def _data2string(data, form):
'''Converts a data element to binary, text or base64 representation'''
if isinstance(data, basestring):
return ('"%s"' % encode_escape(data)).encode()
elif type(data) is np.ndarray:
if form == 'text' or types.numpy_data_isstring(data):
f = types.numpy_data2printer(data)
nrows, ncols = data.shape
return _TEXT_ROWSEP.join([
_TEXT_COLSEP.join([f(data[row, col]) for col in xrange(ncols)])
for row in xrange(nrows)
]).encode()
elif form == 'binary':
data_reshaped = data.reshape((data.shape[1], data.shape[0]))
r = data_reshaped.tostring()
debug(
'NIML', 'Binary encoding (len %d -> %d): [%s]' %
(data_reshaped.size, len(r), _partial_string(r, 0)))
return r
elif form == 'base64':
data_reshaped = data.reshape((data.shape[1], data.shape[0]))
r = base64.b64encode(data_reshaped.tostring())
debug('NIML', 'Encoding ok: [%s]', _partial_string(r, 0))
return r
else:
raise ValueError("illegal format %s" % format)
elif type(data) is list:
# mixed types, each column in its own container
# always use text output format, even if requested form is binary of base64
ncols = len(data)
if ncols == 0:
return "".encode()
else:
nrows = len(data[0])
# separate formatter functions for each column
# if list of strings then take first element of the list to get a string formattr
# else use the entire np array to get a numeric formatter
fs = [
types.numpy_data2printer(d[0] if type(d) is list else d)
for d in data
]
return _TEXT_ROWSEP.join([
_TEXT_COLSEP.join(
[fs[col](data[col][row]) for col in xrange(ncols)])
for row in xrange(nrows)
]).encode()
else:
raise TypeError("Unknown type %r" % type(data))
def _data2string(data, form):
'''Converts a data element to binary, text or base64 representation'''
if isinstance(data, basestring):
return ('"%s"' % encode_escape(data)).encode()
elif type(data) is np.ndarray:
if form == 'text':
f = types.numpy_data2printer(data)
nrows, ncols = data.shape
return _TEXT_ROWSEP.join([_TEXT_COLSEP.join([f(data[row, col])
for col in xrange(ncols)])
for row in xrange(nrows)]).encode()
elif form == 'binary':
data_reshaped = data.reshape((data.shape[1], data.shape[0]))
r = data_reshaped.tostring()
debug('NIML', 'Binary encoding (len %d -> %d): [%s]' %
(data_reshaped.size, len(r), _partial_string(r, 0)))
return r
elif form == 'base64':
data_reshaped = data.reshape((data.shape[1], data.shape[0]))
r = base64.b64encode(data_reshaped.tostring())
debug('NIML', 'Encoding ok: [%s]', _partial_string(r, 0))
return r
else:
raise ValueError("illegal format %s" % format)
elif type(data) is list:
# mixed types, each column in its own container
# always use text output format, even if requested form is binary of base64
ncols = len(data)
if ncols == 0:
return "".encode()
else:
nrows = len(data[0])
# separate formatter functions for each column
# if list of strings then take first element of the list to get a string formattr
# else use the entire np array to get a numeric formatter
fs = [types.numpy_data2printer(d[0] if type(d) is list else d) for d in data]
return _TEXT_ROWSEP.join([_TEXT_COLSEP.join([fs[col](data[col][row])
for col in xrange(ncols)])
for row in xrange(nrows)]).encode()
else:
raise TypeError("Unknown type %r" % type(data))
def _dset2rawniml_datarange(s):
data = s['data']
minpos = np.argmin(data, axis=0)
maxpos = np.argmax(data, axis=0)
f = types.numpy_data2printer(data) # formatter function
r = []
for i in xrange(len(minpos)):
mnpos = minpos[i]
mxpos = maxpos[i]
r.append('%s %s %d %d' % (f(data[mnpos, i]), f(data[mxpos, i]), mnpos, mxpos))
# range of data in each column
return dict(atr_name='COLMS_RANGE',
data=r)
