def data(self):
delim = ','
try: # TODO: better, smarter error checking than this
with open(self.filename, 'r') as f:
lns = f.readlines()
ions = [float(i) for i in lns[0].split(delim)[1:]]
data = np.array(
[np.fromstring(ln, sep=delim) for ln in lns[1:]])
return Chromatogram(data[:, 1:], data[:, 0], ions)
except:
return Chromatogram()
def data(self):
fidx = open(self.filename, 'rb')
fdat = open(op.splitext(self.filename)[0] + '.DAT', 'rb')
ions = set([])
while True:
fdat.seek(fdat.tell() + 2)
try:
i = struct.unpack('<H', fdat.read(2))[0]
except struct.error:
break
ions.add(i)
ions = sorted(ions)
data = []
tme = []
while True:
try:
idx = struct.unpack('<IHHffhhh', fidx.read(22))
tme.append(idx[4])
fdat.seek(idx[0])
new_line = np.zeros(len(ions))
d = struct.unpack('<' + int(idx[1] / 4) * 'HH',
fdat.read(idx[1]))
for i, v in zip(d[1::2], d[0::2]):
new_line[ions.index(i)] = v
data.append(new_line)
except struct.error:
break
fdat.close()
fidx.close()
return Chromatogram(np.array(data), np.array(tme), ions)
def read_amdis_list(db, filename):
def get_val(line, cols, key):
return line.split('\t')[cols.index(key)]
cmp_lvl = 2 # number of directory levels to compare
# TODO: does this work for agilent files only?
mapping = defaultdict(list)
with open(filename, 'r') as f:
cols = f.readline().split('\t')
for line in f:
filename = get_val(line, cols, 'FileName')
fn = op.splitext('/'.join(filename.split('\\')[-cmp_lvl:]))[0]
# find if filtered filename overlaps with anything in the db
for dt in db.children_of_type('file'):
if fn in '/'.join(dt.rawdata.split(op.sep)):
break
else:
continue
info = {}
info['name'] = get_val(line, cols, 'Name')
info['p-s-time'] = get_val(line, cols, 'RT')
info['p-s-area'] = get_val(line, cols, 'Area')
ts = Chromatogram(np.array([np.nan]), np.array([np.nan]), [''])
mapping[dt] += [Peak(info, ts)]
with db:
for dt in mapping:
dt.children += mapping[dt]
def data(self):
fhead = open(self.filename[:-3] + '.sd', 'rb')
fdata = open(self.filename[:-3] + '.sp', 'rb')
fhead.seek(0x50)
nscans = struct.unpack('Q', fhead.read(8))[0]
fhead.seek(0xA4)
ions = []
for scn in range(nscans):
t = struct.unpack('<IdddIQIIdddd', fhead.read(80))
npts = t[7]
if scn == 0:
# TODO: this will fail if the wavelengths collected
# change through the run.
data = np.zeros((nscans, npts))
times = np.zeros((nscans))
ions = [t[8] + x * t[3] for x in range(npts)]
times[scn] = t[1]
fdata.seek(t[5] + 16)
# TODO: use np.fromfile?
data[scn] = struct.unpack('<' + npts * 'd', fdata.read(npts * 8))
return Chromatogram(data, times, ions)
fhead.close()
fdata.close()
def data(self):
print(self.filename)
# Because the spectra are stored in several files in the same
# directory, we need to loop through them and return them together.
ions = []
dtraces = []
foldname = os.path.dirname(self.filename)
# if foldname == '': foldname = os.curdir
for i in [os.path.join(foldname, i) for i in
os.listdir(foldname)]:
if i[-3:].upper() == '.CH':
wv, dtrace = self._read_ind_file(i)
if wv is None:
continue
# generate the time points if this is the first trace
if len(ions) == 0:
f = open(i, 'rb')
f.seek(0x244)
yunits = f.read(struct.unpack('>B', f.read(1))[0]).decode()
f.seek(0x11A)
st_t = struct.unpack('>i', f.read(4))[0] / 60000.
en_t = struct.unpack('>i', f.read(4))[0] / 60000.
