Closed iaraota closed 5 months ago
Attention: Patch coverage is 15.38462%
with 22 lines
in your changes are missing coverage. Please review.
Project coverage is 49.56%. Comparing base (
720bd0c
) to head (8f6113b
). Report is 3 commits behind head on master.
Files | Patch % | Lines |
---|---|---|
gwsumm/data/range.py | 0.00% | 18 Missing :warning: |
gwsumm/data/spectral.py | 66.67% | 2 Missing :warning: |
gwsumm/plot/builtin.py | 0.00% | 2 Missing :warning: |
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
@iaraota Thanks for finding the source of this bug! I think that we also need to fix the spectrograms, correct? For example, check out quiet and noisy state spectrograms here: https://ldas-jobs.ligo.caltech.edu/~detchar/summary/day/20240101/sei/bs/
@iaraota Thanks for finding the source of this bug! I think that we also need to fix the spectrograms, correct? For example, check out quiet and noisy state spectrograms here: https://ldas-jobs.ligo.caltech.edu/~detchar/summary/day/20240101/sei/bs/
I didn't realize that, as the spectrogram for the strain works fine. I will have a look, thank you!
@eagoetz I was looking into the problem of quiet and noise spectrograms and it seem to me that the problem is having a different definition for H1 and L1, in the configuration file we have:
[state-Quiet]
key = Quiet
name = SEI Quiet
[state-H1-Quiet]
name = SEI Quiet
key = H1-quiet
hours = 1-5,H1
[state-L1-Quiet]
name = SEI Quiet
key = L1-quiet
hours = 2-6,L1
And this page uses the state Quiet
and intersect the options for L1 and H1, and that is causing the problem. For example, when using the state H1-quiet/noisy
, the issue doesn't arise: https://ldas-jobs.ligo.caltech.edu/~iara.ota/summary/day/20240404/sei/bs
Thus, the issue does not lie within the get_spectrogram
function, but rather in how the states are generated, considering the discrepancies between the two sites. In my view, this problem falls outside the scope of this PR and should be addressed separately in another PR.
@iaraota Yeah, I had a feeling it was more complicated, but at least we have some fixes for the spectra. Let's go ahead with this PR and continue working on the spectrogram issue in another PR
This PR addresses a bug related to computing the spectrum for different states. Previously, the same spectrogram was being plotted for different states despite having different segments, due to a global variable storing spectrum data with identical keys for all states. This issue has been resolved by introducing the
state
argument to theget_spectrum
andget_spectrum_range
functions, using the state name as a key to differentiate spectrum data.The original intention behind the removed snippet below
was to incorporate the state into the
name
variable. However, this condition always failed because of the decoratoruse_segmentlist
, which was already resetingsegments = segments.active
, therefore making it impossible to insert the state in thename
variable. With the introduction of thestate
variable, this redundant condition has been removed.Additionally, the range spectrogram also required fixing as it failed to change with the state. The code introduced mirrors that used in
_get_spectrogram
to ensure the correct data is returned.Example: