It could be nice to be able to specify a start and end time and return a new timeseries cut to that start and end time. I use code like:

    tt1 = np.array(dataFull1.times)
    indexes = np.intersect1d(np.where(tt1 >= gpss[i])[0],np.where(tt1 <= gpss[i+1])[0])
    indexMin = np.min(indexes)
    indexMax = np.max(indexes)
    dataCut1 = dataFull1[indexMin:indexMax]

It isn't necessary for sure, but I could imagine people doing it enough that something like

dataCut1 = dataFull1.cut(start,end)

could be useful.

Trying to add a new set of data Axes to a plot with a colorbar fails because the colorbar Axes object doesn't have the relevant geometry attributes and methods:

>>> from gwpy.timeseries import TimeSeries
>>> ts = TimeSeries.fetch('H1:LDAS-STRAIN,rds', 'September 16 2010 06:40', 'September 16 2010 06:50')
>>> specgram = ts.spectrogram(5, fftlength=2, overlap=1) ** (1/2.)
>>> plot = specgram.plot()
>>> plot.add_colorbar()
>>> plot.add_timeseries(ts, newax=True)
Traceback (most recent call last):
  File "gwpy-78.py", line 16, in <module>
    plot.add_timeseries(ts, newax=True)
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/timeseries.py", line 403, in add_timeseries
    super(TimeSeriesPlot, self).add_timeseries(timeseries, **kwargs)
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/decorators.py", line 41, in wrapper
    return func(artist, *args, **kwargs)
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/core.py", line 604, in add_timeseries
    ax=ax, newax=newax, **kwargs)
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/decorators.py", line 41, in wrapper
    return func(artist, *args, **kwargs)
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/core.py", line 693, in add_array
    ax = self._add_new_axes(projection=projection)
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/core.py", line 358, in _add_new_axes
    geometry = self._increment_geometry()
  File "/Users/duncan/Library/Python/2.7/lib/python/site-packages/gwpy-0.1b1.dev46-py2.7.egg/gwpy/plotter/core.py", line 342, in _increment_geometry
    current = self.axes[-1].get_geometry()
AttributeError: 'Axes' object has no attribute 'get_geometry'

The "any" and "all" functions are useful for working with boolean arrays, and should therefore work on StateTimeSeries objects. Unfortunately they currently don't:

In [164]: state = gwpy.timeseries.StateTimeSeries([False])

In [165]: any(state)
Out[165]: True

In [166]: all(state)
Out[166]: True

It would be great if these functions worked, since they could be useful in identifying state activity in epochs.

The test suite is failing on python-2.6 as follows:

$ python setup.py test
...
Traceback (most recent call last):
  File "setup.py", line 353, in <module>
    'License :: OSI Approved :: GNU General Public License v3 (GPLv3)',
  File "/usr/lib64/python2.6/distutils/core.py", line 152, in setup
    dist.run_commands()
  File "/usr/lib64/python2.6/distutils/dist.py", line 975, in run_commands
    self.run_command(cmd)
  File "/usr/lib64/python2.6/distutils/dist.py", line 995, in run_command
    cmd_obj.run()
  File "build/bdist.linux-x86_64/egg/setuptools/command/test.py", line 145, in run
  File "build/bdist.linux-x86_64/egg/setuptools/command/test.py", line 126, in with_project_on_sys_path
  File "build/bdist.linux-x86_64/egg/setuptools/command/test.py", line 166, in run_tests
  File "/usr/lib64/python2.6/unittest.py", line 815, in __init__
    self.parseArgs(argv)
  File "/usr/lib64/python2.6/unittest.py", line 842, in parseArgs
    self.createTests()
  File "/usr/lib64/python2.6/unittest.py", line 848, in createTests
    self.module)
  File "/usr/lib64/python2.6/unittest.py", line 612, in loadTestsFromNames
    suites = [self.loadTestsFromName(name, module) for name in names]
  File "/usr/lib64/python2.6/unittest.py", line 586, in loadTestsFromName
    return self.loadTestsFromModule(obj)
  File "build/bdist.linux-x86_64/egg/setuptools/command/test.py", line 40, in loadTestsFromModule
  File "/usr/lib64/python2.6/unittest.py", line 583, in loadTestsFromName
    parent, obj = obj, getattr(obj, part)
AttributeError: 'module' object has no attribute 'segments'

Version 1.6.1 of numpy did not work, while 1.8.2 worked

The gwpy-plot command-line interface doesn't expose any time-axis scale options to the user, meaning they can't choose between the (default) human-readable 'Time (some unit) from epoch', and just displaying the GPS times.