f.close()
times = np.linspace(st_t, en_t, len(dtrace))
# add the wavelength and trace into the appropriate places
ions.append(float(wv))
dtraces.append(dtrace)
data = np.array(dtraces).transpose()
return Chromatogram(data, times, ions, yunits=yunits)
def data(self):
f = open(self.filename, 'rb')
f.seek(19)
while True:
f.seek(f.tell() - 19)
if f.read(19) == b'CRawDataScanStorage':
break
if f.read(1) == b'':
f.close()
return
f.seek(f.tell() + 62)
nscans = struct.unpack('H', f.read(2))[0]
# TODO: this shouldn't be hardcoded
ions = [44, 45, 46]
ni = len(ions)
f.seek(f.tell() + 35)
data = np.array([struct.unpack('<f' + ni * 'd', f.read(4 + ni * 8))
for _ in range(nscans)])
data[:, 0] /= 60. # convert time to minutes
# self.data = TimeSeries(data[:, 1:], data[:, 0], ions)
f.close()
return Chromatogram(data[:, 1:], data[:, 0], ions)
def data(self):
# convenience function for reading in data
def rd(f, st):
return struct.unpack(st, f.read(struct.calcsize(st)))
# open the file
f = open(self.filename, 'rb')
nscans = rd(f, 'ii')[1]
if nscans == 0:
self.data = Trace(np.array([]), np.array([]), [])
return
times = np.array(rd(f, nscans * 'd')) / 60.0
f.seek(f.tell() + 4) # number of scans again
# set up the array of column indices
indptr = np.empty(nscans + 1, dtype=int)
indptr[0] = 0
# figure out the total number of points
dpos = f.tell()
tot_pts = 0
for scn in range(nscans):
npts = rd(f, 'i')[0]
# rd(f, npts * 'f' + 'i' + n_pts * 'f')
f.seek(f.tell() + 8 * npts + 4)
tot_pts += npts
indptr[scn + 1] = tot_pts
f.seek(dpos)
ions = []
i_lkup = {}
idxs = np.empty(tot_pts, dtype=int)
vals = np.empty(tot_pts, dtype=float)
for scn in range(nscans):
npts = rd(f, 'i')[0]
rd_ions = rd(f, npts * 'f')
f.seek(f.tell() + 4) # number of points again
abun = rd(f, npts * 'f')
nions = set([int(i) for i in rd_ions if int(i) not in i_lkup])
i_lkup.update(
dict((ion, i + len(ions) - 1) for i, ion in enumerate(nions)))
ions += nions
idxs[indptr[scn]:indptr[scn + 1]] = \
[i_lkup[int(i)] for i in rd_ions]
vals[indptr[scn]:indptr[scn + 1]] = \
abun
idxs += 1
data = scipy.sparse.csr_matrix((vals, idxs, indptr),
shape=(nscans, len(ions)),
dtype=float)
return Chromatogram(data, times, ions)
def data(self):
# TODO: handle skip mass ranges
with open(self.filename, 'rb') as f:
# read in the time segments/mz ranges for the run
# read in the data itself
doff = find_offset(f, 4 * b'\xff' + 'HapsScan'.encode('ascii'))
if doff is None:
return
f.seek(doff - 20)
data_end = doff + struct.unpack('<I', f.read(4))[0] + 55
f.seek(doff + 56)
times, abns, mzs = [], [], []
cur_seg = None
mz_reader = self._ions(f)
while f.tell() <= data_end:
# record info looks like a standard format
n, t, _, recs, _, seg = struct.unpack('<IiHHHH', f.read(16))
if cur_seg != seg:
# if we've switched segments, update the list of mzs
try:
cur_mzs = next(mz_reader)
except StopIteration:
break
mzs += set(cur_mzs).difference(mzs)
mz_idx = [mzs.index(i) for i in cur_mzs]
cur_seg = seg
# just add the new time in
times.append(t)
# read the list of abundances
cur_abns = struct.unpack('<' + 'f' * recs, f.read(4 * recs))
# convert this list into an array that matches up with
# whatever mzs we currently have
empty_row = np.zeros(len(mzs))
empty_row[mz_idx] = cur_abns
# add that row into the list
abns.append(empty_row)
# convert the time from milliseconds to minutes
times = np.array(times, dtype=float) / 60000
# create the data array and populate it
data = np.zeros((len(times), len(mzs)))
for i, r in enumerate(abns):
data[i, 0:len(r)] = r
return Chromatogram(data, times, mzs)
def data(self):
# TODO: the chromatograms this generates are not exactly the
# same as the ones in the *.CH files. Maybe they need to be 0'd?
f = open(self.filename, 'rb')
f.seek(0x146)
yunits = f.read(struct.unpack('>B',
f.read(1))[0]).decode('ascii').strip()
f.seek(0x116)
nscans = struct.unpack('>i', f.read(4))[0]
times = np.zeros(nscans)
data = nscans * [{}]
ions = []
npos = 0x202
for i in range(nscans):
f.seek(npos)
npos += struct.unpack('<H', f.read(2))[0]
times[i] = struct.unpack('<L', f.read(4))[0] / 60000.
nm_srt = struct.unpack('<H', f.read(2))[0] / 20.
