numexpr 2.14.0__tar.gz → 2.14.2__tar.gz

{numexpr-2.14.0 → numexpr-2.14.2}/ANNOUNCE.rst

@@ -1,27 +1,37 @@
 =========================
-Announcing NumExpr 2.14.
+Announcing NumExpr 2.14.2
 =========================
 
 Hi everyone,
 
-NumExpr 2.14.0 introduces a couple of patches for tan/tanh and
-adds static typing support.
-Thanks to Luke Shaw and Joren Hammudoglu (@jorenham) for these contributions.
+NumExpr 2.14.2 is a maintenance release with several bug fixes, a new
+``disable_cache`` option for ``evaluate()``, and updated build/CI support
+(Windows ARM64 wheels, dropped Python 3.10, no more free-threaded 3.13
+wheels).
 
 Project documentation is available at:
 
 https://numexpr.readthedocs.io/
 
-Changes from 2.13.1 to 2.14.0
+Changes from 2.14.1 to 2.14.2
 -----------------------------
 
-* Numerical stability for overflow has been improved for ``tan`` and ``tanh`` 
-  to handle possible overflows for complex numbers.
-
-* Static typing support has been added, making NumExpr compatible with
-  static type checkers like `mypy` and `pyright`.
-  Thanks to Joren Hammudoglu (@jorenham) for the work.
-
+* Added a ``disable_cache`` parameter to ``evaluate()`` to bypass the
+  internal expression cache. Thanks to 27rabbitlt.
+* Added Windows ARM64 wheel builds.
+* Dropped support for Python 3.10.
+* No longer build free-threaded Python 3.13 wheels, matching NumPy's own
+  support.
+* Avoid keeping arrays passed as ``out=`` alive in the ``re_evaluate`` cache
+  (#558).
+* Guarded out-of-range shift counts (shift amount >= bit width) in the
+  integer ``<<``/``>>`` opcodes, which was undefined behavior in C and could
+  return garbage results. Thanks to uwezkhan (#559).
+* Fixed ``run_interpreter()`` unconditionally returning success even when
+  the VM engine failed, so execution errors are now correctly raised
+  instead of silently discarded (#557).
+* Fixed a reference leak of ``constsig`` on the allocation-failure path in
+  ``NumExpr_init()`` (#561).
 
 What's Numexpr?
 ---------------

{numexpr-2.14.0 → numexpr-2.14.2}/MANIFEST.in

@@ -2,7 +2,7 @@ include MANIFEST.in VERSION
 
 include *.rst *.txt *.cfg site.cfg.example
 
-recursive-include numexpr *.cpp *.hpp *.py *.pyi py.typed
+recursive-include numexpr *.cpp *.hpp *.py
 recursive-include numexpr/win32 *.c *.h
 exclude numexpr/__config__.py RELEASING.txt site.cfg
 

{numexpr-2.14.0 → numexpr-2.14.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: numexpr
-Version: 2.14.0
+Version: 2.14.2
 Summary: Fast numerical expression evaluator for NumPy
 Author-email: "David M. Cooke, Francesc Alted, and others" <blosc@blosc.org>
 Maintainer-email: Blosc Development Team <blosc@blosc.org>
@@ -14,20 +14,18 @@ Classifier: Intended Audience :: Information Technology
 Classifier: Intended Audience :: Science/Research
 Classifier: Programming Language :: Python
 Classifier: Topic :: Software Development :: Libraries :: Python Modules
-Classifier: Typing :: Typed
 Classifier: Operating System :: Microsoft :: Windows
 Classifier: Operating System :: Unix
 Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
 Classifier: Programming Language :: Python :: 3.13
 Classifier: Programming Language :: Python :: 3.14
-Requires-Python: >=3.10
+Requires-Python: >=3.11
 Description-Content-Type: text/x-rst
 License-File: LICENSE.txt
 License-File: LICENSES/cpuinfo.txt
-Requires-Dist: numpy>=1.23.0
+Requires-Dist: numpy>=1.26.0
 Dynamic: license-file
 
 ======================================================

{numexpr-2.14.0 → numexpr-2.14.2}/RELEASE_NOTES.rst

@@ -2,10 +2,37 @@
 Release notes for NumExpr 2.14 series
 =====================================
 
+Changes from 2.14.1 to 2.14.2
+-----------------------------
+
+* Added a ``disable_cache`` parameter to ``evaluate()`` to bypass the
+  internal expression cache. Thanks to 27rabbitlt.
+* Added Windows ARM64 wheel builds.
+* Dropped support for Python 3.10.
+* No longer build free-threaded Python 3.13 wheels, matching NumPy's own
+  support.
+* Avoid keeping arrays passed as ``out=`` alive in the ``re_evaluate`` cache
+  (#558).
+* Guarded out-of-range shift counts (shift amount >= bit width) in the
+  integer ``<<``/``>>`` opcodes, which was undefined behavior in C and could
+  return garbage results. Thanks to uwezkhan (#559).
+* Fixed ``run_interpreter()`` unconditionally returning success even when
+  the VM engine failed, so execution errors are now correctly raised
+  instead of silently discarded (#557).
+* Fixed a reference leak of ``constsig`` on the allocation-failure path in
+  ``NumExpr_init()`` (#561).
+
+Changes from 2.14.0 to 2.14.1
+-----------------------------
+
+* Rolled back static typing support to ensure compatibiity with NumPy 1.26.
+* Added CI tests for NumPy 1.26
+
+
 Changes from 2.13.1 to 2.14.0
 -----------------------------
 
-* Numerical stability for overflow has been improved for ``tan`` and ``tanh`` 
+* Numerical stability for overflow has been improved for ``tan`` / ``tanh``
   to handle possible overflows for complex numbers.
 
 * Static typing support has been added, making NumExpr compatible with
@@ -19,7 +46,6 @@ Changes from 2.13.0 to 2.13.1
 * Patch to maximum/minimum functions in order to match NumPy NaN handling
 * Patch to convert '+'->'|' and '*'->'&' for booleans
 
-
 Changes from 2.12.1 to 2.13.0
 -----------------------------
 

numexpr-2.14.2/VERSION

@@ -0,0 +1 @@
+2.14.2

{numexpr-2.14.0 → numexpr-2.14.2}/bench/large_array_vs_numpy.py

@@ -78,7 +78,7 @@ def benchmark_numexpr_re_evaluate(expr, a, b, c, results, indices):
     for index in indices:
         start = index * chunk_size
         end = (index + 1) * chunk_size
-        if index == 0:
+        if index == 0 or index == 1:
             # Evaluate the first chunk with evaluate
             time_taken = timeit.timeit(
                 lambda: ne.evaluate(

{numexpr-2.14.0 → numexpr-2.14.2}/bench/vml_timing2.py

@@ -13,7 +13,7 @@ import numpy as np
 
 import numexpr as ne
 
-N = int(2**26)
+N = int(2**28)
 
 x = np.linspace(0, 1, N)
 y = np.linspace(0, 1, N)
@@ -31,17 +31,17 @@ gbs = working_set_GB / (t1-t0)
 print("Time for an algebraic expression:     %.3f s / %.3f GB/s" % (t1-t0, gbs))
 
 t0 = time()
-z = np.sin(x)**2 + np.cos(y)**2
+z = np.sin(x)**3.2 + np.cos(y)**3.2
 t1 = time()
 gbs = working_set_GB / (t1-t0)
 print("Time for a transcendental expression: %.3f s / %.3f GB/s" % (t1-t0, gbs))
 
 if ne.use_vml:
     ne.set_vml_num_threads(1)
-    ne.set_num_threads(8)
+    ne.set_num_threads(16)
     print("NumExpr version: %s, Using MKL ver. %s, Num threads: %s" % (ne.__version__, ne.get_vml_version(), ne.nthreads))
 else:
-    ne.set_num_threads(8)
+    ne.set_num_threads(16)
     print("NumExpr version: %s, Not Using MKL, Num threads: %s" % (ne.__version__, ne.nthreads))
 
 t0 = time()
@@ -51,7 +51,7 @@ gbs = working_set_GB / (t1-t0)
 print("Time for an algebraic expression:     %.3f s / %.3f GB/s" % (t1-t0, gbs))
 
 t0 = time()
-ne.evaluate('sin(x)**2 + cos(y)**2', out = z)
+ne.evaluate('sin(x)**3.2 + cos(y)**3.2', out = z)
 t1 = time()
 gbs = working_set_GB / (t1-t0)
 print("Time for a transcendental expression: %.3f s / %.3f GB/s" % (t1-t0, gbs))

numexpr-2.14.2/numexpr/__init__.py

@@ -0,0 +1,66 @@
+###################################################################
+#  Numexpr - Fast numerical array expression evaluator for NumPy.
+#
+#      License: MIT
+#      Author:  See AUTHORS.txt
+#
+#  See LICENSE.txt and LICENSES/*.txt for details about copyright and
+#  rights to use.
+####################################################################
+
+"""
+Numexpr is a fast numerical expression evaluator for NumPy.  With it,
+expressions that operate on arrays (like "3*a+4*b") are accelerated
+and use less memory than doing the same calculation in Python.
+
+See:
+
+https://github.com/pydata/numexpr
+
+for more info about it.
+
+"""
+
+from numexpr.interpreter import __BLOCK_SIZE1__, MAX_THREADS, use_vml
+
+is_cpu_amd_intel = False # DEPRECATION WARNING: WILL BE REMOVED IN FUTURE RELEASE
+
+# cpuinfo imports were moved into the test submodule function that calls them
+# to improve import times.
+
+from numexpr.expressions import E
+from numexpr.necompiler import (NumExpr, disassemble, evaluate, re_evaluate,
+                                validate)
+from numexpr.utils import (_init_num_threads, detect_number_of_cores,
+                           detect_number_of_threads, get_num_threads,
+                           get_vml_version, set_num_threads,
+                           set_vml_accuracy_mode, set_vml_num_threads)
+
+# Detect the number of cores
+ncores = detect_number_of_cores()
+# Initialize the number of threads to be used
+nthreads = _init_num_threads()
+# The default for VML is 1 thread (see #39)
+# set_vml_num_threads(1)
+
+from . import version
+
+__version__ = version.version
+
+def print_versions():
+    """Print the versions of software that numexpr relies on."""
+    try:
+        import numexpr.tests
+        return numexpr.tests.print_versions()
+    except ImportError:
+        # To maintain Python 2.6 compatibility we have simple error handling
+        raise ImportError('`numexpr.tests` could not be imported, likely it was excluded from the distribution.')
+
+def test(verbosity=1):
+    """Run all the tests in the test suite."""
+    try:
+        import numexpr.tests
+        return numexpr.tests.test(verbosity=verbosity)
+    except ImportError:
+        # To maintain Python 2.6 compatibility we have simple error handling
+        raise ImportError('`numexpr.tests` could not be imported, likely it was excluded from the distribution.')