I'm not sure if GWpy can actually display just GPS times, without matplotlib imposing an offset, but this should be workable pretty quickly.

The TimeSeries.spectrogram method should sanity check the relation between the stride, fftlength, and overlap parameters. At the moment if fftlength is greater than stride, an error gets thrown by scipy:

ValueError: noverlap must be less than nperseg.

The spectrogram method should raise exceptions with more sensible error messages.

The Spectrogram has a crop method to cut the array based on a start and stop time, there should probably be an equivalent method to cut from some low to high frequencies.

Adding these lines to gw_ex_plot_spectrogram.py:

import gwpy.spectrum
kwargs = {'log':True,'nbins':500,'norm':True}
specvar = gwpy.spectrum.hist.SpectralVariance.from_spectrogram(specgram,**kwargs)
print specvar
specvar = 100 * specvar
print specvar

triggers an error on the second print line

ERROR: AttributeError: 'SpectralVariance' object has no attribute '_bins' [gwpy.spectrum.hist]
Traceback (most recent call last):
File "gw_ex_plot_spectrogram.py", line 60, in
print specvar
File "/home/mcoughlin/.local/lib/python2.6/site-packages/gwpy-0.0.dev387-py2.6.egg/gwpy/data/array.py", line 164, in str
rval = str(getattr(self, key)).replace('\n',
File "/home/mcoughlin/.local/lib/python2.6/site-packages/gwpy-0.0.dev387-py2.6.egg/gwpy/spectrum/hist.py", line 76, in bins
return self._bins
AttributeError: 'SpectralVariance' object has no attribute '_bins'

I plotted a timeseries using minute trend data. I'm looking for where the HWS camera turns off, and on the minute plot it's at 60 minutes in. When I switched to second trend for better resolution, it's at 80 minutes in. With the raw data, it never turns off. I'm not sure whether this is a bug in the plotting code or maybe in NDS2.

gwpy-plot timeseries --chan "H1:TCS-ITMX_HWS_DALSACAMERASWITCH.mean,m-trend" "H1:TCS-ITMY_HWS_DALSACAMERASWITCH.mean,m-trend" --start 1113368400 --dur 7200 --ymin -0.1 --ymax 1.1 --out test-hws-mtrend.png

gwpy-plot timeseries --chan "H1:TCS-ITMX_HWS_DALSACAMERASWITCH.mean,s-trend" "H1:TCS-ITMY_HWS_DALSACAMERASWITCH.mean,s-trend" --start 1113368400 --dur 7200 --ymin -0.1 --ymax 1.1 --out test-hws-strend.png

gwpy-plot timeseries --chan "H1:TCS-ITMX_HWS_DALSACAMERASWITCH" "H1:TCS-ITMY_HWS_DALSACAMERASWITCH" --start 1113368400 --dur 7200 --ymin -0.1 --ymax 1.1 --out test-hws-raw.png

The default spectrogram created from a time series will produce low frequency of 0 Hz and also assign a logscale to the frequency axis. Upon rendering (matplotlib.savefig) this raises a ValueError since the frequency axis now attempts to span the semi-infinite interval.

(Pdb) l
 49             super(SpectrogramPlot, self).__init__(**kwargs)
 50     
 51             # plot data
 52             for i,spectrogram in enumerate(spectrograms):
 53                 self.add_spectrogram(spectrogram, newax=sep, **plotargs)
 54  ->             self.axes[-1].set_yscale('log')
 55                 self.axes[-1].fmt_ydata = lambda f: ('%s %s'
 56                                                      % (f, spectrogram.yunit))
 57                 self.axes[-1].set_ylabel('Frequency [%s]' % spectrogram.yunit)
 58     
 59             # set matching epoch for each set of axes
(Pdb) spectrogram.f0
<Quantity 0.0 Hz>
(Pdb) bt
  /home/pankow/magnetometer/gnome-data-analysis/8_16_2014_analysis/script_spectrogram(15)<module>()
-> plot = specgram.plot(norm='log',vmin=1e-7,vmax=1e-5)
  /home/detchar/opt/gwpysoft/lib64/python2.6/site-packages/gwpy/spectrogram/core.py(191)plot()
-> return SpectrogramPlot(self, **kwargs)
> /home/detchar/opt/gwpysoft/lib64/python2.6/site-packages/gwpy/plotter/spectrogram.py(54)__init__()
-> self.axes[-1].set_yscale('log')

gwpy.timeseries.TimeSeries should provide a rayleigh statistic method, with options to produce a rayleigh spectrogram.

This LDVW URL request reports an error from the gwpy-plot CLI process:

Traceback (most recent call last):
File "/usr/local/ldvw/bin/gwpy-plot", line 118, in 
result_code = prod.makePlot(args)
File "/usr/lib/python2.6/site-packages/gwpy/cli/cliproduct.py", line 778, in makePlot
self.gen_plot(args)
File "/usr/lib/python2.6/site-packages/gwpy/cli/spectrogram.py", line 87, in gen_plot
stride = max(2*secpfft, stride)
File "/usr/lib64/python2.6/site-packages/numpy/core/fromnumeric.py", line 2135, in amax
out=out, keepdims=keepdims)
File "/usr/lib64/python2.6/site-packages/numpy/core/_methods.py", line 26, in _amax
return umr_maximum(a, axis, None, out, keepdims)
ValueError: 'axis' entry is out of bounds

The original command-line is reported as

gwpy-plot spectrogram --chan 'L1:OAF-CAL_DARM_DQ' --start 1109982616 --duration 600 --out '/usr/local/tomcat_base/temp/ldvw8406441437680960249.png' --secpfft '0.500' --overlap '0.900' --fmin 40 --fmax 4000 --imin '0.500' --imax '10.000' --norm

The error comes from misuse of numpy.max (and probably also numpy.min). I am preparing a fix which will be submitted via PR.

The cli_examples/ directory generated during make html doesn't link properly to the finished HTML.

I believe that gwpy currently uses scipy.resample for resampling timeseries. Could I re commend you look at using decimate for instances where you are downsampling?

http://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.decimate.html

It uses a lowpass filter which prevents resampling artifacts. It hopefully wouldn't be more than a couple line fix, but it would also help make resample more useful when wanting to analyze the resampled data.

Could these be made available as well?

We should be able to read a Spectrum from a two-column text file (frequency, amplitude).

The following code snippet fails with the latest pull from dev, it did work in previous versions:

from gwpy.timeseries import TimeSeries
chan = 'LVE-EX:X3_810BTORR.mean,m-trend'
dur = 36000
start = 1072539420
data = TimeSeries.fetch(chan, start, start+dur, verbose=True)

