AST function handler for the last assignment in a statement block. Promotes output magically if it has not been pre-allocated.

AST factory function for attributes, such as :code:`numpy.pi`. Other than the attribute handle these are always treated similar to :cod

The equivalent AST factory for :code:`ast.Boolop` nodes to :code:`_binop`. :code:`ast.Boolop` nodes typically represent bitshift, bitand, and

AST function factory for a callable, for one to three arguments. One-function arguments are typical, they may take many forms such as ca

_cast3 isn't called by where/tenary so no need for an implementation at present.

The equivalent AST factory for :code:`ast.Compare` nodes to :code:`_binop`. :code:`ast.Compare` node fields are :code:`(left,ops,comparators

AST factory function for building scalar constants from numbers or string literals parsed by the AST. The constants are stored and cas

The statement block can end without an assignment, in which case the output is implicitly allocated and returned.

Parse a list literal into a numpy.array e.g. :code:`ne3.NumExpr( 'a < [1,2,3]' )` Only numbers are supported at present. Mixing floats

AST factory function for NumReg registers from variable names parsed by the AST. Handles three cases: 1. node.id is already in namesR

Currently only :code:`ast.USub`, i.e. the negation operation :code:`'-a', is supported, and the :code:`node.operand` is the value acted upon

Prints a formatted representation of the program and registers for debugging purposes.

Evaluate a mutli-line expression element-wise, using NumPy broadcasting rules. This function is provided as a convience for porting code fro

Needed for allowing the internal casting in numexpr machinery for conjugate operations RAM: What this comment means is if the user calls conj(x) on a

Needed for allowing the internal casting in numexpr machinery for conjugate operations

Multiprocessing cannot deal with bound methods, as they aren't pickleable.

Returns the CPU info by using the best sources of information for your OS. Returns {} if nothing is found.

Get the number of threads currently being used by the virtual machine.

TODO: refactor get_reduction_axis

RAM: hypotf was also missing

RAM: cross-power-spectrum (for Phase Correlation). https://en.wikipedia.org/wiki/Phase_correlation

RAM: algorithm branches, and is really not designed for a modern CPU... TODO: we should be able to branch the exp>-100 && exp<100 test before the for

logaddexp

Print software versions that NumExpr imports and CPU information.

* IMPORTANT: This implementation requires that pthread_cond_signal * is called while the mutex is held that is used in the corresponding * pthread_c

To save namedtuples, convert them to ``dict`` objects via ``namedtuple._asdict()`` before serializing as JSON? They are saved without field n

Sets a number of threads to be used in operations. Generally speaking one should use >= the number of physical cores. NumExpr does not benef