Fix for gaps in segments causing unfilled pie chart plots #392
Conversation
Codecov ReportAttention:
Additional details and impacted files@@ Coverage Diff @@
## master #392 +/- ##
==========================================
- Coverage 50.17% 49.96% -0.21%
==========================================
Files 60 60
Lines 8686 8729 +43
==========================================
+ Hits 4358 4361 +3
- Misses 4328 4368 +40
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
eagoetz
force-pushed
the
fix-network-missing
branch
from
5aed97a to
b79308f
Compare
eagoetz
force-pushed
the
fix-network-missing
branch
from
b79308f to
a600d6c
Compare
|
@iaraota We should go over this together on a Zoom call soon. I think this bug fix also would fix a problem that some folks have been confused about on the summary page pie plots |
|
@eagoetz thank you for working on this! I think that removing the blank "future segment" when the run is not complete could potentially lead to increased confusion, as the percentages do not add up to 100%. Additionally, there's a discrepancy with the O4a.3 percentages, summing to 99.9%. Is this possibly due to a rounding issue? |
|
@iaraota Thanks for the feedback. Maybe I need some more clarification:
|
|
@eagoetz The percentages are calculated using the full span, and they should be calculated using just the time covered so far in the pie chart |
|
There also appears to be a bug in the total time calculation that the DMT-ANALYSIS_READY for H1 in O4a.3 is nearly 100% and yet we're finding a missing segment amount for the network of ~1.5%. This seems excessive and needs to be investigated as a potential bug |
- Better define whether running before, during, or after the span for SegmentPiePlot - suptitle of the plot is only over the full span if 'include_future' is provided, otherwise based on before, during, or after the span of interest - better handle undefined / unknown times - gwsumm.segments.get_segments() now has optional ignore_undefined argument by default False, but if True, as is used in NetworkDutyPiePlot, this prevents an undefined segment time from impacting the network segments - gwsumm.segments.get_segments() improved reading segments from global memory and taking intersection (this also fixes an incorrect code comment claiming incorrectly that the & operator is a union) - fix a bug that was using an iterating variable - some code cleanup, more documentation and comments
eagoetz
force-pushed
the
fix-network-missing
branch
from
f61917a to
0c3d6d7
Compare
|
@eagoetz Is the "O4a.4 page" the last version of the code? I ask because the percentages on this page exceed 100%. |
Good catch, I'll take a look |
|
@iaraota The percentage issue has now been addressed. Let me know if you have further feedback. Thanks! |
eagoetz
force-pushed
the
fix-network-missing
branch
from
f958144 to
e8b796e
Compare
eagoetz
force-pushed
the
fix-network-missing
branch
from
e8b796e to
e0b11f7
Compare
This PR addresses a problem that there could be gaps in the segments, thus resulting in pie chart plots that can remain unfilled.
Here is an example for O4a.3: https://ldas-jobs.ligo.caltech.edu/~evan.goetz/summary/gps/1384873218-1389456018/segments/
Here is an example for "O4a.4" Dec 1 2023 - Jan 31 2024 ran on Jan 25 2024: https://ldas-jobs.ligo.caltech.edu/~evan.goetz/summary/gps/1385424018-1390694418/segments/
For reference, this is what the "O4a.4" interval run using the current gwsumm code: https://ldas-jobs.ligo.caltech.edu/~evan.goetz/summary/gps/1385424018-1390694419/segments/
We might consider changing the colors for the single IFO pie wedges in the
.iniconfiguration file since the red pie wedges tend to blend together. Or we can chose a different color for the "Missing segments" wedge