# -*- coding: utf-8 -*-
"""
Tools for creating and merging previews.
:copyright:
The ObsPy Development Team (devs@obspy.org)
:license:
GNU Lesser General Public License, Version 3
(https://www.gnu.org/copyleft/lesser.html)
"""
from copy import copy
import numpy as np
from obspy.core.stream import Stream
from obspy.core.trace import Trace
from obspy.core.utcdatetime import UTCDateTime
[docs]
def create_preview(trace, delta=60):
"""
Creates a preview trace.
A preview trace consists of maximum minus minimum of all samples within
``delta`` seconds. The parameter ``delta`` must be a multiple of the
sampling rate of the ``trace`` object.
:type delta: int, optional
:param delta: Difference between two preview points. Defaults to ``60``.
:rtype: :class:`~obspy.core.trace.Trace`
:return: New Trace object.
This method will modify the original Trace object. Create a copy of the
Trace object if you want to continue using the original data.
"""
if not isinstance(delta, int) or delta < 1:
msg = 'The delta values need to be an Integer and at least 1.'
raise TypeError(msg)
data = trace.data
start_time = trace.stats.starttime.timestamp
# number of samples for a single slice of delta seconds
samples_per_slice = delta * int(trace.stats.sampling_rate)
if samples_per_slice < 1:
raise ValueError('samples_per_slice is less than 0 - skipping')
# minimum and maximum of samples before a static time marker
start = int((delta - start_time % delta) * int(trace.stats.sampling_rate))
start_time = start_time - start_time % delta
if start > (delta / 2) and data[0:start].size:
first_diff = [data[0:start].max() - data[0:start].min()]
else:
# skip starting samples
first_diff = []
start_time += delta
# number of complete slices of data
number_of_slices = int((len(data) - start) / samples_per_slice)
# minimum and maximum of remaining samples
end = samples_per_slice * number_of_slices + start
if end > (delta / 2) and data[end:].size:
last_diff = [data[end:].max() - data[end:].min()]
else:
# skip tailing samples
last_diff = []
# Fill NaN value with -1.
if len(last_diff) and np.isnan(last_diff)[0]:
last_diff = -1
# reshape matrix
data = trace.data[start:end].reshape([number_of_slices, samples_per_slice])
# get minimum and maximum for each row
diff = data.ptp(axis=1)
# fill masked values with -1 -> means missing data
if isinstance(diff, np.ma.masked_array):
diff = np.ma.filled(diff, -1)
data = np.concatenate([first_diff, diff, last_diff])
data = np.require(data, dtype=np.float32)
tr = Trace(data=data, header=trace.stats)
tr.stats.delta = delta
tr.stats.npts = len(data)
tr.stats.starttime = UTCDateTime(start_time)
tr.stats.preview = True
return tr
[docs]
def merge_previews(stream):
"""
Merges all preview traces in one Stream object. Does not change the
original stream because the data needs to be copied anyway.
:type stream: :class:`~obspy.core.stream.Stream`
:param stream: Stream object to be merged
:rtype: :class:`~obspy.core.stream.Stream`
:return: Merged Stream object.
"""
copied_traces = copy(stream.traces)
stream.sort()
# Group traces by id.
traces = {}
dtypes = []
for trace in stream:
# Throw away empty traces.
if trace.stats.npts == 0:
continue
if not hasattr(trace.stats, 'preview') or not trace.stats.preview:
msg = 'Trace\n%s\n is no preview file.' % str(trace)
raise Exception(msg)
traces.setdefault(trace.id, [])
traces[trace.id].append(trace)
dtypes.append(trace.data.dtype)
if len(traces) == 0:
return Stream()
# Initialize new Stream object.
new_stream = Stream()
for value in traces.values():
if len(value) == 1:
new_stream.append(value[0])