nm_end = struct.unpack('<H', f.read(2))[0] / 20.
nm_stp = struct.unpack('<H', f.read(2))[0] / 20.
f.read(8)
s = {}
v = struct.unpack('<h', f.read(2))[0] / 2000.
s[nm_srt] = v
for wv in np.arange(nm_srt, nm_end, nm_stp):
ov = struct.unpack('<h', f.read(2))[0]
if ov == -32768:
v = struct.unpack('<i', f.read(4))[0] / 2000.
else:
v += ov / 2000.
s[wv] = v
if wv not in ions:
ions.append(wv)
data[i] = s
ndata = np.zeros((nscans, len(ions)))
for i, d in zip(range(nscans), data):
for ion, abn in d.items():
ndata[i, ions.index(ion)] = abn
return Chromatogram(ndata, times, ions, yunits=yunits)
def data(self):
f = open(self.filename, 'rb')
# TODO: use find_offset to find this?
# f.seek(11)
# while True:
# f.seek(f.tell() - 11)
# if f.read(11) == b'CEvalGCData':
# break
# if f.read(1) == b'':
# f.close()
# return
f.seek(find_offset(f, b'CRawData') + 9)
strlen = 2 * struct.unpack('<B', f.read(1))[0]
tname = f.read(strlen).decode('utf_16_le')
if tname == 'CO2':
ions = [44, 45, 46]
elif tname == 'CO':
ions = [28, 29, 30]
elif tname == 'SO2,SO-SO2 Ext,SO':
# TODO: check this is in the right order
ions = [48, 49, 50, 64, 65, 66]
else:
# TODO: should save the tname somewhere for future reference
ions = [1, 2, 3]
f.seek(find_offset(f, b'CEvalGCData') + 4)
# bytes until the end converted to # of records
nscans = int(struct.unpack('<I', f.read(4))[0] /
(4.0 + len(ions) * 8.0))
dtype = np.dtype([('index', '<f4'), ('values', '<f8', len(ions))])
data = np.fromfile(f, dtype=dtype, count=nscans)
# convert time to minutes
data['index'] /= 60.
f.close()
return Chromatogram(data['values'], data['index'], ions)
def data(self):
f = open(self.filename, 'rb')
# get number of scans to read in
# note that GC and LC chemstation store this in slightly different
# places
f.seek(0x5)
if f.read(4) == 'GC':
f.seek(0x142)
else:
f.seek(0x118)
nscans = struct.unpack('>H', f.read(2))[0]
f.seek(0x10A)
f.seek(2 * struct.unpack('>H', f.read(2))[0] - 2)
dstart = f.tell()
# determine total number of measurements in file
tot_pts = 0
rowst = np.empty(nscans + 1, dtype=int)
rowst[0] = 0
for scn in range(nscans):
# get the position of the next scan
npos = f.tell() + 2 * struct.unpack('>H', f.read(2))[0]
# keep a running total of how many measurements
tot_pts += (npos - f.tell() - 26) / 4
rowst[scn + 1] = tot_pts
# move forward
f.seek(npos)
# go back to the beginning and load all the other data
f.seek(dstart)
ions = []
i_lkup = {}
cols = np.empty(tot_pts, dtype=int)
vals = np.empty(tot_pts, dtype=np.int32)
times = np.empty(nscans)
for scn in range(nscans):
npos = f.tell() + 2 * struct.unpack('>H', f.read(2))[0]
# the sampling rate is evidentally 60 kHz on all Agilent's MS's
times[scn] = struct.unpack('>I', f.read(4))[0] / 60000.
f.seek(f.tell() + 12)
npts = rowst[scn + 1] - rowst[scn]
mzs = struct.unpack('>' + npts * 'HH', f.read(npts * 4))
# there's some bug in the numpy implementation that makes this fail
# after the first time
# mzs = np.fromfile(f, dtype='>H', count=npts * 2)
nions = set(mzs[0::2]).difference(i_lkup)
i_lkup.update({ion: i + len(ions) for i, ion in enumerate(nions)})
ions += nions
cols[rowst[scn]:rowst[scn + 1]] = [i_lkup[i] for i in mzs[0::2]]
vals[rowst[scn]:rowst[scn + 1]] = mzs[1::2]
f.seek(npos)
f.close()
vals = ((vals & 16383) * 8 ** (vals >> 14)).astype(float)
data = scipy.sparse.csr_matrix((vals, cols, rowst),
shape=(nscans, len(ions)), dtype=float)
ions = np.array(ions) / 20.
return Chromatogram(data, times, ions)
def old_data(self):
f = open(self.filename, 'rb')
# get number of scans to read in
# note that GC and LC chemstation store this in slightly different
# places
f.seek(0x5)
if f.read(4) == 'GC':
f.seek(0x142)
else:
f.seek(0x118)
nscans = struct.unpack('>H', f.read(2))[0]
# find the starting location of the data
f.seek(0x10A)
f.seek(2 * struct.unpack('>H', f.read(2))[0] - 2)
# make a list of all of the ions and also read in times
ions = set()
times = np.empty(nscans)
scan_locs = np.empty(nscans, dtype=int)
scan_pts = np.empty(nscans, dtype=int)
for scn in range(nscans):
npos = f.tell() + 2 * struct.unpack('>H', f.read(2))[0]
# the sampling rate is evidentally 60 kHz on all Agilent's MS's
times[scn] = struct.unpack('>I', f.read(4))[0] / 60000.
f.seek(f.tell() + 6)
npts = struct.unpack('>H', f.read(2))[0]
# jump to the data and save relevant parameters for later
f.seek(f.tell() + 4)
scan_locs[scn] = f.tell()
scan_pts[scn] = npts
# TODO: use numpy.fromfile?
nions = np.fromfile(f, dtype='>H', count=npts * 2)[0::2]
if scn < 2:
print(npts)
print(nions)
# nions = struct.unpack('>' + npts * 'HH', f.read(npts * 4))[0::2]
ions.update(nions)
f.seek(npos)
ions = np.array(sorted(list(ions)))
data = np.empty((len(times), len(ions)), dtype=float)
for scn in range(nscans):
f.seek(scan_locs[scn])
# TODO: use numpy.fromfile?
mzs = np.fromfile(f, dtype='>H', count=scan_pts[scn] * 2)
# mzs = np.array(struct.unpack('>' + npts * 'HH', f.read(npts * 4))) # noqa
if len(mzs) == 0:
continue
ilocs = np.searchsorted(ions, mzs[0::2])
abn = (mzs[1::2] & 16383) * 8 ** (mzs[1::2] >> 14)
data[scn][ilocs] = abn
f.close()
ions /= 20.
return Chromatogram(data, times, ions)
def read_peaks(db, filename, ftype='isodat'):
if ftype is None:
with open(filename, 'r') as f:
header = f.readline()
if 'd 13C/12C[per mil]vs. VPDB' in header:
ftype = 'isodat'
else:
ftype = 'amdis'
if ftype == 'amdis':
delim = '\t'
cvtr = {'name': 'name',
'p-s-time': 'rt',
'p-s-area': 'area'}
elif ftype == 'isodat':
delim = ','
cvtr = {'name': 'peak nr.',
'p-s-time': 'rt[s]',
'p-s-area': 'area all[vs]',
'p-s-width': 'width[s]',
'p-s-d13c': 'd 13c/12c[per mil]vs. vpdb',
'p-s-d18o': 'd 18o/16o[per mil]vs. vsmow'}
headers = None
mapping = defaultdict(list)
ref_pk_info = {}
def get_val(line, cols, key):
return line.split(delim)[cols.index(key)]
with open(filename, 'r') as f:
for line in f:
if bool(re.match('filename' + delim,
line, re.I)) or headers is None:
headers = line.lower().split(',')
continue
fn = get_val(line, headers, 'filename')
if ftype == 'amdis':
# AMDIS has '.FIN' sufffixes and other stuff, so
# munge Filename to get it into right format
cmp_lvl = 2
fn = op.splitext('/'.join(fn.split('\\')[-cmp_lvl:]))[0]
# find if filtered filename overlaps with anything in the db
for dt in db.children_of_type('file'):
if fn in '/'.join(dt.rawdata.split(op.sep)):
break
else:
continue
info = {}
# load all the predefined fields
for k in cvtr:
info[k] = get_val(line, headers, cvtr[k])
# create peak shapes for plotting
if ftype == 'isodat':
rt = float(info['p-s-time']) / 60.
width = float(info['p-s-width']) / 60.
t = np.linspace(rt - width, rt + width)
data = []
for ion in ['44', '45', '46']:
area = float(get_val(line, headers,
'rarea ' + ion + '[mvs]')) / 60.
