def write_output(self, f_out, f_in):
"""Assemble output file with collected results."""
out = csv.writer(f_out, dialect='excel', delimiter=';')
write = out.writerow
mdpair = self.get_mdpair()
mdpts = self.get_mdpair_coords()
write(['input filename', filename(f_in)])
write([])
write(['distance results'])
write(['largest distance points (indices)', str(mdpair)])
write(['coordinates of point %s' % mdpair[0], mdpts[0]])
write(['coordinates of point %s' % mdpair[1], mdpts[1]])
write(['distance', str(self.get_edm()[mdpair])])
write([])
write(['area results calculated by triangular tesselation'])
write(['longest transversal edge', self.get_longest_edge_len()])
write(['overall area', self.get_area()])
write(['perimeter', self.perimeter])
def __init__(self, xmlfile, namespace=''):
"""Create a new Imaris-XML object from a file.
Parameters
----------
xmlfile : file or str
A filehandle or string for the XML file to parse.
namespace : string, optional
A string denoting the namespace expected in the XML file,
defaults to the one used by MS Excel in its XML format.
"""
self.tree = None
self.cells = {}
# by default, we expect the namespace of Excel XML:
self.namespace = 'urn:schemas-microsoft-com:office:spreadsheet'
if namespace:
self.namespace = namespace
log.info("Parsing XML file: %s" % filename(xmlfile))
self.tree = etree.parse(xmlfile)
log.info("Done parsing XML: %s" % self.tree)
self._check_namespace()
def gen_stats(f_in, f_out, label=False, deltas=[1], thresh=0, verbosity=0):
"""Parse and process tracks and calculate statistics from the data."""
# default loglevel is 30 while 20 and 10 show more details
loglevel = (3 - verbosity) * 10
log.setLevel(loglevel)
log.warn("Infile: %s" % f_in)
log.debug("Outfile: %s" % f_out)
log.info("Stepping width(s): %s" % deltas)
log.info("Angle threshold: %s" % thresh)
ppr = pprint.PrettyPrinter(indent=4)
######### tracks parsing #########
# TODO: parsing can be done in a nicer way be reading the header lines via
# csvreader.next(), checking for the expected values and the number of
# tracks and then directly reading the trackpoints into a numpy ndarray...
mtrack2_file = filehandle(f_in, "r")
csvreader = csv.reader(mtrack2_file, delimiter="\t")
# parse all lines into memory
# NOTE: this is bad if the files get too large, but we haven't seen result
# files from MTrack2 that are bigger than a couple of MB.
data = []
for row in csvreader:
data.append([parse_cell(x) for x in row])
# data.append(row)
# start parsing the header
header = []
header.append(data.pop(0))
header.append(data.pop(0))
if not header[0][0] == "Frame":
# exit because file is broken...
raise SystemExit("Unable to find correct header, stopping.")
log.debug("Header:\n%s\n" % ppr.pformat(header))
# second line is 'Tracks 1 to N', so we can read the total number there:
trackmax = int(header[1][0].split(" ")[3])
log.info("Total number of tracks: %s" % ppr.pformat(trackmax))
# last N lines are the stats per track
trackstats = []
while True:
# pop returns the last element if no index is given
cur = data.pop()
if cur[0] == "Track":
# remove one more line (empty), then we're done
cur = data.pop()
break
else:
trackstats.append(cur)
# as we parsed from the last element, we need to reverse the list
trackstats.reverse()
log.warn("Track statistics:\n%s" % ppr.pformat(trackstats))
# this code can help debugging problematic files:
# for row in data:
# try:
# np.array(row, dtype='float')
# except ValueError:
# raise SystemExit(row)
# create the ndarray from the remaining data while removing column 0
# (indices), and every subsequent third column (flags)
todelete = range(0, (trackmax + 1) * 3, 3)
npdata = np.delete(data, todelete, axis=1)
npdata_bool = npdata > 0
######### tracks processing (combining etc.) #########
tracklen = [0] * trackmax
t_overlap = npdata_bool[:, 0]
for track in range(trackmax):
tracklen[track] = sum(npdata_bool[:, track * 2])
t_overlap = t_overlap * npdata_bool[:, track * 2]
if trackmax > 1 and sum(t_overlap) > 0:
raise SystemExit("*** WARNING: Found overlapping tracks! ***")
t_combined = np.zeros((npdata.shape[0], 2))
for track in range(trackmax):
t_combined += npdata[:, track * 2 : (track + 1) * 2]
comb_mask = np.zeros(t_combined.shape[0])
for i, row in enumerate(t_combined):
if (row == [0.0, 0.0]).all():
# print 'row %i is zerooooo' % i
comb_mask[i] = True
t_combined = np.ma.compress_rows(np.ma.array(t_combined, mask=np.repeat(comb_mask, 2)))
######### calculations #########
mov_v = {}
mov_n = {}
rot = {}
rot_t = {}
outdata = t_combined
if label:
label = "pos_x\tpos_y"
for step in deltas:
# calculate movement vectors (mov_v):
mov_v[step] = movement_vectors(t_combined, step)
# calculate vector normals (mov_n):
mov_n[step] = np.zeros((mov_v[step].shape[0], 1))
for pos in range(1, mov_n[step].shape[0]):
mov_n[step][pos] = np.linalg.norm(mov_v[step][pos])
# calculate rotation:
rot[step] = calc_rotation(mov_v[step], mov_n[step], step)
# for the movement vectors all values need to be written to the output,
# but it is not necessary to repeat them for every stepping, so they
# are only added for stepping '1':
if step == 1:
outdata = np.hstack((outdata, mov_v[1]))
if label:
label += "\tdelta_x\tdelta_y"
outdata = np.hstack((outdata, mov_n[step], rot[step]))
# threshold rotation angles:
if thresh > 0:
rot_t[step] = np.where(abs(rot[step]) > thresh, rot[step], 0)
outdata = np.hstack((outdata, rot_t[step]))
if label:
label += "\tdistance_%s\tangle_%s" % (step, step)
if thresh > 0:
label += "\tthresholded_angle_%s" % step
if label:
log.info("label: %s" % label)
_save_results(f_out, outdata, label)
log.warn("Wrote results to '%s'" % filename(f_out))