produces the error:
```AttributeError: 'Channel' object has no attribute '_ifo' ```

This is one of the examples in gwpy-plot documentation.  The plot from a previous version looks like: 

https://ligodv.areeda.com/gwpy/cli_examples/cli-ts-02.png

The gwpy-plot CLI ends up raising a NotImplementedError when the legend is removed with matplotlib < 1.4:

$ gwpy-plot timeseries --chan L1:OAF-CAL_DARM_DQ --start 1102766651 --out /tmp/cli-01.png
Traceback (most recent call last):
  File "/home/duncan.macleod/.local/bin/gwpy-plot", line 5, in <module>
    pkg_resources.run_script('gwpy==0.1a10.dev427', 'gwpy-plot')
  File "build/bdist.linux-x86_64/egg/pkg_resources.py", line 534, in run_script
    dist = self.by_key.get(req.key)
  File "build/bdist.linux-x86_64/egg/pkg_resources.py", line 1434, in run_script
    def _listdir(self,fspath):
  File "/mnt/qfs4/duncan.macleod/.local/lib/python2.6/site-packages/gwpy-0.1a10.dev427-py2.6.egg/EGG-INFO/scripts/gwpy-plot", line 118, in <module>
    result_code = prod.makePlot(args)
  File "/home/duncan.macleod/.local/lib/python2.6/site-packages/gwpy-0.1a10.dev427-py2.6.egg/gwpy/cli/cliproduct.py", line 777, in makePlot
    self.annotate_save_plot(args)
  File "/home/duncan.macleod/.local/lib/python2.6/site-packages/gwpy-0.1a10.dev427-py2.6.egg/gwpy/cli/cliproduct.py", line 650, in annotate_save_plot
    self.ax.legend_.remove()
  File "/home/duncan.macleod/.local/lib/python2.6/site-packages/matplotlib/artist.py", line 137, in remove
    raise NotImplementedError('cannot remove artist')
NotImplementedError: cannot remove artist

This has been confirmed with matplotlib 1.3.1, and confirmed as not a bug with matplotlib 1.4.2.

Trying to get Omicron triggers using the fetch method of SnglBurstTable, I found that no error message is returned if the channel is not analyzed by Omicron (or if the channel name is wrong). Also no error message is given if the name of the ETG is wrong, and the help message doesn't list the possible ETGs, or specify formatting (e.g. omicron vs Omicron).

Examples:

SnglBurstTable.fetch('H1:NOT-A_CHANNEL','omicron',1111230000,1111240000)

SnglBurstTable.fetch('H1:LSC-DARM_IN1_DQ','iota',1111230000,1111240000)

Both return an empty list.

gwpy.timeseries.TimeSeries should provide coherence spectrogram method.

ERROR: AttributeError: 'Array' object has no attribute 'value'
You can trigger it by creating a spectrogram and calling its .plot() function.

[gwpy.plotter.spectrogram]
Traceback (most recent call last):
File "pylal_seismon_run", line 279, in
pylal.pylal_seismon_psd.spectra(params,channel,segment)
File "/home/mcoughlin/LAL/master/lib64/python2.6/site-packages/pylal/pylal_seismon_psd.py", line 468, in spectra
plot = medratio.plot()
File "/home/mcoughlin/.local/lib/python2.6/site-packages/gwpy-0.0.dev247-py2.6.egg/gwpy/spectrogram/core.py", line 154, in plot
return SpectrogramPlot(self, **kwargs)
File "/home/mcoughlin/.local/lib/python2.6/site-packages/gwpy-0.0.dev247-py2.6.egg/gwpy/plotter/spectrogram.py", line 52, in init
ymax = max(s.frequencies.max().value for s in spectrograms)
File "/home/mcoughlin/.local/lib/python2.6/site-packages/gwpy-0.0.dev247-py2.6.egg/gwpy/plotter/spectrogram.py", line 52, in
ymax = max(s.frequencies.max().value for s in spectrograms)
AttributeError: 'Array' object has no attribute 'value'

I've been making 12hr spectrograms and they look blocky. I believe the problem is with the stride calculation.

The referenced methods should use the url string to try and intelligently select the correct query method.

The Table.read() unified IO method for formats txt and csv doesn't accept other kwargs that can be passed to numpy.loadtxt:

>>> events = MultiBurstTable.read('/home/marissa.walker/L1_cwb_Feb28_2015/report/postprod/M1.R_RSRA_i0cc60_i1rho0_freq64_2048/data/EVENTS.txt', columns=['snr', 'flow', 'fhigh', 'duration', 'time'], usecols=[2, 13, 14, 15, 20])
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)

/mnt/qfs4/duncan.macleod/<ipython console> in <module>()

/home/duncan.macleod/.local/lib/python2.6/site-packages/astropy-1.0.1-py2.6-linux-x86_64.egg/astropy/io/registry.pyc in read(cls, *args, **kwargs)
    324 
    325         reader = get_reader(format, cls)
--> 326         data = reader(*args, **kwargs)
    327 
    328         if not isinstance(data, cls):

TypeError: table_from_ascii_rows() got an unexpected keyword argument 'usecols'

run from the LIGO-CIT data centre. I think just passing kwargs onto numpy.loadtxt is the easiest way to go.

Each of the LIGO sites uses a number of channels such as LVE-EX:X3_810BTORR for monitoring the vacuum system. Since the prefix LVE-EX doesn't parse as a proper IFO prefix, the default NDS2 server is selected as that at CIT.

While the prefix doesn't identify the IFO, it does identify the site, since they are either e.g. LVE-EX or HVE-EX, for LLO and LHO respectively.

The NDS2 server selection algorithm should be able to select a server for these channels as well.

The LIGO_LW format tables imported from glue should be able to read space-, tab-, and comma-separated text files using the standard unified read() method.

The following simple example works fine for reading CSV and is easily adaptable for other delimiters:

import numpy
from gwpy.table.lsctables import (New, SnglBurstTable)
from gwpy.time import LIGOTimeGPS
trigs = New(SnglBurstTable, columns=['peak_time', 'peak_time_ns', 'snr', 'central_freq'])
for t, f, snr in numpy.loadtxt('mytrigs.csv', delimiter=','):
    row = trigs.RowType()
    row.set_peak(LIGOTimeGPS(float(t)))
    row.central_freq = f
    row.snr = snr
    trigs.append(row)

There is a use-case for stacking a list of Spectra to generate a Spectrogram.

All filtering is out by a factor of 2_pi at the moment. On line 169 in:
/mnt/qfs5/detchar/opt/gwpysoft/lib64/python2.6/site-packages/gwpy/spectrum/core.py
frequencies should actually be w (so 2_pi*frequencies)

The use of percentiles to set the intensity axis has proved confusing. Also ldvw does not have an option to use specific values rather than percentiles.

Better help messages and labeling in ldvw is needed.

On the current HEAD of the GWpy repository, the CLI example cli-spg-01.png fails to build as follows:

$ make ../build/sphinx/cli_examples/cli-spg-01.png
cli-spg-01
Traceback (most recent call last):
  File "../bin/gwpy-plot", line 118, in <module>
    result_code = prod.makePlot(args)
  File "/home/duncan.macleod/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/cli/cliproduct.py", line 797, in makePlot
    self.gen_plot(args)
  File "/home/duncan.macleod/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/cli/spectrogram.py", line 94, in gen_plot
    snr_nfft = self.dur / (secpfft * stride)
NameError: global name 'stride' is not defined
make: *** [../build/sphinx/cli_examples/cli-spg-01.png] Error 1

@areeda, can you take a look?

As of today, the latest version of LAL in master does not jive well with GWpy. It seems that from lal import lal doesn't work anymore (although import lal does). For instance:

>>> from gwpy.timeseries import TimeSeries
>>> data = TimeSeries.read("/archive/frames/trend/minute-trend/40m/C-M-11/C-M-1111554000-3600.gwf", "C1:LSC-DARM_OUT_DQ.mean")
>>> spectrum = data.asd(method='median-mean')
  File "<stdin>", line 1, in <module>
  File "/home/max.isi/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/timeseries/core.py", line 492, in asd
    overlap=overlap, **kwargs) ** (1/2.)
  File "/home/max.isi/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/timeseries/core.py", line 459, in psd
    psd_ = method_func(self, nfft, **kwargs)
  File "/home/max.isi/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/spectrum/lal_.py", line 220, in _spectrum
    return lal_psd(*args, **kwargs)
  File "/home/max.isi/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/utils/deps.py", line 81, in wrapper
    stacklevel=2)
  File "/home/max.isi/.local/lib/python2.6/site-packages/gwpy-0.1b1.dev17-py2.6.egg/gwpy/utils/deps.py", line 50, in import_method_dependency
    return __import__(module, fromlist=[''])
ImportError: Cannot import lal.lal required by the _spectrum() method: 'No module named lal'

The current Spectrogram calculation (from a TimeSeries) uses non-overlapping chunks, with each one being used to calculate a PSD. This means the edges of each chunk are not represented properly in the overall result, and that the resolution of the resulting Spectrogram is low.

We need to rethink the way the Spectrogram is calculated, mainly for plotting purposes.

The default colour settings for DataQualityFlag.plot show the known segments as a hashed block, and the active segments in green. By default they should probably be red and green.

The TimeSeries.read method is poorly documented. One user tried the following:

data = TimeSeries.read('/archive/frames/A6/raw/H1/H-H1_R-11112/H-H1_R-1111231744-64.gwf', 'H1:LSC-DARM_IN1_DQ', 1111231750, 1111231770)

which should work, but doesn't because of the way the keyword arguments are passed around.

Improvements need to be made to the keyword argument handling, and the docstring for the read() classmethod.

    plot = dataHighpass.plot(label="highpass")
    plot.add_timeseries(dataFull,label="data")
    plot.add_timeseries(dataLowpass,label="lowpass")

This type of behavior throws a keyword error. But if you leave off the first label, the second label gets pointed to the first line, the third label to the second, etc.

It could be useful to have a fft call for a TimeSeries, just like the psd and asd.

The doctrings for the listed methods is out-of-date, referencing LAL variables and the wrong parameters, these strings need to be re-written for the current args, and kwargs.

If I want to plot the transfer function of a filter I naively tried BodePlot('zpk filter') which does not work.

Y-axis is labeled as log10 of ASD when it should be just ASD

CLI program is missing a scale kwarg when creating a time object from GPS times so the labels are off by 35 seconds.

Should have a TimeSeries method to run an auto-coherence, i.e. coherence of a time-series with a time-shifted version of itself.

Data at the sites LLO/LHO are archived to tape after about a month, perhaps 3 weeks. Data can take hours to retrieve from tape if multiple people are using it.

LDR sometimes gets behind so while data usually arrives at CIT in 10 minutes, we seen it take a week when something nasty happens.

A solution, or at least improvement would be to have the choice of server to try first depend on the age of the data being requested. A good cut-off value seems to be 2 weeks. If the data is older than that NDS2 server at CIT should be preferred over the server at the appropriate observatory.

Example:

gwpy-plot timeseries --chan H1:LSC-Y_TR_A_LF_OUT_DQ --start
Gives:

/home/detchar/opt/gwpysoft/lib64/python2.6/site-packages/matplotlib/axes.py:2760: UserWarning: Attempting to set identical bottom==top results
in singular transformations; automatically expanding.
bottom=615.266601562, top=615.266601562

• 'bottom=%s, top=%s') % (bottom, top))

gwpy-plot should detect this and do some expansion of the axis.

Calling coherence on a time series fails with an exception in mlab.py on line 1362 (my version) where it checks
if len(x) < 2 * NFFT
But in core.py around line 991 it sets
fftlength = self_.duration.value
then later turns that to samples.

Setting fftlen explicitly works fine.

I have a time series whose .value property is 35 seconds before the correct answer. I did

data2 = TimeSeries.fetch('H1:SUS-MC2_M3_MASTER_OUT_UR_DQ', 1112015670, 1112015676, verbose=True)

Then data2.epoch gives

but data2.epoch.value is 1112015635.0

For certain choices of fftlength and overlap, spectrogram2 is scaled wrong. In the example, the data is 6 seconds but the plot is squashed into 5.76 seconds. I injected a delta function, and it shows up at the wrong time, consistent with the squashing. This is only really noticeable for certain choices and not for others. Maybe there's a weird rounding error?

from gwpy.timeseries import TimeSeries

start_time = 1113031000
dur = 6

probe = 4.

darm=TimeSeries.fetch('L1:OAF-CAL_DARM_DQ',start_time,start_time+dur)

idx = int(16384*probe)
darm[idx] = 0.*darm[idx]

specgram = darm.spectrogram2(fftlength=0.05, overlap=0.049) ** (1/2.)
medratio = specgram.ratio('median')
plot = medratio.plot(norm='log', vmin=0.5, vmax=10)
plot.set_yscale('log')
plot.set_ylim(32, 8192)
plot.set_xlim(start_time, start_time+dur)
ax = plot.gca()

ax.axvline(start_time+probe,c='k',label='correct loc')
ax.axvline(start_time+(5.76/6.)*probe,c='r',label='scaled by .96')
ax.legend()
plot.savefig('test-specgram.png')

The TimeSeriesPlot prints y-axis values as float, instead of scientific notation, for values that are very small:

from gwpy.timeseries import TimeSeries
data = TimeSeries.fetch('L1:OAF-CAL_DARM_DQ', 1108807722, 1108807724)
plot = data.plot()
plot.show()

Ideally the y-axis tick labels would be -11e-5 up to -4e-5 or similar.

Bitten once again by this. I guess floats can't walk like a Quantity and Quantities can't walk like a float.

The TimeSeries.plot method should be able to handle a custom label for the plot. Currently the following code raises a TypeError:

from urllib2 import urlopen
from numpy import asarray
from gwpy.timeseries import TimeSeries

data = urlopen(
    'http://www.ligo.org/science/GW100916/L-strain_hp30-968654552-10.txt'
).read()
ts = TimeSeries(asarray(data.splitlines(), dtype=float),
                epoch=968654552, sample_rate=16384, unit='strain')
plot = ts.plot(label='TEST')
plot.set_title('LIGO Livingston Observatory data for GW100916')
plot.set_ylabel('Gravitational-wave strain amplitude')
plot.show()