spaCy/include/numpy/multiarray_api.txt

===========
Numpy C-API
===========
::

  unsigned int
  PyArray_GetNDArrayCVersion(void )


Included at the very first so not auto-grabbed and thus not labeled.

::

  int
  PyArray_SetNumericOps(PyObject *dict)

Set internal structure with number functions that all arrays will use

::

  PyObject *
  PyArray_GetNumericOps(void )

Get dictionary showing number functions that all arrays will use

::

  int
  PyArray_INCREF(PyArrayObject *mp)

For object arrays, increment all internal references.

::

  int
  PyArray_XDECREF(PyArrayObject *mp)

Decrement all internal references for object arrays.
(or arrays with object fields)

::

  void
  PyArray_SetStringFunction(PyObject *op, int repr)

Set the array print function to be a Python function.

::

  PyArray_Descr *
  PyArray_DescrFromType(int type)

Get the PyArray_Descr structure for a type.

::

  PyObject *
  PyArray_TypeObjectFromType(int type)

Get a typeobject from a type-number -- can return NULL.

New reference

::

  char *
  PyArray_Zero(PyArrayObject *arr)

Get pointer to zero of correct type for array.

::

  char *
  PyArray_One(PyArrayObject *arr)

Get pointer to one of correct type for array

::

  PyObject *
  PyArray_CastToType(PyArrayObject *arr, PyArray_Descr *dtype, int
                     is_f_order)

For backward compatibility

Cast an array using typecode structure.
steals reference to at --- cannot be NULL

This function always makes a copy of arr, even if the dtype
doesn't change.

::

  int
  PyArray_CastTo(PyArrayObject *out, PyArrayObject *mp)

Cast to an already created array.

::

  int
  PyArray_CastAnyTo(PyArrayObject *out, PyArrayObject *mp)

Cast to an already created array.  Arrays don't have to be "broadcastable"
Only requirement is they have the same number of elements.

::

  int
  PyArray_CanCastSafely(int fromtype, int totype)

Check the type coercion rules.

::

  npy_bool
  PyArray_CanCastTo(PyArray_Descr *from, PyArray_Descr *to)

leaves reference count alone --- cannot be NULL

PyArray_CanCastTypeTo is equivalent to this, but adds a 'casting'
parameter.

::

  int
  PyArray_ObjectType(PyObject *op, int minimum_type)

Return the typecode of the array a Python object would be converted to

Returns the type number the result should have, or NPY_NOTYPE on error.

::

  PyArray_Descr *
  PyArray_DescrFromObject(PyObject *op, PyArray_Descr *mintype)

new reference -- accepts NULL for mintype

::

  PyArrayObject **
  PyArray_ConvertToCommonType(PyObject *op, int *retn)


::

  PyArray_Descr *
  PyArray_DescrFromScalar(PyObject *sc)

Return descr object from array scalar.

New reference

::

  PyArray_Descr *
  PyArray_DescrFromTypeObject(PyObject *type)


::

  npy_intp
  PyArray_Size(PyObject *op)

Compute the size of an array (in number of items)

::

  PyObject *
  PyArray_Scalar(void *data, PyArray_Descr *descr, PyObject *base)

Get scalar-equivalent to a region of memory described by a descriptor.

::

  PyObject *
  PyArray_FromScalar(PyObject *scalar, PyArray_Descr *outcode)

Get 0-dim array from scalar

0-dim array from array-scalar object
always contains a copy of the data
unless outcode is NULL, it is of void type and the referrer does
not own it either.

steals reference to outcode

::

  void
  PyArray_ScalarAsCtype(PyObject *scalar, void *ctypeptr)

Convert to c-type

no error checking is performed -- ctypeptr must be same type as scalar
in case of flexible type, the data is not copied
into ctypeptr which is expected to be a pointer to pointer

::

  int
  PyArray_CastScalarToCtype(PyObject *scalar, void
                            *ctypeptr, PyArray_Descr *outcode)

Cast Scalar to c-type

The output buffer must be large-enough to receive the value
Even for flexible types which is different from ScalarAsCtype
where only a reference for flexible types is returned

This may not work right on narrow builds for NumPy unicode scalars.

::

  int
  PyArray_CastScalarDirect(PyObject *scalar, PyArray_Descr
                           *indescr, void *ctypeptr, int outtype)

Cast Scalar to c-type

::

  PyObject *
  PyArray_ScalarFromObject(PyObject *object)

Get an Array Scalar From a Python Object

Returns NULL if unsuccessful but error is only set if another error occurred.
Currently only Numeric-like object supported.

::

  PyArray_VectorUnaryFunc *
  PyArray_GetCastFunc(PyArray_Descr *descr, int type_num)

Get a cast function to cast from the input descriptor to the
output type_number (must be a registered data-type).
Returns NULL if un-successful.

::

  PyObject *
  PyArray_FromDims(int nd, int *d, int type)

Construct an empty array from dimensions and typenum

::

  PyObject *
  PyArray_FromDimsAndDataAndDescr(int nd, int *d, PyArray_Descr
                                  *descr, char *data)

Like FromDimsAndData but uses the Descr structure instead of typecode
as input.

::

  PyObject *
  PyArray_FromAny(PyObject *op, PyArray_Descr *newtype, int
                  min_depth, int max_depth, int flags, PyObject
                  *context)

Does not check for NPY_ARRAY_ENSURECOPY and NPY_ARRAY_NOTSWAPPED in flags
Steals a reference to newtype --- which can be NULL

::

  PyObject *
  PyArray_EnsureArray(PyObject *op)

This is a quick wrapper around PyArray_FromAny(op, NULL, 0, 0, ENSUREARRAY)
that special cases Arrays and PyArray_Scalars up front
It *steals a reference* to the object
It also guarantees that the result is PyArray_Type
Because it decrefs op if any conversion needs to take place
so it can be used like PyArray_EnsureArray(some_function(...))

::

  PyObject *
  PyArray_EnsureAnyArray(PyObject *op)


::

  PyObject *
  PyArray_FromFile(FILE *fp, PyArray_Descr *dtype, npy_intp num, char
                   *sep)


Given a ``FILE *`` pointer ``fp``, and a ``PyArray_Descr``, return an
array corresponding to the data encoded in that file.

If the dtype is NULL, the default array type is used (double).
If non-null, the reference is stolen.

The number of elements to read is given as ``num``; if it is < 0, then
then as many as possible are read.

If ``sep`` is NULL or empty, then binary data is assumed, else
text data, with ``sep`` as the separator between elements. Whitespace in
the separator matches any length of whitespace in the text, and a match
for whitespace around the separator is added.

For memory-mapped files, use the buffer interface. No more data than
necessary is read by this routine.

::

  PyObject *
  PyArray_FromString(char *data, npy_intp slen, PyArray_Descr
                     *dtype, npy_intp num, char *sep)


Given a pointer to a string ``data``, a string length ``slen``, and
a ``PyArray_Descr``, return an array corresponding to the data
encoded in that string.

