def numpy2ri(o):
""" Augmented conversion function, converting numpy arrays into
rpy2.rinterface-level R structures. """
if not o.dtype.isnative:
raise(ValueError("Cannot pass numpy arrays with non-native byte orders at the moment."))
# Most types map onto R arrays:
if o.dtype.kind in _kinds:
# "F" means "use column-major order"
vec = SexpVector(o.ravel("F"), _kinds[o.dtype.kind])
dim = SexpVector(o.shape, INTSXP)
#FIXME: no dimnames ?
#FIXME: optimize what is below needed/possible ? (other ways to create R arrays ?)
res = rinterface.baseenv['array'](vec, dim=dim)
# R does not support unsigned types:
elif o.dtype.kind == "u":
raise(ValueError("Cannot convert numpy array of unsigned values -- R does not have unsigned integers."))
# Array-of-PyObject is treated like a Python list:
elif o.dtype.kind == "O":
res = conversion.py2ri(list(o))
# Record arrays map onto R data frames:
elif o.dtype.kind == "V":
if o.dtype.names is None:
raise(ValueError("Nothing can be done for this numpy array type %s at the moment." % (o.dtype,)))
df_args = []
for field_name in o.dtype.names:
df_args.append((field_name,
conversion.py2ri(o[field_name])))
res = ro.baseenv["data.frame"].rcall(tuple(df_args), ro.globalenv)
# It should be impossible to get here:
else:
raise(ValueError("Unknown numpy array type."))
return res
def py2ri_pandasseries(obj):
if obj.dtype.name == 'category':
res = py2ri_categoryseries(obj)
res = FactorVector(res)
elif obj.dtype == dt_datetime64ns_type:
# time series
d = [
IntVector([x.year for x in obj]),
IntVector([x.month for x in obj]),
IntVector([x.day for x in obj]),
IntVector([x.hour for x in obj]),
IntVector([x.minute for x in obj]),
IntVector([x.second for x in obj])
]
res = ISOdatetime(*d)
#FIXME: can the POSIXct be created from the POSIXct constructor ?
# (is '<M8[ns]' mapping to Python datetime.datetime ?)
res = POSIXct(res)
else:
# converted as a numpy array
func = numpy2ri.converter.py2ri.registry[numpy.ndarray]
# current conversion as performed by numpy
res = func(obj)
if len(obj.shape) == 1:
if (obj.dtype != dt_O_type):
# force into an R vector
res = as_vector(res)
# "index" is equivalent to "names" in R
if obj.ndim == 1:
res.do_slot_assign('names',
StrVector(tuple(str(x) for x in obj.index)))
else:
res.do_slot_assign('dimnames', SexpVector(conversion.py2ri(obj.index)))
return res
def py2rpy_pandasseries(obj):
if obj.dtype.name == 'O':
warnings.warn('Element "%s" is of dtype "O" and converted '
'to R vector of strings.' % obj.name)
res = StrVector(obj)
elif obj.dtype.name == 'category':
res = py2rpy_categoryseries(obj)
res = FactorVector(res)
elif is_datetime64_any_dtype(obj.dtype):
# time series
tzname = obj.dt.tz.zone if obj.dt.tz else ''
d = [
IntVector([x.year for x in obj]),
IntVector([x.month for x in obj]),
IntVector([x.day for x in obj]),
IntVector([x.hour for x in obj]),
IntVector([x.minute for x in obj]),
FloatSexpVector([x.second + x.microsecond * 1e-6 for x in obj])
]
res = ISOdatetime(*d, tz=StrSexpVector([tzname]))
# TODO: can the POSIXct be created from the POSIXct constructor ?
# (is '<M8[ns]' mapping to Python datetime.datetime ?)
res = POSIXct(res)
elif (obj.dtype == dt_O_type):
homogeneous_type = None
for x in obj.values:
if x is None:
continue
if homogeneous_type is None:
homogeneous_type = type(x)
continue
if type(x) is not homogeneous_type:
raise ValueError('Series can only be of one type, or None.')
# TODO: Could this be merged with obj.type.name == 'O' case above ?
res = {
int: IntVector,
bool: BoolVector,
None: BoolVector,
str: StrVector,
bytes: numpy2ri.converter.py2rpy.registry[numpy.ndarray]
}[homogeneous_type](obj)
else:
# converted as a numpy array
func = numpy2ri.converter.py2rpy.registry[numpy.ndarray]
# current conversion as performed by numpy
res = func(obj)
if len(obj.shape) == 1:
if (obj.dtype != dt_O_type):
# force into an R vector
res = as_vector(res)
# "index" is equivalent to "names" in R
if obj.ndim == 1:
res.do_slot_assign('names',
StrVector(tuple(str(x) for x in obj.index)))
else:
res.do_slot_assign('dimnames',
SexpVector(conversion.py2rpy(obj.index)))
return res
def py2ri_pandasseries(obj):
if obj.dtype == '<M8[ns]':
# time series
d = [
IntVector([x.year for x in obj]),
IntVector([x.month for x in obj]),
IntVector([x.day for x in obj]),
IntVector([x.hour for x in obj]),
IntVector([x.minute for x in obj]),
IntVector([x.second for x in obj])
]
res = ISOdatetime(*d)
#FIXME: can the POSIXct be created from the POSIXct constructor ?
# (is '<M8[ns]' mapping to Python datetime.datetime ?)
res = POSIXct(res)
else:
# converted as a numpy array
res = numpy2ri.numpy2ri(obj.values)
# "index" is equivalent to "names" in R
if obj.ndim == 1:
res.do_slot_assign('names',
StrVector(tuple(str(x) for x in obj.index)))
else:
res.do_slot_assign('dimnames', SexpVector(conversion.py2ri(obj.index)))
return res
def _convert_to_r_array(x, force_2d=False):
# see <https://bitbucket.org/rpy2/rpy2/src/62be5f3e447776bcb808f48cb68cbbcd8a75d4d3/rpy/robjects/numpy2ri.py>
# line 76-82
# copy to avoid any problem.
x = np.asarray(x).copy()
if force_2d:
assert x.ndim in {1, 2}
# this is necessary for y variable. otherwise, it would fail for cv.glmnet.
if x.ndim == 1:
x = x[:, np.newaxis]
if x.dtype == np.float64:
vec = SexpVector(x.ravel("F"), REALSXP)
elif x.dtype == np.int64:
vec = SexpVector(x.ravel("F"), INTSXP)
else:
raise RuntimeError('not implemented for this dtype {}'.format(x.dtype))
dim = SexpVector(x.shape, INTSXP)
return rinterface.baseenv['array'](vec, dim=dim)
def py2rpy_pandasseries(obj):
if numpy.dtype.name == 'O':
warnings.warn('Element "%s" is of dtype "O" and converted to R vector of strings.' % obj.name)
res = StrVector(obj)
elif obj.dtype.name == 'category':
res = py2rpy_categoryseries(obj)
res = FactorVector(res)
elif is_datetime64_any_dtype(obj.dtype):
# time series
tzname = obj.dt.tz.zone if obj.dt.tz else ''
d = [IntVector([x.year for x in obj]),
IntVector([x.month for x in obj]),
IntVector([x.day for x in obj]),
IntVector([x.hour for x in obj]),
IntVector([x.minute for x in obj]),
IntVector([x.second for x in obj])]
res = ISOdatetime(*d, tz=StrSexpVector([tzname]))
#FIXME: can the POSIXct be created from the POSIXct constructor ?
# (is '<M8[ns]' mapping to Python datetime.datetime ?)
res = POSIXct(res)
else:
# converted as a numpy array
func = numpy2ri.converter.py2rpy.registry[numpy.ndarray]
# current conversion as performed by numpy
res = func(obj)
if len(obj.shape) == 1:
if (obj.dtype != dt_O_type):
# force into an R vector
res=as_vector(res)
# "index" is equivalent to "names" in R
if obj.ndim == 1:
res.do_slot_assign('names',
StrVector(tuple(str(x) for x in obj.index)))
else:
res.do_slot_assign('dimnames',
SexpVector(conversion.py2rpy(obj.index)))
return res
def py2rpy_pandasseries(obj):
if obj.dtype.name == 'O':
warnings.warn('Element "%s" is of dtype "O" and converted '
'to R vector of strings.' % obj.name)
res = StrVector(obj)
elif obj.dtype.name == 'category':
res = py2rpy_categoryseries(obj)
res = FactorVector(res)
elif is_datetime64_any_dtype(obj.dtype):
# time series
tzname = obj.dt.tz.zone if obj.dt.tz else ''
d = [IntVector([x.year for x in obj]),
IntVector([x.month for x in obj]),
IntVector([x.day for x in obj]),
IntVector([x.hour for x in obj]),
IntVector([x.minute for x in obj]),
FloatSexpVector([x.second + x.microsecond * 1e-6 for x in obj])]
res = ISOdatetime(*d, tz=StrSexpVector([tzname]))
# TODO: can the POSIXct be created from the POSIXct constructor ?
# (is '<M8[ns]' mapping to Python datetime.datetime ?)
res = POSIXct(res)
elif obj.dtype.type == str:
res = _PANDASTYPE2RPY2[str](obj)
elif obj.dtype.name in integer_array_types:
res = _PANDASTYPE2RPY2[int](obj)
if len(obj.shape) == 1:
if obj.dtype != dt_O_type:
# force into an R vector
res = as_vector(res)
elif (obj.dtype == dt_O_type):
homogeneous_type = None
for x in obj.values:
if x is None:
continue
if homogeneous_type is None:
homogeneous_type = type(x)
continue
if ((type(x) is not homogeneous_type)
and not
((isinstance(x, float) and math.isnan(x))
or pandas.isna(x))):
raise ValueError(
'Series can only be of one type, or None '
'(and here we have %s and %s). If happening with '
'a pandas DataFrame the method infer_objects() '
'will normalize data types before conversion.' %
(homogeneous_type, type(x)))
# TODO: Could this be merged with obj.type.name == 'O' case above ?
res = _PANDASTYPE2RPY2[homogeneous_type](obj)
else:
# converted as a numpy array
func = numpy2ri.converter.py2rpy.registry[numpy.ndarray]
# current conversion as performed by numpy
res = func(obj.values)
if len(obj.shape) == 1:
if (obj.dtype != dt_O_type):
# force into an R vector
res = as_vector(res)
# "index" is equivalent to "names" in R
if obj.ndim == 1:
res.do_slot_assign('names',
StrVector(tuple(str(x) for x in obj.index)))
else:
res.do_slot_assign('dimnames',
SexpVector(conversion.py2rpy(obj.index)))
return res
def simple_string_vector_repr(self):
if len(self) == 1:
return self[0]
return SexpVector.__old_repr__(self)
def simple_string_vector_repr(self):
if len(self) == 1:
return self[0]
return SexpVector.__old_repr__(self)