riptide_cpp
riptide_cpp copied to clipboard
Multihreaded 64 bit c++ files for processing numba arrays
numba 0.55.2 was recently released with compatibility for numpy 1.22; that means _riptide_cpp_ can now add builds targeting numpy 1.22 since _riptable_ will now be able to use it.
When compiling a CMake release build, CMake doesn't invoke the compiler / linker with flags needed to preserve debugging information. There is a ``RelWithDebInfo`` target which does set these flags,...
Fixes rtosholdings/riptable#87 and rtosholdings/riptable#175. This change fixes the max/nanmax/min/nanmin reduction kernels so they propgate NaNs (max, min) or ignore NaNs (nanmax, nanmin). The changes also consolidate + streamline some logic;...
The AllocateLikeNumpyArray() function seems to intend to create a new array with the same shape as the template array, but it uses the [PyArray_NewLikeArray](https://numpy.org/doc/1.17/reference/c-api.array.html#c.PyArray_NewLikeArray) API with subOK=true, which implies a...
#51 added the ``ut`` framework and C++-based unit tests for this project (vs. the Python-based tests in rtosholdings/riptable that exercise the C++ code indirectly). rtosholdings/riptable also contains some property-based tests...
We're in C++17, we have [[maybe_unused]], as well as empty-declaration parameters available.
RipTide.cpp defines min/max macros using the naive implementation which is subtly wrong, depending on the arguments passed. We should be using the STL version instead.
First available in Python 3.8, [PEP 578](https://www.python.org/dev/peps/pep-0578/) introduced audit hooks for certain functions which could be considered security-sensitive, such as opening a file or accessing the Windows registry. To better...
This change is fine for now, but it'd be a good idea to make another pass through the code (in a separate changelist/PR) to get rid of these constants sprinkled...