If the dtype is NULL, the default array type is used (double).
If non-null, the reference is stolen.

If ``slen`` is < 0, then the end of string is used for text data.
It is an error for ``slen`` to be < 0 for binary data (since embedded NULLs
would be the norm).

The number of elements to read is given as ``num``; if it is < 0, then
then as many as possible are read.

If ``sep`` is NULL or empty, then binary data is assumed, else
text data, with ``sep`` as the separator between elements. Whitespace in
the separator matches any length of whitespace in the text, and a match
for whitespace around the separator is added.

::

  PyObject *
  PyArray_FromBuffer(PyObject *buf, PyArray_Descr *type, npy_intp
                     count, npy_intp offset)


::

  PyObject *
  PyArray_FromIter(PyObject *obj, PyArray_Descr *dtype, npy_intp count)


steals a reference to dtype (which cannot be NULL)

::

  PyObject *
  PyArray_Return(PyArrayObject *mp)


Return either an array or the appropriate Python object if the array
is 0d and matches a Python type.

::

  PyObject *
  PyArray_GetField(PyArrayObject *self, PyArray_Descr *typed, int
                   offset)

Get a subset of bytes from each element of the array

::

  int
  PyArray_SetField(PyArrayObject *self, PyArray_Descr *dtype, int
                   offset, PyObject *val)

Set a subset of bytes from each element of the array

::

  PyObject *
  PyArray_Byteswap(PyArrayObject *self, npy_bool inplace)


::

  PyObject *
  PyArray_Resize(PyArrayObject *self, PyArray_Dims *newshape, int
                 refcheck, NPY_ORDER order)

Resize (reallocate data).  Only works if nothing else is referencing this
array and it is contiguous.  If refcheck is 0, then the reference count is
not checked and assumed to be 1.  You still must own this data and have no
weak-references and no base object.

::

  int
  PyArray_MoveInto(PyArrayObject *dst, PyArrayObject *src)

Move the memory of one array into another, allowing for overlapping data.

Returns 0 on success, negative on failure.

::

  int
  PyArray_CopyInto(PyArrayObject *dst, PyArrayObject *src)

Copy an Array into another array.
Broadcast to the destination shape if necessary.

Returns 0 on success, -1 on failure.

::

  int
  PyArray_CopyAnyInto(PyArrayObject *dst, PyArrayObject *src)

Copy an Array into another array -- memory must not overlap
Does not require src and dest to have "broadcastable" shapes
(only the same number of elements).

TODO: For NumPy 2.0, this could accept an order parameter which
only allows NPY_CORDER and NPY_FORDER.  Could also rename
this to CopyAsFlat to make the name more intuitive.

Returns 0 on success, -1 on error.

::

  int
  PyArray_CopyObject(PyArrayObject *dest, PyObject *src_object)


::

  PyObject *
  PyArray_NewCopy(PyArrayObject *obj, NPY_ORDER order)

Copy an array.

::

  PyObject *
  PyArray_ToList(PyArrayObject *self)

To List

::

  PyObject *
  PyArray_ToString(PyArrayObject *self, NPY_ORDER order)


::

  int
  PyArray_ToFile(PyArrayObject *self, FILE *fp, char *sep, char *format)

To File

::

  int
  PyArray_Dump(PyObject *self, PyObject *file, int protocol)


::

  PyObject *
  PyArray_Dumps(PyObject *self, int protocol)


::

  int
  PyArray_ValidType(int type)

Is the typenum valid?

::

  void
  PyArray_UpdateFlags(PyArrayObject *ret, int flagmask)

Update Several Flags at once.

::

  PyObject *
  PyArray_New(PyTypeObject *subtype, int nd, npy_intp *dims, int
              type_num, npy_intp *strides, void *data, int itemsize, int
              flags, PyObject *obj)

Generic new array creation routine.

::

  PyObject *
  PyArray_NewFromDescr(PyTypeObject *subtype, PyArray_Descr *descr, int
                       nd, npy_intp *dims, npy_intp *strides, void
                       *data, int flags, PyObject *obj)

Generic new array creation routine.

steals a reference to descr (even on failure)

::

  PyArray_Descr *
  PyArray_DescrNew(PyArray_Descr *base)

base cannot be NULL

::

  PyArray_Descr *
  PyArray_DescrNewFromType(int type_num)


::

  double
  PyArray_GetPriority(PyObject *obj, double default_)

Get Priority from object

::

  PyObject *
  PyArray_IterNew(PyObject *obj)

Get Iterator.

::

  PyObject *
  PyArray_MultiIterNew(int n, ... )

Get MultiIterator,

::

  int
  PyArray_PyIntAsInt(PyObject *o)


::

  npy_intp
  PyArray_PyIntAsIntp(PyObject *o)


::

  int
  PyArray_Broadcast(PyArrayMultiIterObject *mit)


::

  void
  PyArray_FillObjectArray(PyArrayObject *arr, PyObject *obj)

Assumes contiguous

::

  int
  PyArray_FillWithScalar(PyArrayObject *arr, PyObject *obj)


::

  npy_bool
  PyArray_CheckStrides(int elsize, int nd, npy_intp numbytes, npy_intp
                       offset, npy_intp *dims, npy_intp *newstrides)


::

  PyArray_Descr *
  PyArray_DescrNewByteorder(PyArray_Descr *self, char newendian)


returns a copy of the PyArray_Descr structure with the byteorder
altered:
no arguments:  The byteorder is swapped (in all subfields as well)
single argument:  The byteorder is forced to the given state
(in all subfields as well)

Valid states:  ('big', '>') or ('little' or '<')
('native', or '=')

If a descr structure with | is encountered it's own
byte-order is not changed but any fields are:


Deep bytorder change of a data-type descriptor
Leaves reference count of self unchanged --- does not DECREF self ***

::

  PyObject *
  PyArray_IterAllButAxis(PyObject *obj, int *inaxis)

Get Iterator that iterates over all but one axis (don't use this with
PyArray_ITER_GOTO1D).  The axis will be over-written if negative
with the axis having the smallest stride.

::

  PyObject *
  PyArray_CheckFromAny(PyObject *op, PyArray_Descr *descr, int
                       min_depth, int max_depth, int requires, PyObject
                       *context)

steals a reference to descr -- accepts NULL

::

  PyObject *
  PyArray_FromArray(PyArrayObject *arr, PyArray_Descr *newtype, int
                    flags)

steals reference to newtype --- acc. NULL

::

  PyObject *
  PyArray_FromInterface(PyObject *origin)


::

  PyObject *
  PyArray_FromStructInterface(PyObject *input)


::

  PyObject *
  PyArray_FromArrayAttr(PyObject *op, PyArray_Descr *typecode, PyObject
                        *context)


::

  NPY_SCALARKIND
  PyArray_ScalarKind(int typenum, PyArrayObject **arr)

ScalarKind

Returns the scalar kind of a type number, with an
optional tweak based on the scalar value itself.
If no scalar is provided, it returns INTPOS_SCALAR
for both signed and unsigned integers, otherwise
it checks the sign of any signed integer to choose
INTNEG_SCALAR when appropriate.

