Open brendan-m-murphy opened 4 months ago
The issue came in the reading of dask-arrays in a loop in _local_ratio
. It was taking 5-6 seconds to read from disk for every measurement, meaning a whole year's worth of data (say 1500 measurements) would take hours to filter. Fixed by adding a dataset.load()
line before the loop starts. Won't add this to devel as I think it messes with some CO2 stuff, but putting here for reference
In Issue #179, @qq23840 noted that the
local_influence
filter was running for 30 minutes without finishing.The bug in Issue #176 is addressed by PR #179, but the performance issue isn't addressed by that PR.