Open kvantricht opened 1 week ago
In fact, the sample I'm looking at (2019Usda1053623_1753
) has start_date = 2018-12-01
and end_date = 2019-11-30
.
This is exactly 12 months, but the computation of available_timesteps
:
df_pivot["available_timesteps"] = (
df_pivot["end_date"].dt.year * 12 + df_pivot["end_date"].dt.month
) - (df_pivot["start_date"].dt.year * 12 + df_pivot["start_date"].dt.month)
results in 11
timesteps. I guess this formula should do a +1
to be correct?
Converting this issue to a bug as I think something is not working correctly here (might be wrong though).
The determination of
available_timesteps
is undocumented and therefore not entirely clear for a user on what this means. For example I'm investigating a sample whereavailable_timesteps==11
, which is odd when the required number of timesteps is12
. Some documentation could help clarify this.https://github.com/WorldCereal/presto-worldcereal/blob/c1eadb50a102f99d9c7de152cd3eecfee30a2c9b/presto/utils.py#L325