a*b-4.1*a > -2.5*b'`), although much higher speed-ups can be achieved for some functions -and complex math operations (up to 15x in some cases). +:code:`'a + 1'`) and 4x (for relatively complex ones like :code:`'a*b-4.1*a > 2.5*b'`), +although much higher speed-ups can be achieved for some functions and complex +math operations (up to 15x in some cases). NumExpr performs best on matrices that are too large to fit in L1 CPU cache. In order to get a better idea on the different speed-ups that can be achieved @@ -114,17 +115,17 @@ From Source ^^^^^^^^^^^ -On most `Nix systems your compilers will already be present. However if you +On most \*nix systems your compilers will already be present. However if you are using a virtual environment with a substantially newer version of Python than your system Python you may be prompted to install a new version of `gcc` or `clang`. For Windows, you will need to install the Microsoft Visual C++ Build Tools -(which are free) first.The version depends on which version of Python you have +(which are free) first. The version depends on which version of Python you have installed: https://wiki.python.org/moin/WindowsCompilers -For Python 3.6+ simply installating the latest version of MSVC build tools should +For Python 3.6+ simply installing the latest version of MSVC build tools should be sufficient. Note that wheels found via pip do not include MKL support. Wheels available via `conda` will have MKL, if the MKL backend is used for NumPy. @@ -145,7 +146,7 @@ NumExpr includes support for Intel's MKL library. This may provide better performance on Intel architectures, mainly when evaluating transcendental -functions (trigonometrical, exponential...). +functions (trigonometrical, exponential, ...). If you have Intel's MKL, copy the `site.cfg.example` that comes with the distribution to `site.cfg` and edit the latter file to provide correct paths to diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/numexpr.egg-info/requires.txt new/numexpr-2.8.4/numexpr.egg-info/requires.txt --- old/numexpr-2.8.3/numexpr.egg-info/requires.txt 2022-06-25 22:35:21.000000000 +0200 +++ new/numexpr-2.8.4/numexpr.egg-info/requires.txt 2022-10-26 04:59:26.000000000 +0200 @@ -1,2 +1 @@ numpy>=1.13.3 -packaging diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/requirements.txt new/numexpr-2.8.4/requirements.txt --- old/numexpr-2.8.3/requirements.txt 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/requirements.txt 2022-10-26 04:43:06.000000000 +0200 @@ -1,2 +1 @@ numpy >= 1.13.3 -packaging \ No newline at end of file diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/setup.cfg new/numexpr-2.8.4/setup.cfg --- old/numexpr-2.8.3/setup.cfg 2022-06-25 22:35:21.621027200 +0200 +++ new/numexpr-2.8.4/setup.cfg 2022-10-26 04:59:27.071168200 +0200 @@ -1,6 +1,6 @@ [metadata] name = numexpr -version = 2.8.3 +version = 2.8.4 description = Fast numerical expression evaluator for NumPy author = David M. Cooke, Francesc Alted, and others maintainer = Robert A. McLeod @@ -19,6 +19,7 @@ Programming Language :: Python :: 3.8 Programming Language :: Python :: 3.9 Programming Language :: Python :: 3.10 + Programming Language :: Python :: 3.11 Operating System :: Microsoft :: Windows Operating System :: POSIX Operating System :: MacOS diff -urN '--exclude=CVS' '--exclude=.cvsignore' '--exclude=.svn' '--exclude=.svnignore' old/numexpr-2.8.3/setup.py new/numexpr-2.8.4/setup.py --- old/numexpr-2.8.3/setup.py 2022-06-25 22:25:22.000000000 +0200 +++ new/numexpr-2.8.4/setup.py 2022-10-26 04:43:06.000000000 +0200 @@ -74,6 +74,7 @@ libs.extend( site['mkl']['libraries'].split(os.pathsep)) def_macros.append(('USE_VML', None)) + print(f'FOUND MKL IMPORT') def setup_package():
