JannisNe / timewise / 6026209157

87%

push

github-actions

Speed up position masks and add clustering algorithm to spatially queried data (#102)

* avoid astropy when calculating psoition masks

* add progress bar when calculating position masks

* bump version

* use minimum(v, 1) to avoid values larger than 1 due to precision issues

* use astropy.coordinates.angular_utilities.angular_separation for stable sep calculation, consistent with main

* do INCLUSIVE cut on 90 percentile

* BUGFIX: do not convert radians to radians again

* rotate from position if positions cross 360 to 0 degree border

* use position of source rather than try to calculate median of datapoints

* supply source position

* use 90th percentile to estimate normal dist

* add 2d gaussian correction factor

* BUGFIX: double conversion from deg to rad!

* adjust docstring to degrees

* add scikit-learn for clustering algos

* fix types in docstring for calculate_position_mask

* create masks that indicate the closest source in allwise when calculating position masks

* add clustering algorithm to correctly associate datapoints after positional query

* BUGFIX: skip allwise data if there is none

* add logging, handle case of no clusters

* BUGFIX: convert input source position to rad

* add logging

* small bugfixes

* remove allwise mask

* accept non-default colors in plot_lc

* more returns from calculate_position_mask

* BUGFIX: closest_allwise_mask has to be at least None

* return more by calculate_position_mask

* remove weird formatting

* plot unused darta in grey

* remove dead code

* remove more dead code

* adjust logging

* set the 70th quantile to 1e-10 instead of 0 to avoid warnings when multiplying with np.inf

* make index right dtype in get_position_mask

* select all data within 1 arcsec, skip clustering if only one datapoint

* separate WISEBigDataLocal and WISEBigDataTestVersion

* make rings indicating query region and whitelist region thicke... (continued)

85 of 85 new or added lines in 3 files covered. (100.0%)

1779 of 2135 relevant lines covered (83.33%)

0.83 hits per line