void Fn::NumpyInput::reset()
{
_io = std::make_shared<IO_double>(_dim,_inputType);
// setup numpy iterator
if (_iter) {
NpyIter_Deallocate(_iter);
_iter = NULL;
}
PyArray_Descr* dtype = PyArray_DescrFromType(NPY_DOUBLE);
_iter = NpyIter_New( (PyArrayObject*)_arr, NPY_ITER_READONLY|
NPY_ITER_EXTERNAL_LOOP|
NPY_ITER_REFS_OK,
NPY_KEEPORDER, NPY_UNSAFE_CASTING, dtype );
Py_DECREF(dtype);
if (_iter == NULL) {
throw std::invalid_argument("An unknown error was encountered. Please contact developers.");
}
_iternext = NpyIter_GetIterNext(_iter, NULL);
if (_iternext == NULL) {
NpyIter_Deallocate(_iter);
throw std::invalid_argument("An unknown error was encountered. Please contact developers.");
}
/* The location of the data pointer which the iterator may update */
_dataptr = NpyIter_GetDataPtrArray(_iter);
/* The location of the stride which the iterator may update */
_strideptr = NpyIter_GetInnerStrideArray(_iter);
/* The location of the inner loop size which the iterator may update */
_innersizeptr = NpyIter_GetInnerLoopSizePtr(_iter);
_count = *_innersizeptr;
_data = *_dataptr;
grab();
}
/*
* This function executes all the standard NumPy reduction function
* boilerplate code, just calling assign_identity and the appropriate
* inner loop function where necessary.
*
* operand : The array to be reduced.
* out : NULL, or the array into which to place the result.
* wheremask : NOT YET SUPPORTED, but this parameter is placed here
* so that support can be added in the future without breaking
* API compatibility. Pass in NULL.
* operand_dtype : The dtype the inner loop expects for the operand.
* result_dtype : The dtype the inner loop expects for the result.
* casting : The casting rule to apply to the operands.
* axis_flags : Flags indicating the reduction axes of 'operand'.
* reorderable : If True, the reduction being done is reorderable, which
* means specifying multiple axes of reduction at once is ok,
* and the reduction code may calculate the reduction in an
* arbitrary order. The calculation may be reordered because
* of cache behavior or multithreading requirements.
* keepdims : If true, leaves the reduction dimensions in the result
* with size one.
* subok : If true, the result uses the subclass of operand, otherwise
* it is always a base class ndarray.
* assign_identity : If NULL, PyArray_InitializeReduceResult is used, otherwise
* this function is called to initialize the result to
* the reduction's unit.
* loop : The loop which does the reduction.
* data : Data which is passed to assign_identity and the inner loop.
* buffersize : Buffer size for the iterator. For the default, pass in 0.
* funcname : The name of the reduction function, for error messages.
* errormask : forwarded from _get_bufsize_errmask
*
* TODO FIXME: if you squint, this is essentially an second independent
* implementation of generalized ufuncs with signature (i)->(), plus a few
* extra bells and whistles. (Indeed, as far as I can tell, it was originally
* split out to support a fancy version of count_nonzero... which is not
* actually a reduction function at all, it's just a (i)->() function!) So
* probably these two implementation should be merged into one. (In fact it
* would be quite nice to support axis= and keepdims etc. for arbitrary
* generalized ufuncs!)
*/
NPY_NO_EXPORT PyArrayObject *
PyUFunc_ReduceWrapper(PyArrayObject *operand, PyArrayObject *out,
PyArrayObject *wheremask,
PyArray_Descr *operand_dtype,
PyArray_Descr *result_dtype,
NPY_CASTING casting,
npy_bool *axis_flags, int reorderable,
int keepdims,
int subok,
PyArray_AssignReduceIdentityFunc *assign_identity,
PyArray_ReduceLoopFunc *loop,
void *data, npy_intp buffersize, const char *funcname,
int errormask)
{
PyArrayObject *result = NULL, *op_view = NULL;
npy_intp skip_first_count = 0;
/* Iterator parameters */
NpyIter *iter = NULL;
PyArrayObject *op[2];
PyArray_Descr *op_dtypes[2];
npy_uint32 flags, op_flags[2];
/* Validate that the parameters for future expansion are NULL */
if (wheremask != NULL) {
PyErr_SetString(PyExc_RuntimeError,
"Reduce operations in NumPy do not yet support "
"a where mask");
return NULL;
}
/*
* This either conforms 'out' to the ndim of 'operand', or allocates
* a new array appropriate for this reduction.
*
* A new array with WRITEBACKIFCOPY is allocated if operand and out have memory
* overlap.
*/
Py_INCREF(result_dtype);
result = PyArray_CreateReduceResult(operand, out,
result_dtype, axis_flags,
keepdims, subok, funcname);
if (result == NULL) {
goto fail;
}
/*
* Initialize the result to the reduction unit if possible,
* otherwise copy the initial values and get a view to the rest.
*/
if (assign_identity != NULL) {
/*
* If this reduction is non-reorderable, make sure there are
* only 0 or 1 axes in axis_flags.
*/
if (!reorderable && check_nonreorderable_axes(PyArray_NDIM(operand),
axis_flags, funcname) < 0) {
goto fail;
}
if (assign_identity(result, data) < 0) {
goto fail;
}
op_view = operand;
Py_INCREF(op_view);
}
else {
op_view = PyArray_InitializeReduceResult(result, operand,
axis_flags, reorderable,
&skip_first_count, funcname);
if (op_view == NULL) {
goto fail;
}
/* empty op_view signals no reduction; but 0-d arrays cannot be empty */
if ((PyArray_SIZE(op_view) == 0) || (PyArray_NDIM(operand) == 0)) {
Py_DECREF(op_view);
op_view = NULL;
goto finish;
}
}
/* Set up the iterator */
op[0] = result;
op[1] = op_view;
op_dtypes[0] = result_dtype;
op_dtypes[1] = operand_dtype;
flags = NPY_ITER_BUFFERED |
NPY_ITER_EXTERNAL_LOOP |
NPY_ITER_GROWINNER |
NPY_ITER_DONT_NEGATE_STRIDES |
NPY_ITER_ZEROSIZE_OK |
NPY_ITER_REDUCE_OK |
NPY_ITER_REFS_OK;
op_flags[0] = NPY_ITER_READWRITE |
NPY_ITER_ALIGNED |
NPY_ITER_NO_SUBTYPE;
op_flags[1] = NPY_ITER_READONLY |
NPY_ITER_ALIGNED;
iter = NpyIter_AdvancedNew(2, op, flags,
NPY_KEEPORDER, casting,
op_flags,
op_dtypes,
-1, NULL, NULL, buffersize);
if (iter == NULL) {
goto fail;
}
/* Start with the floating-point exception flags cleared */
PyUFunc_clearfperr();
if (NpyIter_GetIterSize(iter) != 0) {
NpyIter_IterNextFunc *iternext;
char **dataptr;
npy_intp *strideptr;
npy_intp *countptr;
int needs_api;
iternext = NpyIter_GetIterNext(iter, NULL);
if (iternext == NULL) {
goto fail;
}
dataptr = NpyIter_GetDataPtrArray(iter);
strideptr = NpyIter_GetInnerStrideArray(iter);
countptr = NpyIter_GetInnerLoopSizePtr(iter);
needs_api = NpyIter_IterationNeedsAPI(iter);
/* Straightforward reduction */
if (loop == NULL) {
PyErr_Format(PyExc_RuntimeError,
"reduction operation %s did not supply an "
"inner loop function", funcname);
goto fail;
}
if (loop(iter, dataptr, strideptr, countptr,
iternext, needs_api, skip_first_count, data) < 0) {
goto fail;
}
}
/* Check whether any errors occurred during the loop */
if (PyErr_Occurred() ||
_check_ufunc_fperr(errormask, NULL, "reduce") < 0) {
goto fail;
}
NpyIter_Deallocate(iter);
Py_DECREF(op_view);
finish:
/* Strip out the extra 'one' dimensions in the result */
