latqcdtools.base.check

_usesAccelerateBLAS() -> bool:
"""
Apple's Accelerate BLAS/LAPACK, used by the default macOS arm64 numpy wheels, is known to leave the
hardware FPU status flags "tainted" after certain matmul/linalg calls involving matrices with many
exact-zero entries (e.g. diagonal or orthogonal matrices), even though the numerical result is fine.
Once tainted, numpy reports ALL FOUR floating-point exception bits (divide, overflow, underflow,
invalid; flag=15) simultaneously for essentially any later operation, whether or not it actually did
anything problematic. This has no equivalent on the OpenBLAS/MKL backends normally used on Linux, so
we only need to work around it when Accelerate is actually in use.
"""

checkDomain(obj, expectedDomain):
"""
Check that obj lies in expectedDomain.

Args:
    obj (obj)
    expectedDomain (array-like): collection of values obj is allowed to take
"""

checkEqualLengths(*args):
"""
Check that all array-like objects passed have the same length. 
"""

checkExtension(filename, extension):
"""
Check the extension of a file

Args:
    filename (str)
    extension (str)
"""

checkType(expectedType, **kwargs):
"""
Check the type of an object. If it thinks the type is wrong, it will tell you what the
name of obj is (as you named it in your code) along with its type and what was expected.
Grabbing the name doesn't work if you pass him a dictionary element like myDict['key'];
it can only tell the name myDict. One could use type hints, but at the time of writing,
type hints will not necessarily crash the program, which I want.

Args:
    obj (obj)
    expectedType (type): what type do you expect? Also accepts "array", "real", "int", and "scalar".
"""

err_handler(err, flag):
"""
This method lets us control in detail how different types of errors are treated.
"""

ignoreDivideByZero():
"""
Turn off zero division crashes. 
"""

ignoreInvalidValue():
"""
Turn off invalid value crashes. 
"""

ignoreOverflow():
"""
Turn off overflow crashes. 
"""

ignoreUnderflow():
"""
Turn off underflow crashes. 
"""

class DivideByZeroError:

class InvalidValueError:

class UnderflowError: