def open_gwf(filename, mode='r'):
"""Open a filename for reading or writing GWF format data
Parameters
----------
filename : `str`
the path to read from, or write to
mode : `str`, optional
either ``'r'`` (read) or ``'w'`` (write)
Returns
-------
`LDAStools.frameCPP.IFrameFStream`
the input frame stream (if `mode='r'`), or
`LDAStools.frameCPP.IFrameFStream`
the output frame stream (if `mode='w'`)
"""
if mode not in ('r', 'w'):
raise ValueError("mode must be either 'r' or 'w'")
from LDAStools import frameCPP
filename = file_path(filename)
if mode == 'r':
return frameCPP.IFrameFStream(str(filename))
return frameCPP.OFrameFStream(str(filename))
def table_from_gwf(filename, name, columns=None, selection=None):
"""Read a Table from FrEvent structures in a GWF file (or files)
Parameters
----------
filename : `str`
path of GWF file to read
name : `str`
name associated with the `FrEvent` structures
columns : `list` of `str`
list of column names to read
selection : `str`, `list` of `str`
one or more column selection strings to apply, e.g. ``'snr>6'``
"""
from LDAStools import frameCPP
# open frame file
if isinstance(filename, FILE_LIKE):
filename = filename.name
stream = frameCPP.IFrameFStream(filename)
# parse selections and map to column indices
if selection is None:
selection = []
selection = parse_column_filters(selection)
# read events row by row
data = []
i = 0
while True:
try:
frevent = stream.ReadFrEvent(i, name)
except IndexError:
break
i += 1
# read first event to get column names (and map selection)
if columns is None:
columns = get_columns_from_frevent(frevent)
# read row with filter
row = row_from_frevent(frevent, columns=columns, selection=selection)
if row is not None: # if passed selection
data.append(row)
return Table(rows=data, names=columns)
def _read_framefile(framefile,
channels,
start=None,
end=None,
ctype=None,
series_class=TimeSeries):
"""Internal function to read data from a single frame.
"""
if not start:
start = 0
if not end:
end = 0
# open file
stream = frameCPP.IFrameFStream(framefile)
# get number of frames in file
try:
nframe = int(stream.GetNumberOfFrames())
except (AttributeError, ValueError):
nframe = None
# if single frame, trust filename to provide GPS epoch of data
# as required by the file-naming convention
epochs = None
try:
if nframe == 1:
epochs = [file_segment(framefile)[0]]
except ValueError:
pass
if epochs is None:
toc = stream.GetTOC()
epochs = [LIGOTimeGPS(s, n) for s, n in zip(toc.GTimeS, toc.GTimeN)]
toclist = {} # only get names once
for channel in channels:
# if ctype not declared, find it from the table-of-contents
if not ctype.get(channel, None):
toc = stream.GetTOC()
for typename in ['Sim', 'Proc', 'ADC']:
if typename not in toclist:
get_ = getattr(toc, 'Get%s' % typename)
try:
toclist[typename] = get_().keys()
except AttributeError:
toclist[typename] = get_()
if str(channel) in toclist[typename]:
ctype[channel] = typename.lower()
break
# if still not found, channel isn't in the frame
if not ctype.get(channel, None):
raise ValueError(
"Channel %s not found in frame table of contents" %
str(channel))
# find channels
out = series_class.DictClass()
for channel in channels:
name = str(channel)
read_ = getattr(stream, 'ReadFr%sData' % ctype[channel].title())
ts = None
i = 0
while True:
try:
data = read_(i, name)
except IndexError as e:
if 'exceeds the range' in str(e): # no more frames
break
else: # some other problem (likely channel not present)
raise
offset = data.GetTimeOffset()
datastart = epochs[i] + offset
i += 1 # increment frame index before any 'continue'
# check overlap with user-requested span
if end and datastart >= end and nframe == 1:
raise ValueError("Cannot read %s from FrVect in %s "
"ending at %s" % (name, framefile, end))
elif end and datastart >= end: # don't need this frame
continue
try:
dataend = datastart + data.GetTRange()
except AttributeError: # not proc channel
pass
else:
if datastart == dataend: # tRange not set
# tRange is not required, so if it is 0, it may have been
# omitted, rather than actually representing an empty
# data set
pass
elif start and dataend < start: # don't need this frame
continue
for vect in data.data: # loop hopefully over single vector
# only read FrVect with matching name (or no name set)
# frame spec allows for arbitrary other FrVects
# to hold other information
if vect.GetName() and vect.GetName() != name:
continue
# decompress data
arr = vect.GetDataArray()
dim = vect.GetDim(0)
dx = dim.dx
sx = dim.startX
if isinstance(arr, buffer):
arr = numpy.frombuffer(
arr, dtype=NUMPY_TYPE_FROM_FRVECT[vect.GetType()])
# crop to required subset
dimstart = datastart + sx
dimend = dimstart + arr.size * dx
a = int(max(0., float(start - dimstart)) / dx)
if end:
b = arr.size - int(max(0., float(dimend - end)) / dx)
else:
b = None
# if file only has ony frame, error on overlap problems
if a >= arr.size and nframe == 1: # start too large
raise ValueError("Cannot read %s from FrVect in %s "
"starting at %s" %
(name, framefile, start))
# otherwise just skip to the next frame
if a >= arr.size: # skip frame
continue
if a or b:
arr = arr[a:b]
# cast as series or append
if ts is None:
# get unit
unit = vect.GetUnitY() or None
# create array - need require() to prevent segfault
ts = numpy.require(series_class(arr,
t0=dimstart + a * dx,
dt=dx,
name=name,
channel=name,
unit=unit,
copy=False),
requirements=['O'])
# add information to channel
ts.channel.sample_rate = ts.sample_rate.value
ts.channel.unit = unit
ts.channel.dtype = ts.dtype
else:
ts.append(arr)
if ts is None:
raise ValueError("Failed to read '%s' from file '%s'" %
(str(channel), framefile))
else:
out[channel] = ts
return out
def _read_framefile(framefile, channels, start=None, end=None, ctype=None,
series_class=TimeSeries):
"""Internal function to read data from a single frame.
"""
# construct span segment
span = Segment(start is not None and start or 0,
end is not None and end or 0)
# open file
if isinstance(framefile, CacheEntry):
fp = framefile.path
else:
fp = framefile
stream = frameCPP.IFrameFStream(fp)
# get number of frames in file
try:
nframe = int(stream.GetNumberOfFrames())
except (AttributeError, ValueError):
nframe = None
# if single frame, trust filename to provide GPS epoch of data
# as required by the file-naming convention
epochs = None
try:
ce = CacheEntry.from_T050017(fp)
except ValueError:
pass
else:
if nframe == 1:
epochs = [float(ce.segment[0])]
if epochs is None:
toc = stream.GetTOC()
epochs = [LIGOTimeGPS(s, n) for s, n in zip(toc.GTimeS, toc.GTimeN)]
toclist = {} # only get names once
for channel in channels:
# if ctype not declared, find it from the table-of-contents
if not ctype.get(channel, None):
toc = stream.GetTOC()
for typename in ['Sim', 'Proc', 'ADC']:
if not typename in toclist:
get_ = getattr(toc, 'Get%s' % typename)
try:
toclist[typename] = get_().keys()
except AttributeError:
toclist[typename] = get_()
if str(channel) in toclist[typename]:
ctype[channel] = typename.lower()
break
# if still not found, channel isn't in the frame
if not ctype.get(channel, None):
raise ValueError("Channel %s not found in frame table of contents"
% str(channel))
# find channels
out = series_class.DictClass()
for channel in channels:
name = str(channel)
read_ = getattr(stream, 'ReadFr%sData' % ctype[channel].title())
ts = None
i = 0
while True:
try:
data = read_(i, name)
except IndexError as e:
if 'exceeds the range' in str(e): # no more frames
break
else: # some other problem (likely channel not present)
raise
offset = data.GetTimeOffset()
datastart = epochs[i] + offset
for vect in data.data: # loop hopefully over single vector
# decompress data
arr = vect.GetDataArray()
dim = vect.GetDim(0)
dx = dim.dx
sx = dim.startX
if isinstance(arr, buffer):
arr = numpy.frombuffer(
arr, dtype=NUMPY_TYPE_FROM_FRVECT[vect.GetType()])
# crop to required subset
dimstart = datastart + sx
dimend = dimstart + arr.size * dx
a = int(max(0., float(span[0]-dimstart)) / dx)
if span[1]:
b = arr.size - int(max(0., float(dimend-span[1])) / dx)
else:
b = None
if a >= arr.size or b is not None and b <= a:
raise ValueError("Span %s not covered by FrVect for %r "
"in %s" % (span, name, fp))
if a or b:
arr = arr[a:b]
# cast as series or append
if ts is None:
# get unit
unit = vect.GetUnitY() or None
# create array - need require() to prevent segfault
ts = numpy.require(
series_class(arr, t0=dimstart+a*dx, dt=dx, name=name,
channel=channel, unit=unit, copy=False),
requirements=['O'])
# add information to channel
ts.channel.sample_rate = ts.sample_rate.value
ts.channel.unit = unit
else:
ts.append(arr)
i += 1
if ts is None:
raise ValueError("Channel '%s' not found in frame '%s'"
% (str(channel), fp))
else:
out[channel] = ts
return out