Closed brancomat closed 2 years ago
It did not work when there was only a single "good" step to be copied from input to output without need to compute any difference. I pushed an update that corrects, but this matter of statistical processing on intervals is very complicate and I would like to improve it further.
In commit 9a7dddfd9c42ff3129baa69761618a609031eb09 and surroundings a better search algorithm for existing ready intervals has been implemented. It is exact for gridded data, while for sparse data, due to how data are copied from input to output, it may not work perfectly with full steps, however it is netter than before. Some tests in analysis mode are needed.
(as seen in https://github.com/ARPA-SIMC/arkimaps/issues/117)
test data: tp_2Iruc.tar.gz
vg6d_transform --comp-stat-proc 1 --comp-frac-valid=0 --comp-step="0 12" tp_2Iruc.grib tp_12h.grib
vg6d_transform --comp-stat-proc 1 --comp-frac-valid=0 --comp-step="0 12" --comp-full-steps tp_2Iruc.grib tp_12h_fullsteps.grib
forcing
--comp-start="2022-05-18 00"
(the reftime of the example data) still doesn't work but forcing--comp-start
to a subsequent hours gives the expected output:This is curious since in the first example (the one without
--comp-full-steps
) it seems to be able to compute the 0-12 interval.(bonus issue: exporting
LOG4C_PRIORITY="debug"
gives no debug output)