* Defer numpy initialization to first use in rdchem, rdmolops, cDataStructs
`from rdkit import Chem` unconditionally bootstrapped numpy (~120ms) via
import_array()/boost::python::numpy::initialize() in module init functions,
even when no numpy-dependent APIs were called. This is costly in cold-start
environments like AWS Lambda.
Move numpy initialization behind lazy guards (static bool + first-call init)
in rdchem.so, rdmolops.so, and cDataStructs.so. Numpy now loads only when
an API that actually needs it is invoked (GetDistanceMatrix, GetPositions,
SetPositions, GetAdjacencyMatrix, ConvertToNumpyArray, etc.).
Also change Conformer::SetPos to accept python::object instead of
np::ndarray to prevent Boost.Python from requiring numpy type conversion
before the lazy guard runs.
Adds test_lazy_numpy.py with subprocess-based tests verifying:
- `from rdkit import Chem` does not load numpy
- SmilesToMol/MolToSmiles work without numpy
- numpy loads on demand when array APIs are called
* skip inchi tests if not available
* switch to threadsafe once_flag, like elsewhere
* finish ifdef style
* switch to magic static style
* Revert "switch to magic static style"
This reverts commit 7300188db7.
* Add RealValueVect.
* Add UniformRealValueGrid3D
* Add Molecular Interaction Fields (MIFs)
* line endings
* cherry-pick f1bc94a4c8
* format
* Adapt tests for python3.
* Adapt RealValueVector pickling for python3.
* Speed-up of MIF calculations.
* Bugfix in MIFDescriptors.cpp.
* all tests pass
* clean up some memory leaks
* update copyrights
* rename
* rename the library
* complete the rename
* lost file
* another forgotten file
* cleanup
* clang-tidy
* clang-tidy
* windows DLL builds work
* python wrapper and tests cleanup
* convert to catch2 testing
* switch RealValueVect to use std::vector
* remove obsolete friend
* - Replace explicit loops with stdlib implicit equivalents
- Replace explicit types with auto where possible
- Avoid unnecessary copy operations where possible
- Replace raw pointers with exception-safe unique_ptr
- Replace C-style #define with constexpr
- Replace C-style casts with C++ casts
- Replace C-style arrays with std::vector
- Avoid code duplication with templated operators
- Replace VdWaals class taking multiple atom type definitions and force-field name as string parameter with force-field-specific classes deriving from an abstract VdWaals class
- Replace x,y,z doubles with Point3D class where possible
- Removed unused (and untested) DistanceToClosestAtom class
- Renamed some variables and functions for better clarity
- Converted tabs to spaces
- Made the mol parameter in cube read/write functions optional for convenience
- Made the Python wrappers more pythonic (e.g., avoid C++-style passing objects as parameters which are modified in place)
- Implemented alternative Python class constructors using boost::python::make_constructor rather than with external non-class functions
- The Python wrappers taking a sequence of Point3D now take a sequence of sequences, such that the output of Conformer.GetPositions() can be passed
- Made the Python wrapper sequence parsing more robust
- Removed duplicated code from Python wrappers
* - avoid an unnecessary copy
* progress
* works
* more cleanup
* all tests pass
* changes in response to review
---------
Co-authored-by: dfhahn <dfhahn@users.noreply.github.com>
Co-authored-by: ptosco <paolo.tosco@novartis.com>
* clean up of python and c++ for-loops
* readd assignement
* Remove misleading walrus operators (#7323)
* Run installed tests if applicable (#7325)
* simpler check for equality
* Revert "simpler check for equality"
py assert does not work pointwise on lists on all platforms,
we need to iterate through all elements
This reverts commit 03cc0ad1a6.
* fix ambiguity on truth value of np arrays
* compare list elementwise
---------
Co-authored-by: Matt Swain <m.swain@me.com>
* run clang-tidy with readability-braces-around-statements
clang-format the results
clean up all the parts that clang-tidy-8 broke
* fix problem on windows