# bgd = float(get_val(line, headers, \
# 'bgd ' + ion + '[mv]'))
height = float(get_val(line, headers,
'ampl. ' + ion + '[mv]'))
# save the height at 44 into the info for linearity
if ion == '44':
info['p-s-ampl44'] = height
# 0.8 is a empirical number to make things look better
data.append(gaussian(t, x=rt, w=0.5 * area / height,
h=height))
# save info if this is the main ref gas peak
if info['name'].endswith('*'):
ref_pk_info[dt] = info
ts = Chromatogram(np.array(data).T, t, [44, 45, 46])
else:
ts = Chromatogram(np.array([np.nan]), np.array([np.nan]), [''])
mapping[dt] += [Peak(info, ts)]
# do drift correction
if ftype == 'isodat':
for dt in mapping:
ref_pks = []
hgt44 = ref_pk_info[dt]['p-s-ampl44']
d18o = float(ref_pk_info[dt]['p-s-d18o'])
d13c = float(ref_pk_info[dt]['p-s-d13c'])
for pk in mapping[dt]:
# if the d18o and height are similar, it's a ref peak
if abs(pk.info['p-s-ampl44'] - hgt44) < 10. and \
abs(float(pk.info['p-s-d18o']) - d18o) < 2.:
ref_pks.append(pk)
# get out the dd13C values and times for the ref gas peaks
d13cs = [float(pk.info['p-s-d13c']) for pk in ref_pks]
dd13cs = np.array(d13cs) - d13c
rts = [float(pk.info['p-s-time']) for pk in ref_pks]
# try to fit a linear model through all of them
p0 = [d13cs[0], 0]
def errfunc(p, x, y):
return p[0] + p[1] * x - y
try:
p, succ = leastsq(errfunc, p0, args=(np.array(rts), dd13cs))
except Exception:
p = p0
# apply the linear model to get the dd13C linearity correction
# for a given time and add it to the value of this peak
for pk in mapping[dt]:
pk.info['p-s-d13c'] = str(-errfunc(p,
float(pk.info['p-s-time']),
float(pk.info['p-s-d13c'])))
# save everything
with db:
for dt in mapping:
dt.children += mapping[dt]
def data(self):
f = open(self.filename, 'rb')
f.seek(0xC15)
yunits = string_read(f)
f.seek(0x116)
nscans = struct.unpack('>i', f.read(4))[0]
# get all wavelengths and times
wvs = set()
times = np.empty(nscans)
npos = 0x1002
for i in range(nscans):
f.seek(npos)
npos += struct.unpack('<H', f.read(2))[0]
times[i] = struct.unpack('<L', f.read(4))[0]
nm_srt, nm_end, nm_stp = struct.unpack('<HHH', f.read(6))
n_wvs = np.arange(nm_srt, nm_end, nm_stp) / 20.
wvs.update(set(n_wvs).difference(wvs))
wvs = list(wvs)
ndata = np.empty((nscans, len(wvs)), dtype="<i4")
npos = 0x1002
# try to speed up by preloading the function
unpack = struct.unpack
seek = f.seek
read = f.read
tell = f.tell
for i in range(nscans):
seek(npos)
dlen = unpack('<H', read(2))[0]
npos += dlen
seek(tell() + 4) # skip time
nm_srt, nm_end, nm_stp = unpack('<HHH', read(6))
seek(tell() + 8)
# OLD CODE
v = 0
pos = f.tell()
for wv in np.arange(nm_srt, nm_end, nm_stp) / 20.:
ov = unpack('<h', read(2))[0]
if ov == -32768:
v = unpack('<i', read(4))[0]
else:
v += ov
ndata[i, wvs.index(wv)] = v
seek(pos)
# WORKING ON A FASTER WAY TO READ ALL THIS DATA BELOW
# read in all the data
# data = np.fromfile(f, dtype="<i2", count=int((dlen - 24) / 2))
# if there are any records marked -32768, we need to reinterpret
# parts of the array as i4's
# oob_idxs = np.where(data == -32768)[0]
# locations of the cells to merge into 32-bit ints
# big_idxs = np.repeat(oob_idxs, 2) + \
# np.tile([1, 2], oob_idxs.shape[0])
# big_data = data[big_idxs].view('<i4').copy()
# remove the 32-bit cells, so the arrays the right size
# data = np.delete(data, big_idxs).astype("<i4")
# set the marker cells to the right values
# oob_idxs = np.where(data == -32768)[0]
# data[oob_idxs] = big_data
# compute cumulative sums for each chunk
# pidx = 0
# if data.shape[0] != ndata.shape[1]:
# print(big_data)
# for idx in np.hstack([oob_idxs, data.shape[0]]):
# ndata[i, pidx:idx] = np.cumsum(data[pidx:idx])
# pidx = idx
return Chromatogram(ndata / 2000., times / 60000., wvs, yunits=yunits)
def data(self):
if self._data is not None:
return self._data
else:
return Chromatogram()
def wrap_func(df, *args):
# TODO: should vectorize to apply over all columns?
return Chromatogram(f(df.values, *args), df.index, df.columns)