::

  int
  PyArray_CanCoerceScalar(int thistype, int neededtype, NPY_SCALARKIND
                          scalar)


Determines whether the data type 'thistype', with
scalar kind 'scalar', can be coerced into 'neededtype'.

::

  PyObject *
  PyArray_NewFlagsObject(PyObject *obj)


Get New ArrayFlagsObject

::

  npy_bool
  PyArray_CanCastScalar(PyTypeObject *from, PyTypeObject *to)

See if array scalars can be cast.

TODO: For NumPy 2.0, add a NPY_CASTING parameter.

::

  int
  PyArray_CompareUCS4(npy_ucs4 *s1, npy_ucs4 *s2, size_t len)


::

  int
  PyArray_RemoveSmallest(PyArrayMultiIterObject *multi)

Adjusts previously broadcasted iterators so that the axis with
the smallest sum of iterator strides is not iterated over.
Returns dimension which is smallest in the range [0,multi->nd).
A -1 is returned if multi->nd == 0.

don't use with PyArray_ITER_GOTO1D because factors are not adjusted

::

  int
  PyArray_ElementStrides(PyObject *obj)


::

  void
  PyArray_Item_INCREF(char *data, PyArray_Descr *descr)


::

  void
  PyArray_Item_XDECREF(char *data, PyArray_Descr *descr)


::

  PyObject *
  PyArray_FieldNames(PyObject *fields)

Return the tuple of ordered field names from a dictionary.

::

  PyObject *
  PyArray_Transpose(PyArrayObject *ap, PyArray_Dims *permute)

Return Transpose.

::

  PyObject *
  PyArray_TakeFrom(PyArrayObject *self0, PyObject *indices0, int
                   axis, PyArrayObject *out, NPY_CLIPMODE clipmode)

Take

::

  PyObject *
  PyArray_PutTo(PyArrayObject *self, PyObject*values0, PyObject
                *indices0, NPY_CLIPMODE clipmode)

Put values into an array

::

  PyObject *
  PyArray_PutMask(PyArrayObject *self, PyObject*values0, PyObject*mask0)

Put values into an array according to a mask.

::

  PyObject *
  PyArray_Repeat(PyArrayObject *aop, PyObject *op, int axis)

Repeat the array.

::

  PyObject *
  PyArray_Choose(PyArrayObject *ip, PyObject *op, PyArrayObject
                 *out, NPY_CLIPMODE clipmode)


::

  int
  PyArray_Sort(PyArrayObject *op, int axis, NPY_SORTKIND which)

Sort an array in-place

::

  PyObject *
  PyArray_ArgSort(PyArrayObject *op, int axis, NPY_SORTKIND which)

ArgSort an array

::

  PyObject *
  PyArray_SearchSorted(PyArrayObject *op1, PyObject *op2, NPY_SEARCHSIDE
                       side, PyObject *perm)


Search the sorted array op1 for the location of the items in op2. The
result is an array of indexes, one for each element in op2, such that if
the item were to be inserted in op1 just before that index the array
would still be in sorted order.

Parameters
----------
op1 : PyArrayObject *
Array to be searched, must be 1-D.
op2 : PyObject *
Array of items whose insertion indexes in op1 are wanted
side : {NPY_SEARCHLEFT, NPY_SEARCHRIGHT}
If NPY_SEARCHLEFT, return first valid insertion indexes
If NPY_SEARCHRIGHT, return last valid insertion indexes
perm : PyObject *
Permutation array that sorts op1 (optional)

Returns
-------
ret : PyObject *
New reference to npy_intp array containing indexes where items in op2
could be validly inserted into op1. NULL on error.

Notes
-----
Binary search is used to find the indexes.

::

  PyObject *
  PyArray_ArgMax(PyArrayObject *op, int axis, PyArrayObject *out)

ArgMax

::

  PyObject *
  PyArray_ArgMin(PyArrayObject *op, int axis, PyArrayObject *out)

ArgMin

::

  PyObject *
  PyArray_Reshape(PyArrayObject *self, PyObject *shape)

Reshape

::

  PyObject *
  PyArray_Newshape(PyArrayObject *self, PyArray_Dims *newdims, NPY_ORDER
                   order)

New shape for an array

::

  PyObject *
  PyArray_Squeeze(PyArrayObject *self)


return a new view of the array object with all of its unit-length
dimensions squeezed out if needed, otherwise
return the same array.

::

  PyObject *
  PyArray_View(PyArrayObject *self, PyArray_Descr *type, PyTypeObject
               *pytype)

View
steals a reference to type -- accepts NULL

::

  PyObject *
  PyArray_SwapAxes(PyArrayObject *ap, int a1, int a2)

SwapAxes

::

  PyObject *
  PyArray_Max(PyArrayObject *ap, int axis, PyArrayObject *out)

Max

::

  PyObject *
  PyArray_Min(PyArrayObject *ap, int axis, PyArrayObject *out)

Min

::

  PyObject *
  PyArray_Ptp(PyArrayObject *ap, int axis, PyArrayObject *out)

Ptp

::

  PyObject *
  PyArray_Mean(PyArrayObject *self, int axis, int rtype, PyArrayObject
               *out)

Mean

::

  PyObject *
  PyArray_Trace(PyArrayObject *self, int offset, int axis1, int
                axis2, int rtype, PyArrayObject *out)

Trace

::

  PyObject *
  PyArray_Diagonal(PyArrayObject *self, int offset, int axis1, int
                   axis2)

Diagonal

In NumPy versions prior to 1.7,  this function always returned a copy of
the diagonal array. In 1.7, the code has been updated to compute a view
onto 'self', but it still copies this array before returning, as well as
setting the internal WARN_ON_WRITE flag. In a future version, it will
simply return a view onto self.

::

  PyObject *
  PyArray_Clip(PyArrayObject *self, PyObject *min, PyObject
               *max, PyArrayObject *out)

Clip

::

  PyObject *
  PyArray_Conjugate(PyArrayObject *self, PyArrayObject *out)

Conjugate

::

  PyObject *
  PyArray_Nonzero(PyArrayObject *self)

Nonzero

TODO: In NumPy 2.0, should make the iteration order a parameter.

::

  PyObject *
  PyArray_Std(PyArrayObject *self, int axis, int rtype, PyArrayObject
              *out, int variance)

Set variance to 1 to by-pass square-root calculation and return variance
Std

::

  PyObject *
  PyArray_Sum(PyArrayObject *self, int axis, int rtype, PyArrayObject
              *out)

Sum

::

  PyObject *
  PyArray_CumSum(PyArrayObject *self, int axis, int rtype, PyArrayObject
                 *out)

CumSum

::

  PyObject *
  PyArray_Prod(PyArrayObject *self, int axis, int rtype, PyArrayObject
               *out)

Prod

::

  PyObject *
  PyArray_CumProd(PyArrayObject *self, int axis, int
                  rtype, PyArrayObject *out)

CumProd

::

  PyObject *
  PyArray_All(PyArrayObject *self, int axis, PyArrayObject *out)

All

::

  PyObject *
  PyArray_Any(PyArrayObject *self, int axis, PyArrayObject *out)

Any

::

  PyObject *
  PyArray_Compress(PyArrayObject *self, PyObject *condition, int
                   axis, PyArrayObject *out)

Compress

::

  PyObject *
  PyArray_Flatten(PyArrayObject *a, NPY_ORDER order)

Flatten

::

  PyObject *
  PyArray_Ravel(PyArrayObject *arr, NPY_ORDER order)

Ravel
Returns a contiguous array

::

  npy_intp
  PyArray_MultiplyList(npy_intp *l1, int n)

Multiply a List

::

  int
  PyArray_MultiplyIntList(int *l1, int n)

Multiply a List of ints

::

  void *
  PyArray_GetPtr(PyArrayObject *obj, npy_intp*ind)

Produce a pointer into array

::

  int
  PyArray_CompareLists(npy_intp *l1, npy_intp *l2, int n)

Compare Lists

::

  int
  PyArray_AsCArray(PyObject **op, void *ptr, npy_intp *dims, int
                   nd, PyArray_Descr*typedescr)

Simulate a C-array
steals a reference to typedescr -- can be NULL

::

  int
  PyArray_As1D(PyObject **op, char **ptr, int *d1, int typecode)

Convert to a 1D C-array

::

  int
  PyArray_As2D(PyObject **op, char ***ptr, int *d1, int *d2, int
               typecode)

Convert to a 2D C-array

::

  int
  PyArray_Free(PyObject *op, void *ptr)

Free pointers created if As2D is called

::

  int
  PyArray_Converter(PyObject *object, PyObject **address)


Useful to pass as converter function for O& processing in PyArgs_ParseTuple.

This conversion function can be used with the "O&" argument for
PyArg_ParseTuple.  It will immediately return an object of array type
or will convert to a NPY_ARRAY_CARRAY any other object.

If you use PyArray_Converter, you must DECREF the array when finished
as you get a new reference to it.

::

  int
  PyArray_IntpFromSequence(PyObject *seq, npy_intp *vals, int maxvals)

PyArray_IntpFromSequence
Returns the number of dimensions or -1 if an error occurred.
vals must be large enough to hold maxvals

::

  PyObject *
  PyArray_Concatenate(PyObject *op, int axis)

Concatenate

Concatenate an arbitrary Python sequence into an array.
op is a python object supporting the sequence interface.
Its elements will be concatenated together to form a single
multidimensional array. If axis is NPY_MAXDIMS or bigger, then
each sequence object will be flattened before concatenation

::

  PyObject *
  PyArray_InnerProduct(PyObject *op1, PyObject *op2)

Numeric.innerproduct(a,v)

::

  PyObject *
  PyArray_MatrixProduct(PyObject *op1, PyObject *op2)

Numeric.matrixproduct(a,v)
just like inner product but does the swapaxes stuff on the fly

::

  PyObject *
  PyArray_CopyAndTranspose(PyObject *op)

Copy and Transpose

Could deprecate this function, as there isn't a speed benefit over
calling Transpose and then Copy.

::

  PyObject *
  PyArray_Correlate(PyObject *op1, PyObject *op2, int mode)

Numeric.correlate(a1,a2,mode)

::

  int
  PyArray_TypestrConvert(int itemsize, int gentype)

Typestr converter

::

  int
  PyArray_DescrConverter(PyObject *obj, PyArray_Descr **at)

Get typenum from an object -- None goes to NPY_DEFAULT_TYPE
This function takes a Python object representing a type and converts it
to a the correct PyArray_Descr * structure to describe the type.

Many objects can be used to represent a data-type which in NumPy is
quite a flexible concept.

This is the central code that converts Python objects to
Type-descriptor objects that are used throughout numpy.

Returns a new reference in *at, but the returned should not be
modified as it may be one of the canonical immutable objects or
a reference to the input obj.

::

  int
  PyArray_DescrConverter2(PyObject *obj, PyArray_Descr **at)

Get typenum from an object -- None goes to NULL

::

  int
  PyArray_IntpConverter(PyObject *obj, PyArray_Dims *seq)

Get intp chunk from sequence

This function takes a Python sequence object and allocates and
fills in an intp array with the converted values.

Remember to free the pointer seq.ptr when done using
PyDimMem_FREE(seq.ptr)**

::

  int
  PyArray_BufferConverter(PyObject *obj, PyArray_Chunk *buf)

Get buffer chunk from object

this function takes a Python object which exposes the (single-segment)
buffer interface and returns a pointer to the data segment

You should increment the reference count by one of buf->base
if you will hang on to a reference

You only get a borrowed reference to the object. Do not free the
memory...

::

  int
  PyArray_AxisConverter(PyObject *obj, int *axis)

Get axis from an object (possibly None) -- a converter function,

See also PyArray_ConvertMultiAxis, which also handles a tuple of axes.

::

  int
  PyArray_BoolConverter(PyObject *object, npy_bool *val)

Convert an object to true / false

::

  int
  PyArray_ByteorderConverter(PyObject *obj, char *endian)

Convert object to endian

::

  int
  PyArray_OrderConverter(PyObject *object, NPY_ORDER *val)

Convert an object to FORTRAN / C / ANY / KEEP

::

  unsigned char
  PyArray_EquivTypes(PyArray_Descr *type1, PyArray_Descr *type2)


This function returns true if the two typecodes are
equivalent (same basic kind and same itemsize).

::

  PyObject *
  PyArray_Zeros(int nd, npy_intp *dims, PyArray_Descr *type, int
                is_f_order)

Zeros

steal a reference
accepts NULL type

::

  PyObject *
  PyArray_Empty(int nd, npy_intp *dims, PyArray_Descr *type, int
                is_f_order)

Empty

accepts NULL type
steals referenct to type

::

  PyObject *
  PyArray_Where(PyObject *condition, PyObject *x, PyObject *y)

Where

::

  PyObject *
  PyArray_Arange(double start, double stop, double step, int type_num)

Arange,

::

  PyObject *
  PyArray_ArangeObj(PyObject *start, PyObject *stop, PyObject
                    *step, PyArray_Descr *dtype)


ArangeObj,

this doesn't change the references

::

  int
  PyArray_SortkindConverter(PyObject *obj, NPY_SORTKIND *sortkind)

Convert object to sort kind

::

  PyObject *
  PyArray_LexSort(PyObject *sort_keys, int axis)

LexSort an array providing indices that will sort a collection of arrays
lexicographically.  The first key is sorted on first, followed by the second key
-- requires that arg"merge"sort is available for each sort_key

Returns an index array that shows the indexes for the lexicographic sort along
the given axis.

::

  PyObject *
  PyArray_Round(PyArrayObject *a, int decimals, PyArrayObject *out)

Round

::

  unsigned char
  PyArray_EquivTypenums(int typenum1, int typenum2)


::

  int
  PyArray_RegisterDataType(PyArray_Descr *descr)

Register Data type
Does not change the reference count of descr

::

  int
  PyArray_RegisterCastFunc(PyArray_Descr *descr, int
                           totype, PyArray_VectorUnaryFunc *castfunc)

Register Casting Function
Replaces any function currently stored.

::

  int
  PyArray_RegisterCanCast(PyArray_Descr *descr, int
                          totype, NPY_SCALARKIND scalar)

Register a type number indicating that a descriptor can be cast
to it safely

::

  void
  PyArray_InitArrFuncs(PyArray_ArrFuncs *f)

Initialize arrfuncs to NULL

::

  PyObject *
  PyArray_IntTupleFromIntp(int len, npy_intp *vals)

PyArray_IntTupleFromIntp

::

  int
  PyArray_TypeNumFromName(char *str)


::

  int
  PyArray_ClipmodeConverter(PyObject *object, NPY_CLIPMODE *val)

Convert an object to NPY_RAISE / NPY_CLIP / NPY_WRAP

::

  int
  PyArray_OutputConverter(PyObject *object, PyArrayObject **address)

Useful to pass as converter function for O& processing in
PyArgs_ParseTuple for output arrays

::

  PyObject *
  PyArray_BroadcastToShape(PyObject *obj, npy_intp *dims, int nd)

Get Iterator broadcast to a particular shape

::

  void
  _PyArray_SigintHandler(int signum)


::

  void*
  _PyArray_GetSigintBuf(void )


::

  int
  PyArray_DescrAlignConverter(PyObject *obj, PyArray_Descr **at)


Get type-descriptor from an object forcing alignment if possible
None goes to DEFAULT type.

any object with the .fields attribute and/or .itemsize attribute (if the
.fields attribute does not give the total size -- i.e. a partial record
naming).  If itemsize is given it must be >= size computed from fields

The .fields attribute must return a convertible dictionary if present.
Result inherits from NPY_VOID.

::

  int
  PyArray_DescrAlignConverter2(PyObject *obj, PyArray_Descr **at)


Get type-descriptor from an object forcing alignment if possible
None goes to NULL.

::

  int
  PyArray_SearchsideConverter(PyObject *obj, void *addr)

Convert object to searchsorted side

::

  PyObject *
  PyArray_CheckAxis(PyArrayObject *arr, int *axis, int flags)

PyArray_CheckAxis

check that axis is valid
convert 0-d arrays to 1-d arrays

::

  npy_intp
  PyArray_OverflowMultiplyList(npy_intp *l1, int n)

Multiply a List of Non-negative numbers with over-flow detection.

::

  int
  PyArray_CompareString(char *s1, char *s2, size_t len)


::

  PyObject *
  PyArray_MultiIterFromObjects(PyObject **mps, int n, int nadd, ... )

Get MultiIterator from array of Python objects and any additional

PyObject **mps -- array of PyObjects
int n - number of PyObjects in the array
int nadd - number of additional arrays to include in the iterator.

Returns a multi-iterator object.

::

  int
  PyArray_GetEndianness(void )


::

  unsigned int
  PyArray_GetNDArrayCFeatureVersion(void )

Returns the built-in (at compilation time) C API version

::

  PyObject *
  PyArray_Correlate2(PyObject *op1, PyObject *op2, int mode)

correlate(a1,a2,mode)

This function computes the usual correlation (correlate(a1, a2) !=
correlate(a2, a1), and conjugate the second argument for complex inputs

::

  PyObject*
  PyArray_NeighborhoodIterNew(PyArrayIterObject *x, npy_intp
                              *bounds, int mode, PyArrayObject*fill)

A Neighborhood Iterator object.

::

  void
  PyArray_SetDatetimeParseFunction(PyObject *op)

This function is scheduled to be removed

TO BE REMOVED - NOT USED INTERNALLY.

::

  void
  PyArray_DatetimeToDatetimeStruct(npy_datetime val, NPY_DATETIMEUNIT
                                   fr, npy_datetimestruct *result)

Fill the datetime struct from the value and resolution unit.

TO BE REMOVED - NOT USED INTERNALLY.

::

  void
  PyArray_TimedeltaToTimedeltaStruct(npy_timedelta val, NPY_DATETIMEUNIT
                                     fr, npy_timedeltastruct *result)

Fill the timedelta struct from the timedelta value and resolution unit.

TO BE REMOVED - NOT USED INTERNALLY.

::

  npy_datetime
  PyArray_DatetimeStructToDatetime(NPY_DATETIMEUNIT
                                   fr, npy_datetimestruct *d)

Create a datetime value from a filled datetime struct and resolution unit.

TO BE REMOVED - NOT USED INTERNALLY.

::

  npy_datetime
  PyArray_TimedeltaStructToTimedelta(NPY_DATETIMEUNIT
                                     fr, npy_timedeltastruct *d)

Create a timdelta value from a filled timedelta struct and resolution unit.

TO BE REMOVED - NOT USED INTERNALLY.

::

  NpyIter *
  NpyIter_New(PyArrayObject *op, npy_uint32 flags, NPY_ORDER
              order, NPY_CASTING casting, PyArray_Descr*dtype)

Allocate a new iterator for one array object.

::

  NpyIter *
  NpyIter_MultiNew(int nop, PyArrayObject **op_in, npy_uint32
                   flags, NPY_ORDER order, NPY_CASTING
                   casting, npy_uint32 *op_flags, PyArray_Descr
                   **op_request_dtypes)

Allocate a new iterator for more than one array object, using
standard NumPy broadcasting rules and the default buffer size.

::

  NpyIter *
  NpyIter_AdvancedNew(int nop, PyArrayObject **op_in, npy_uint32
                      flags, NPY_ORDER order, NPY_CASTING
                      casting, npy_uint32 *op_flags, PyArray_Descr
                      **op_request_dtypes, int oa_ndim, int
                      **op_axes, npy_intp *itershape, npy_intp
                      buffersize)

Allocate a new iterator for multiple array objects, and advanced
options for controlling the broadcasting, shape, and buffer size.

::

  NpyIter *
  NpyIter_Copy(NpyIter *iter)

Makes a copy of the iterator

::

  int
  NpyIter_Deallocate(NpyIter *iter)

Deallocate an iterator

::

  npy_bool
  NpyIter_HasDelayedBufAlloc(NpyIter *iter)

Whether the buffer allocation is being delayed

::

  npy_bool
  NpyIter_HasExternalLoop(NpyIter *iter)

Whether the iterator handles the inner loop

::

  int
  NpyIter_EnableExternalLoop(NpyIter *iter)

Removes the inner loop handling (so HasExternalLoop returns true)

::

  npy_intp *
  NpyIter_GetInnerStrideArray(NpyIter *iter)

Get the array of strides for the inner loop (when HasExternalLoop is true)

This function may be safely called without holding the Python GIL.

::

  npy_intp *
  NpyIter_GetInnerLoopSizePtr(NpyIter *iter)

Get a pointer to the size of the inner loop  (when HasExternalLoop is true)

This function may be safely called without holding the Python GIL.

::

  int
  NpyIter_Reset(NpyIter *iter, char **errmsg)

Resets the iterator to its initial state

If errmsg is non-NULL, it should point to a variable which will
receive the error message, and no Python exception will be set.
This is so that the function can be called from code not holding
the GIL.

::

  int
  NpyIter_ResetBasePointers(NpyIter *iter, char **baseptrs, char
                            **errmsg)

Resets the iterator to its initial state, with new base data pointers.
This function requires great caution.

If errmsg is non-NULL, it should point to a variable which will
receive the error message, and no Python exception will be set.
This is so that the function can be called from code not holding
the GIL.

::

  int
  NpyIter_ResetToIterIndexRange(NpyIter *iter, npy_intp istart, npy_intp
                                iend, char **errmsg)

Resets the iterator to a new iterator index range

If errmsg is non-NULL, it should point to a variable which will
receive the error message, and no Python exception will be set.
This is so that the function can be called from code not holding
the GIL.

::

  int
  NpyIter_GetNDim(NpyIter *iter)

Gets the number of dimensions being iterated

::

  int
  NpyIter_GetNOp(NpyIter *iter)

Gets the number of operands being iterated

::

  NpyIter_IterNextFunc *
  NpyIter_GetIterNext(NpyIter *iter, char **errmsg)

Compute the specialized iteration function for an iterator

If errmsg is non-NULL, it should point to a variable which will
receive the error message, and no Python exception will be set.
This is so that the function can be called from code not holding
the GIL.

::

  npy_intp
  NpyIter_GetIterSize(NpyIter *iter)

Gets the number of elements being iterated

::

  void
  NpyIter_GetIterIndexRange(NpyIter *iter, npy_intp *istart, npy_intp
                            *iend)

Gets the range of iteration indices being iterated

::

  npy_intp
  NpyIter_GetIterIndex(NpyIter *iter)

Gets the current iteration index

::

  int
  NpyIter_GotoIterIndex(NpyIter *iter, npy_intp iterindex)

Sets the iterator position to the specified iterindex,
which matches the iteration order of the iterator.

Returns NPY_SUCCEED on success, NPY_FAIL on failure.

::

  npy_bool
  NpyIter_HasMultiIndex(NpyIter *iter)

Whether the iterator is tracking a multi-index

::

  int
  NpyIter_GetShape(NpyIter *iter, npy_intp *outshape)

Gets the broadcast shape if a multi-index is being tracked by the iterator,
otherwise gets the shape of the iteration as Fortran-order
(fastest-changing index first).

The reason Fortran-order is returned when a multi-index
is not enabled is that this is providing a direct view into how
the iterator traverses the n-dimensional space. The iterator organizes
its memory from fastest index to slowest index, and when
a multi-index is enabled, it uses a permutation to recover the original
order.

Returns NPY_SUCCEED or NPY_FAIL.

::

  NpyIter_GetMultiIndexFunc *
  NpyIter_GetGetMultiIndex(NpyIter *iter, char **errmsg)

Compute a specialized get_multi_index function for the iterator

If errmsg is non-NULL, it should point to a variable which will
receive the error message, and no Python exception will be set.
This is so that the function can be called from code not holding
the GIL.

::

  int
  NpyIter_GotoMultiIndex(NpyIter *iter, npy_intp *multi_index)

Sets the iterator to the specified multi-index, which must have the
correct number of entries for 'ndim'.  It is only valid
when NPY_ITER_MULTI_INDEX was passed to the constructor.  This operation
fails if the multi-index is out of bounds.

Returns NPY_SUCCEED on success, NPY_FAIL on failure.

::

  int
  NpyIter_RemoveMultiIndex(NpyIter *iter)

Removes multi-index support from an iterator.

Returns NPY_SUCCEED or NPY_FAIL.

::

  npy_bool
  NpyIter_HasIndex(NpyIter *iter)

Whether the iterator is tracking an index

::

  npy_bool
  NpyIter_IsBuffered(NpyIter *iter)

Whether the iterator is buffered

::

  npy_bool
  NpyIter_IsGrowInner(NpyIter *iter)

Whether the inner loop can grow if buffering is unneeded

::

  npy_intp
  NpyIter_GetBufferSize(NpyIter *iter)

Gets the size of the buffer, or 0 if buffering is not enabled

::

  npy_intp *
  NpyIter_GetIndexPtr(NpyIter *iter)

Get a pointer to the index, if it is being tracked

::

  int
  NpyIter_GotoIndex(NpyIter *iter, npy_intp flat_index)

If the iterator is tracking an index, sets the iterator
to the specified index.

Returns NPY_SUCCEED on success, NPY_FAIL on failure.

::

  char **
  NpyIter_GetDataPtrArray(NpyIter *iter)

Get the array of data pointers (1 per object being iterated)

This function may be safely called without holding the Python GIL.

::

  PyArray_Descr **
  NpyIter_GetDescrArray(NpyIter *iter)

Get the array of data type pointers (1 per object being iterated)

::

  PyArrayObject **
  NpyIter_GetOperandArray(NpyIter *iter)

Get the array of objects being iterated

::

  PyArrayObject *
  NpyIter_GetIterView(NpyIter *iter, npy_intp i)

Returns a view to the i-th object with the iterator's internal axes

::

  void
  NpyIter_GetReadFlags(NpyIter *iter, char *outreadflags)

Gets an array of read flags (1 per object being iterated)

::

  void
  NpyIter_GetWriteFlags(NpyIter *iter, char *outwriteflags)

Gets an array of write flags (1 per object being iterated)

::

  void
  NpyIter_DebugPrint(NpyIter *iter)

For debugging

::

  npy_bool
  NpyIter_IterationNeedsAPI(NpyIter *iter)

Whether the iteration loop, and in particular the iternext()
function, needs API access.  If this is true, the GIL must
be retained while iterating.

::

  void
  NpyIter_GetInnerFixedStrideArray(NpyIter *iter, npy_intp *out_strides)

Get an array of strides which are fixed.  Any strides which may
change during iteration receive the value NPY_MAX_INTP.  Once
the iterator is ready to iterate, call this to get the strides
which will always be fixed in the inner loop, then choose optimized
inner loop functions which take advantage of those fixed strides.

This function may be safely called without holding the Python GIL.

::

  int
  NpyIter_RemoveAxis(NpyIter *iter, int axis)

Removes an axis from iteration. This requires that NPY_ITER_MULTI_INDEX
was set for iterator creation, and does not work if buffering is
enabled. This function also resets the iterator to its initial state.

Returns NPY_SUCCEED or NPY_FAIL.

::

  npy_intp *
  NpyIter_GetAxisStrideArray(NpyIter *iter, int axis)

Gets the array of strides for the specified axis.
If the iterator is tracking a multi-index, gets the strides
for the axis specified, otherwise gets the strides for
the iteration axis as Fortran order (fastest-changing axis first).

Returns NULL if an error occurs.

::

  npy_bool
  NpyIter_RequiresBuffering(NpyIter *iter)

Whether the iteration could be done with no buffering.

::

  char **
  NpyIter_GetInitialDataPtrArray(NpyIter *iter)

Get the array of data pointers (1 per object being iterated),
directly into the arrays (never pointing to a buffer), for starting
unbuffered iteration. This always returns the addresses for the
iterator position as reset to iterator index 0.

These pointers are different from the pointers accepted by
NpyIter_ResetBasePointers, because the direction along some
axes may have been reversed, requiring base offsets.

This function may be safely called without holding the Python GIL.

::

  int
  NpyIter_CreateCompatibleStrides(NpyIter *iter, npy_intp
                                  itemsize, npy_intp *outstrides)

Builds a set of strides which are the same as the strides of an
output array created using the NPY_ITER_ALLOCATE flag, where NULL
was passed for op_axes.  This is for data packed contiguously,
but not necessarily in C or Fortran order. This should be used
together with NpyIter_GetShape and NpyIter_GetNDim.

A use case for this function is to match the shape and layout of
the iterator and tack on one or more dimensions.  For example,
in order to generate a vector per input value for a numerical gradient,
you pass in ndim*itemsize for itemsize, then add another dimension to
the end with size ndim and stride itemsize.  To do the Hessian matrix,
you do the same thing but add two dimensions, or take advantage of
the symmetry and pack it into 1 dimension with a particular encoding.

This function may only be called if the iterator is tracking a multi-index
and if NPY_ITER_DONT_NEGATE_STRIDES was used to prevent an axis from
being iterated in reverse order.

If an array is created with this method, simply adding 'itemsize'
for each iteration will traverse the new array matching the
iterator.

Returns NPY_SUCCEED or NPY_FAIL.

::

  int
  PyArray_CastingConverter(PyObject *obj, NPY_CASTING *casting)

Convert any Python object, *obj*, to an NPY_CASTING enum.

::

  npy_intp
  PyArray_CountNonzero(PyArrayObject *self)

Counts the number of non-zero elements in the array.

Returns -1 on error.

::

  PyArray_Descr *
  PyArray_PromoteTypes(PyArray_Descr *type1, PyArray_Descr *type2)

Produces the smallest size and lowest kind type to which both
input types can be cast.

::

  PyArray_Descr *
  PyArray_MinScalarType(PyArrayObject *arr)

If arr is a scalar (has 0 dimensions) with a built-in number data type,
finds the smallest type size/kind which can still represent its data.
Otherwise, returns the array's data type.


::

  PyArray_Descr *
  PyArray_ResultType(npy_intp narrs, PyArrayObject **arr, npy_intp
                     ndtypes, PyArray_Descr **dtypes)

Produces the result type of a bunch of inputs, using the UFunc
type promotion rules. Use this function when you have a set of
input arrays, and need to determine an output array dtype.

If all the inputs are scalars (have 0 dimensions) or the maximum "kind"
of the scalars is greater than the maximum "kind" of the arrays, does
a regular type promotion.

Otherwise, does a type promotion on the MinScalarType
of all the inputs.  Data types passed directly are treated as array
types.


::

  npy_bool
  PyArray_CanCastArrayTo(PyArrayObject *arr, PyArray_Descr
                         *to, NPY_CASTING casting)

Returns 1 if the array object may be cast to the given data type using
the casting rule, 0 otherwise.  This differs from PyArray_CanCastTo in
that it handles scalar arrays (0 dimensions) specially, by checking
their value.

::

  npy_bool
  PyArray_CanCastTypeTo(PyArray_Descr *from, PyArray_Descr
                        *to, NPY_CASTING casting)

Returns true if data of type 'from' may be cast to data of type
'to' according to the rule 'casting'.

::

  PyArrayObject *
  PyArray_EinsteinSum(char *subscripts, npy_intp nop, PyArrayObject
                      **op_in, PyArray_Descr *dtype, NPY_ORDER
                      order, NPY_CASTING casting, PyArrayObject *out)

This function provides summation of array elements according to
the Einstein summation convention.  For example:
- trace(a)        -> einsum("ii", a)
- transpose(a)    -> einsum("ji", a)
- multiply(a,b)   -> einsum(",", a, b)
- inner(a,b)      -> einsum("i,i", a, b)
- outer(a,b)      -> einsum("i,j", a, b)
- matvec(a,b)     -> einsum("ij,j", a, b)
- matmat(a,b)     -> einsum("ij,jk", a, b)

subscripts: The string of subscripts for einstein summation.
nop:        The number of operands
op_in:      The array of operands
dtype:      Either NULL, or the data type to force the calculation as.
order:      The order for the calculation/the output axes.
casting:    What kind of casts should be permitted.
out:        Either NULL, or an array into which the output should be placed.

By default, the labels get placed in alphabetical order
at the end of the output. So, if c = einsum("i,j", a, b)
then c[i,j] == a[i]*b[j], but if c = einsum("j,i", a, b)
then c[i,j] = a[j]*b[i].

Alternatively, you can control the output order or prevent
an axis from being summed/force an axis to be summed by providing
indices for the output. This allows us to turn 'trace' into
'diag', for example.
- diag(a)         -> einsum("ii->i", a)
- sum(a, axis=0)  -> einsum("i...->", a)

Subscripts at the beginning and end may be specified by
putting an ellipsis "..." in the middle.  For example,
the function einsum("i...i", a) takes the diagonal of
the first and last dimensions of the operand, and
einsum("ij...,jk...->ik...") takes the matrix product using
the first two indices of each operand instead of the last two.

When there is only one operand, no axes being summed, and
no output parameter, this function returns a view
into the operand instead of making a copy.

::

  PyObject *
  PyArray_NewLikeArray(PyArrayObject *prototype, NPY_ORDER
                       order, PyArray_Descr *dtype, int subok)

Creates a new array with the same shape as the provided one,
with possible memory layout order and data type changes.

prototype - The array the new one should be like.
order     - NPY_CORDER - C-contiguous result.
NPY_FORTRANORDER - Fortran-contiguous result.
NPY_ANYORDER - Fortran if prototype is Fortran, C otherwise.
NPY_KEEPORDER - Keeps the axis ordering of prototype.
dtype     - If not NULL, overrides the data type of the result.
subok     - If 1, use the prototype's array subtype, otherwise
always create a base-class array.

NOTE: If dtype is not NULL, steals the dtype reference.

::

  int
  PyArray_GetArrayParamsFromObject(PyObject *op, PyArray_Descr
                                   *requested_dtype, npy_bool
                                   writeable, PyArray_Descr
                                   **out_dtype, int *out_ndim, npy_intp
                                   *out_dims, PyArrayObject
                                   **out_arr, PyObject *context)

Retrieves the array parameters for viewing/converting an arbitrary
PyObject* to a NumPy array. This allows the "innate type and shape"
of Python list-of-lists to be discovered without
actually converting to an array.

In some cases, such as structured arrays and the __array__ interface,
a data type needs to be used to make sense of the object.  When
this is needed, provide a Descr for 'requested_dtype', otherwise
provide NULL. This reference is not stolen. Also, if the requested
dtype doesn't modify the interpretation of the input, out_dtype will
still get the "innate" dtype of the object, not the dtype passed
in 'requested_dtype'.

If writing to the value in 'op' is desired, set the boolean
'writeable' to 1.  This raises an error when 'op' is a scalar, list
of lists, or other non-writeable 'op'.

Result: When success (0 return value) is returned, either out_arr
is filled with a non-NULL PyArrayObject and
the rest of the parameters are untouched, or out_arr is
filled with NULL, and the rest of the parameters are
filled.

Typical usage:

PyArrayObject *arr = NULL;
PyArray_Descr *dtype = NULL;
int ndim = 0;
npy_intp dims[NPY_MAXDIMS];

if (PyArray_GetArrayParamsFromObject(op, NULL, 1, &dtype,
&ndim, &dims, &arr, NULL) < 0) {
return NULL;
}
if (arr == NULL) {
... validate/change dtype, validate flags, ndim, etc ...
// Could make custom strides here too
arr = PyArray_NewFromDescr(&PyArray_Type, dtype, ndim,
dims, NULL,
is_f_order ? NPY_ARRAY_F_CONTIGUOUS : 0,
NULL);
if (arr == NULL) {
return NULL;
}
if (PyArray_CopyObject(arr, op) < 0) {
Py_DECREF(arr);
return NULL;
}
}
else {
... in this case the other parameters weren't filled, just
validate and possibly copy arr itself ...
}
... use arr ...

::

  int
  PyArray_ConvertClipmodeSequence(PyObject *object, NPY_CLIPMODE
                                  *modes, int n)

Convert an object to an array of n NPY_CLIPMODE values.
This is intended to be used in functions where a different mode
could be applied to each axis, like in ravel_multi_index.

::

  PyObject *
  PyArray_MatrixProduct2(PyObject *op1, PyObject
                         *op2, PyArrayObject*out)

Numeric.matrixproduct(a,v,out)
just like inner product but does the swapaxes stuff on the fly

::

  npy_bool
  NpyIter_IsFirstVisit(NpyIter *iter, int iop)

Checks to see whether this is the first time the elements
of the specified reduction operand which the iterator points at are
being seen for the first time. The function returns
a reasonable answer for reduction operands and when buffering is
disabled. The answer may be incorrect for buffered non-reduction
operands.

This function is intended to be used in EXTERNAL_LOOP mode only,
and will produce some wrong answers when that mode is not enabled.

If this function returns true, the caller should also
check the inner loop stride of the operand, because if
that stride is 0, then only the first element of the innermost
external loop is being visited for the first time.

WARNING: For performance reasons, 'iop' is not bounds-checked,
it is not confirmed that 'iop' is actually a reduction
operand, and it is not confirmed that EXTERNAL_LOOP
mode is enabled. These checks are the responsibility of
the caller, and should be done outside of any inner loops.

::

  int
  PyArray_SetBaseObject(PyArrayObject *arr, PyObject *obj)

Sets the 'base' attribute of the array. This steals a reference
to 'obj'.

Returns 0 on success, -1 on failure.

::

  void
  PyArray_CreateSortedStridePerm(int ndim, npy_intp
                                 *strides, npy_stride_sort_item
                                 *out_strideperm)


This function populates the first ndim elements
of strideperm with sorted descending by their absolute values.
For example, the stride array (4, -2, 12) becomes
[(2, 12), (0, 4), (1, -2)].

::

  void
  PyArray_RemoveAxesInPlace(PyArrayObject *arr, npy_bool *flags)


Removes the axes flagged as True from the array,
modifying it in place. If an axis flagged for removal
has a shape entry bigger than one, this effectively selects
index zero for that axis.

WARNING: If an axis flagged for removal has a shape equal to zero,
the array will point to invalid memory. The caller must
validate this!

For example, this can be used to remove the reduction axes
from a reduction result once its computation is complete.

::

  void
  PyArray_DebugPrint(PyArrayObject *obj)

Prints the raw data of the ndarray in a form useful for debugging
low-level C issues.

::

  int
  PyArray_FailUnlessWriteable(PyArrayObject *obj, const char *name)


This function does nothing if obj is writeable, and raises an exception
(and returns -1) if obj is not writeable. It may also do other
house-keeping, such as issuing warnings on arrays which are transitioning
to become views. Always call this function at some point before writing to
an array.

'name' is a name for the array, used to give better error
messages. Something like "assignment destination", "output array", or even
just "array".

::

  int
  PyArray_SetUpdateIfCopyBase(PyArrayObject *arr, PyArrayObject *base)


Precondition: 'arr' is a copy of 'base' (though possibly with different
strides, ordering, etc.). This function sets the UPDATEIFCOPY flag and the
->base pointer on 'arr', so that when 'arr' is destructed, it will copy any
changes back to 'base'.

Steals a reference to 'base'.

Returns 0 on success, -1 on failure.

::

  void *
  PyDataMem_NEW(size_t size)

Allocates memory for array data.

::

  void
  PyDataMem_FREE(void *ptr)

Free memory for array data.

::

  void *
  PyDataMem_RENEW(void *ptr, size_t size)

Reallocate/resize memory for array data.

::

  PyDataMem_EventHookFunc *
  PyDataMem_SetEventHook(PyDataMem_EventHookFunc *newhook, void
                         *user_data, void **old_data)

Sets the allocation event hook for numpy array data.
Takes a PyDataMem_EventHookFunc *, which has the signature:
void hook(void *old, void *new, size_t size, void *user_data).
Also takes a void *user_data, and void **old_data.

Returns a pointer to the previous hook or NULL.  If old_data is
non-NULL, the previous user_data pointer will be copied to it.

If not NULL, hook will be called at the end of each PyDataMem_NEW/FREE/RENEW:
result = PyDataMem_NEW(size)        -> (*hook)(NULL, result, size, user_data)
PyDataMem_FREE(ptr)                 -> (*hook)(ptr, NULL, 0, user_data)
result = PyDataMem_RENEW(ptr, size) -> (*hook)(ptr, result, size, user_data)

When the hook is called, the GIL will be held by the calling
thread.  The hook should be written to be reentrant, if it performs
operations that might cause new allocation events (such as the
creation/descruction numpy objects, or creating/destroying Python
objects which might cause a gc)