if (out == NULL) {
if (!keepdims) {
PyArray_RemoveAxesInPlace(result, axis_flags);
}
}
else {
PyArray_ResolveWritebackIfCopy(result); /* prevent spurious warnings */
Py_DECREF(result);
result = out;
Py_INCREF(result);
}
return result;
fail:
PyArray_ResolveWritebackIfCopy(result); /* prevent spurious warnings */
Py_XDECREF(result);
Py_XDECREF(op_view);
if (iter != NULL) {
NpyIter_Deallocate(iter);
}
return NULL;
}
/* unravel_index implementation - see add_newdocs.py */
NPY_NO_EXPORT PyObject *
arr_unravel_index(PyObject *self, PyObject *args, PyObject *kwds)
{
PyObject *indices0 = NULL, *ret_tuple = NULL;
PyArrayObject *ret_arr = NULL;
PyArrayObject *indices = NULL;
PyArray_Descr *dtype = NULL;
PyArray_Dims dimensions={0,0};
NPY_ORDER order = NPY_CORDER;
npy_intp unravel_size;
NpyIter *iter = NULL;
int i, ret_ndim;
npy_intp ret_dims[NPY_MAXDIMS], ret_strides[NPY_MAXDIMS];
char *kwlist[] = {"indices", "dims", "order", NULL};
if (!PyArg_ParseTupleAndKeywords(args, kwds, "OO&|O&:unravel_index",
kwlist,
&indices0,
PyArray_IntpConverter, &dimensions,
PyArray_OrderConverter, &order)) {
goto fail;
}
if (dimensions.len == 0) {
PyErr_SetString(PyExc_ValueError,
"dims must have at least one value");
goto fail;
}
unravel_size = PyArray_MultiplyList(dimensions.ptr, dimensions.len);
if (!PyArray_Check(indices0)) {
indices = (PyArrayObject*)PyArray_FromAny(indices0,
NULL, 0, 0, 0, NULL);
if (indices == NULL) {
goto fail;
}
}
else {
indices = (PyArrayObject *)indices0;
Py_INCREF(indices);
}
dtype = PyArray_DescrFromType(NPY_INTP);
if (dtype == NULL) {
goto fail;
}
iter = NpyIter_New(indices, NPY_ITER_READONLY|
NPY_ITER_ALIGNED|
NPY_ITER_BUFFERED|
NPY_ITER_ZEROSIZE_OK|
NPY_ITER_DONT_NEGATE_STRIDES|
NPY_ITER_MULTI_INDEX,
NPY_KEEPORDER, NPY_SAME_KIND_CASTING,
dtype);
if (iter == NULL) {
goto fail;
}
/*
* Create the return array with a layout compatible with the indices
* and with a dimension added to the end for the multi-index
*/
ret_ndim = PyArray_NDIM(indices) + 1;
if (NpyIter_GetShape(iter, ret_dims) != NPY_SUCCEED) {
goto fail;
}
ret_dims[ret_ndim-1] = dimensions.len;
if (NpyIter_CreateCompatibleStrides(iter,
dimensions.len*sizeof(npy_intp), ret_strides) != NPY_SUCCEED) {
goto fail;
}
ret_strides[ret_ndim-1] = sizeof(npy_intp);
/* Remove the multi-index and inner loop */
if (NpyIter_RemoveMultiIndex(iter) != NPY_SUCCEED) {
goto fail;
}
if (NpyIter_EnableExternalLoop(iter) != NPY_SUCCEED) {
goto fail;
}
ret_arr = (PyArrayObject *)PyArray_NewFromDescr(&PyArray_Type, dtype,
ret_ndim, ret_dims, ret_strides, NULL, 0, NULL);
dtype = NULL;
if (ret_arr == NULL) {
goto fail;
}
if (order == NPY_CORDER) {
if (NpyIter_GetIterSize(iter) != 0) {
NpyIter_IterNextFunc *iternext;
char **dataptr;
npy_intp *strides;
npy_intp *countptr, count;
npy_intp *coordsptr = (npy_intp *)PyArray_DATA(ret_arr);
iternext = NpyIter_GetIterNext(iter, NULL);
if (iternext == NULL) {
goto fail;
}
dataptr = NpyIter_GetDataPtrArray(iter);
strides = NpyIter_GetInnerStrideArray(iter);
countptr = NpyIter_GetInnerLoopSizePtr(iter);
do {
count = *countptr;
if (unravel_index_loop_corder(dimensions.len, dimensions.ptr,
unravel_size, count, *dataptr, *strides,
coordsptr) != NPY_SUCCEED) {
goto fail;
}
coordsptr += count*dimensions.len;
} while(iternext(iter));
}
}
else if (order == NPY_FORTRANORDER) {
if (NpyIter_GetIterSize(iter) != 0) {
NpyIter_IterNextFunc *iternext;
char **dataptr;
npy_intp *strides;
npy_intp *countptr, count;
npy_intp *coordsptr = (npy_intp *)PyArray_DATA(ret_arr);
iternext = NpyIter_GetIterNext(iter, NULL);
if (iternext == NULL) {
goto fail;
}
dataptr = NpyIter_GetDataPtrArray(iter);
strides = NpyIter_GetInnerStrideArray(iter);
countptr = NpyIter_GetInnerLoopSizePtr(iter);
do {
count = *countptr;
if (unravel_index_loop_forder(dimensions.len, dimensions.ptr,
unravel_size, count, *dataptr, *strides,
coordsptr) != NPY_SUCCEED) {
goto fail;
}
coordsptr += count*dimensions.len;
} while(iternext(iter));
}
}
else {
PyErr_SetString(PyExc_ValueError,
"only 'C' or 'F' order is permitted");
goto fail;
}
/* Now make a tuple of views, one per index */
ret_tuple = PyTuple_New(dimensions.len);
if (ret_tuple == NULL) {
goto fail;
}
for (i = 0; i < dimensions.len; ++i) {
PyArrayObject *view;
view = (PyArrayObject *)PyArray_New(&PyArray_Type, ret_ndim-1,
ret_dims, NPY_INTP,
ret_strides,
PyArray_BYTES(ret_arr) + i*sizeof(npy_intp),
0, NPY_ARRAY_WRITEABLE, NULL);
if (view == NULL) {
goto fail;
}
Py_INCREF(ret_arr);
if (PyArray_SetBaseObject(view, (PyObject *)ret_arr) < 0) {
Py_DECREF(view);
goto fail;
}
PyTuple_SET_ITEM(ret_tuple, i, PyArray_Return(view));
}
Py_DECREF(ret_arr);
Py_XDECREF(indices);
PyDimMem_FREE(dimensions.ptr);
NpyIter_Deallocate(iter);
return ret_tuple;
fail:
Py_XDECREF(ret_tuple);
Py_XDECREF(ret_arr);
Py_XDECREF(dtype);
Py_XDECREF(indices);
PyDimMem_FREE(dimensions.ptr);
NpyIter_Deallocate(iter);
return NULL;
}
/* ravel_multi_index implementation - see add_newdocs.py */
NPY_NO_EXPORT PyObject *
arr_ravel_multi_index(PyObject *self, PyObject *args, PyObject *kwds)
{
int i, s;
PyObject *mode0=NULL, *coords0=NULL;
PyArrayObject *ret = NULL;
PyArray_Dims dimensions={0,0};
npy_intp ravel_strides[NPY_MAXDIMS];
NPY_ORDER order = NPY_CORDER;
NPY_CLIPMODE modes[NPY_MAXDIMS];
PyArrayObject *op[NPY_MAXARGS];
PyArray_Descr *dtype[NPY_MAXARGS];
npy_uint32 op_flags[NPY_MAXARGS];
NpyIter *iter = NULL;
char *kwlist[] = {"multi_index", "dims", "mode", "order", NULL};
memset(op, 0, sizeof(op));
dtype[0] = NULL;
if (!PyArg_ParseTupleAndKeywords(args, kwds,
"OO&|OO&:ravel_multi_index", kwlist,
&coords0,
PyArray_IntpConverter, &dimensions,
&mode0,
PyArray_OrderConverter, &order)) {
goto fail;
}
if (dimensions.len+1 > NPY_MAXARGS) {
PyErr_SetString(PyExc_ValueError,
"too many dimensions passed to ravel_multi_index");
goto fail;
}
if (!PyArray_ConvertClipmodeSequence(mode0, modes, dimensions.len)) {
goto fail;
}
switch (order) {
case NPY_CORDER:
s = 1;
for (i = dimensions.len-1; i >= 0; --i) {
ravel_strides[i] = s;
s *= dimensions.ptr[i];
}
break;
case NPY_FORTRANORDER:
s = 1;
for (i = 0; i < dimensions.len; ++i) {
ravel_strides[i] = s;
s *= dimensions.ptr[i];
}
break;
default:
PyErr_SetString(PyExc_ValueError,
"only 'C' or 'F' order is permitted");
goto fail;
}
/* Get the multi_index into op */
if (sequence_to_arrays(coords0, op, dimensions.len, "multi_index") < 0) {
goto fail;
}
for (i = 0; i < dimensions.len; ++i) {
op_flags[i] = NPY_ITER_READONLY|
NPY_ITER_ALIGNED;
}
op_flags[dimensions.len] = NPY_ITER_WRITEONLY|
NPY_ITER_ALIGNED|
NPY_ITER_ALLOCATE;
dtype[0] = PyArray_DescrFromType(NPY_INTP);
for (i = 1; i <= dimensions.len; ++i) {
dtype[i] = dtype[0];
}
iter = NpyIter_MultiNew(dimensions.len+1, op, NPY_ITER_BUFFERED|
NPY_ITER_EXTERNAL_LOOP|
NPY_ITER_ZEROSIZE_OK,
NPY_KEEPORDER,
NPY_SAME_KIND_CASTING,
op_flags, dtype);
if (iter == NULL) {
goto fail;
}
if (NpyIter_GetIterSize(iter) != 0) {
NpyIter_IterNextFunc *iternext;
char **dataptr;
npy_intp *strides;
npy_intp *countptr;
iternext = NpyIter_GetIterNext(iter, NULL);
if (iternext == NULL) {
goto fail;
}
dataptr = NpyIter_GetDataPtrArray(iter);
strides = NpyIter_GetInnerStrideArray(iter);
countptr = NpyIter_GetInnerLoopSizePtr(iter);
do {
if (ravel_multi_index_loop(dimensions.len, dimensions.ptr,
ravel_strides, *countptr, modes,
dataptr, strides) != NPY_SUCCEED) {
goto fail;
}
} while(iternext(iter));
}
ret = NpyIter_GetOperandArray(iter)[dimensions.len];
Py_INCREF(ret);
Py_DECREF(dtype[0]);
for (i = 0; i < dimensions.len; ++i) {
Py_XDECREF(op[i]);
}
PyDimMem_FREE(dimensions.ptr);
NpyIter_Deallocate(iter);
return PyArray_Return(ret);
fail:
Py_XDECREF(dtype[0]);
for (i = 0; i < dimensions.len; ++i) {
Py_XDECREF(op[i]);
}
PyDimMem_FREE(dimensions.ptr);
NpyIter_Deallocate(iter);
return NULL;
}
/*
* Returns a boolean array with True for input dates which are valid
* business days, and False for dates which are not. This is the
* low-level function which requires already cleaned input data.
*
* dates: An array of dates with 'datetime64[D]' data type.