continue
# All traces need to have the same delta value and also be on the same
# grid spacing. It is enough to only check the sampling rate because
# the algorithm that creates the preview assures that the grid spacing
# is correct.
sampling_rates = {tr.stats.sampling_rate for tr in value}
if len(sampling_rates) != 1:
msg = 'More than one sampling rate for traces with id %s.' % \
value[0].id
raise Exception(msg)
delta = value[0].stats.delta
# Check dtype.
dtypes = {tr.data.dtype for tr in value}
if len(dtypes) > 1:
msg = 'Different dtypes for traces with id %s' % value[0].id
raise Exception(msg)
dtype = dtypes.pop()
# Get the minimum start and maximum end time for all traces.
min_starttime = min([tr.stats.starttime for tr in value])
max_endtime = max([tr.stats.endtime for tr in value])
samples = int(round((max_endtime - min_starttime) / delta)) + 1
data = np.empty(samples, dtype=dtype)
# Fill with negative one values which corresponds to a gap.
data[:] = -1
# Create trace and give starttime.
new_trace = Trace(data=data, header=value[0].stats)
# Loop over all traces in value and add to data.
for trace in value:
start_index = int((trace.stats.starttime - min_starttime) / delta)
end_index = start_index + len(trace.data)
# Element-by-element comparison.
data[start_index:end_index] = \
np.maximum(data[start_index:end_index], trace.data)
# set npts again, because data is changed in place
new_trace.stats.npts = len(data)
new_stream.append(new_trace)
stream.traces = copied_traces
return new_stream
[docs]
def resample_preview(trace, samples, method='accurate'):
"""
Resamples a preview Trace to the chosen number of samples.
:type trace: :class:`~obspy.core.trace.Trace`
:param trace: Trace object to be resampled.
:type samples: int
:param samples: Desired number of samples.
:type method: str, optional
:param method: Resample method. Available are ``'fast'`` and
``'accurate'``. Defaults to ``'accurate'``.
.. rubric:: Notes
This method will destroy the data in the original Trace object.
Deepcopy the Trace if you want to continue using the original data.
The fast method works by reshaping the data array to a
sample x int(npts/samples) matrix (npts are the number of samples in
the original trace) and taking the maximum of each row. Therefore
the last npts - int(npts/samples)*samples will be omitted. The worst
case scenario is resampling a 1999 samples array to 1000 samples. 999
samples, almost half the data will be omitted.
The accurate method has no such problems because it will move a window
over the whole array and take the maximum for each window. It loops
over each window and is up to 10 times slower than the fast method.
This of course is highly depended on the number of wished samples and
the original trace and usually the accurate method is still fast
enough.
"""
# Only works for preview traces.
if not hasattr(trace.stats, 'preview') or not trace.stats.preview:
msg = 'Trace\n%s\n is no preview file.' % str(trace)
raise Exception(msg)
# Save same attributes for later use.
endtime = trace.stats.endtime
dtype = trace.data.dtype
npts = trace.stats.npts
# XXX: Interpolate?
if trace.stats.npts < samples:
msg = 'Can only downsample so far. Interpolation not yet implemented.'
raise NotImplementedError(msg)
# Return if no change is necessary. There obviously are no omitted samples.
elif trace.stats.npts == samples:
return 0
# Fast method.
if method == 'fast':
data = trace.data[:int(npts / samples) * samples]
data = data.reshape(samples, len(data) // samples)
trace.data = data.max(axis=1)
# Set new sampling rate.
trace.stats.delta = (endtime - trace.stats.starttime) / \
float(samples - 1)
# Return number of omitted samples.
return npts - int(npts / samples) * samples
# Slow but accurate method.
elif method == 'accurate':
new_data = np.empty(samples, dtype=dtype)
step = trace.stats.npts / float(samples)
for _i in range(samples):
new_data[_i] = trace.data[int(_i * step):
int((_i + 1) * step)].max()
trace.data = new_data
# Set new sampling rate.
trace.stats.delta = (endtime - trace.stats.starttime) / \
float(samples - 1)
# Return number of omitted samples. Should be 0 for this method.
return npts - int(samples * step)
else:
raise NotImplementedError('Unknown method')