* out: Either NULL, or an array with 'bool' data type
* in which to place the resulting dates.
* weekmask: A 7-element boolean mask, 1 for possible business days and 0
* for non-business days.
* busdays_in_weekmask: A count of how many 1's there are in weekmask.
* holidays_begin/holidays_end: A sorted list of dates matching '[D]'
* unit metadata, with any dates falling on a day of the
* week without weekmask[i] == 1 already filtered out.
*/
NPY_NO_EXPORT PyArrayObject *
is_business_day(PyArrayObject *dates, PyArrayObject *out,
npy_bool *weekmask, int busdays_in_weekmask,
npy_datetime *holidays_begin, npy_datetime *holidays_end)
{
PyArray_DatetimeMetaData temp_meta;
PyArray_Descr *dtypes[2] = {NULL, NULL};
NpyIter *iter = NULL;
PyArrayObject *op[2] = {NULL, NULL};
npy_uint32 op_flags[2], flags;
PyArrayObject *ret = NULL;
if (busdays_in_weekmask == 0) {
PyErr_SetString(PyExc_ValueError,
"the business day weekmask must have at least one "
"valid business day");
return NULL;
}
/* First create the data types for the dates and the bool output */
temp_meta.base = NPY_FR_D;
temp_meta.num = 1;
dtypes[0] = create_datetime_dtype(NPY_DATETIME, &temp_meta);
if (dtypes[0] == NULL) {
goto fail;
}
dtypes[1] = PyArray_DescrFromType(NPY_BOOL);
if (dtypes[1] == NULL) {
goto fail;
}
/* Set up the iterator parameters */
flags = NPY_ITER_EXTERNAL_LOOP|
NPY_ITER_BUFFERED|
NPY_ITER_ZEROSIZE_OK;
op[0] = dates;
op_flags[0] = NPY_ITER_READONLY | NPY_ITER_ALIGNED;
op[1] = out;
op_flags[1] = NPY_ITER_WRITEONLY | NPY_ITER_ALLOCATE | NPY_ITER_ALIGNED;
/* Allocate the iterator */
iter = NpyIter_MultiNew(2, op, flags, NPY_KEEPORDER, NPY_SAFE_CASTING,
op_flags, dtypes);
if (iter == NULL) {
goto fail;
}
/* Loop over all elements */
if (NpyIter_GetIterSize(iter) > 0) {
NpyIter_IterNextFunc *iternext;
char **dataptr;
npy_intp *strideptr, *innersizeptr;
iternext = NpyIter_GetIterNext(iter, NULL);
if (iternext == NULL) {
goto fail;
}
dataptr = NpyIter_GetDataPtrArray(iter);
strideptr = NpyIter_GetInnerStrideArray(iter);
innersizeptr = NpyIter_GetInnerLoopSizePtr(iter);
do {
char *data_dates = dataptr[0];
char *data_out = dataptr[1];
npy_intp stride_dates = strideptr[0];
npy_intp stride_out = strideptr[1];
npy_intp count = *innersizeptr;
npy_datetime date;
int day_of_week;
while (count--) {
/* Check if it's a business day */
date = *(npy_datetime *)data_dates;
day_of_week = get_day_of_week(date);
*(npy_bool *)data_out = weekmask[day_of_week] &&
!is_holiday(date,
holidays_begin, holidays_end) &&
date != NPY_DATETIME_NAT;
data_dates += stride_dates;
data_out += stride_out;
}
} while (iternext(iter));
}
/* Get the return object from the iterator */
ret = NpyIter_GetOperandArray(iter)[1];
Py_INCREF(ret);
goto finish;
fail:
Py_XDECREF(ret);
ret = NULL;
finish:
Py_XDECREF(dtypes[0]);
Py_XDECREF(dtypes[1]);
if (iter != NULL) {
if (NpyIter_Deallocate(iter) != NPY_SUCCEED) {
Py_XDECREF(ret);
ret = NULL;
}
}
return ret;
}
