Searching in a catalog for xrfreq='YS' will not match the entries where that frequency is written as AS-JAN. Same thing for QS vs QS-APR vs QS-JUL vs QS-OCT (and the other 2 groups).
Potential Solution
I see three solutions:
In search, expand the requested list with all equivalent xrfreq.
Normalize all frequencies that come in the catalog.
Number 2 is my favorite, but the issue is that we don't control all the ways to add entries to the catalog. I guess the normalization could be run upon loading the catalog, as well as in update.
Additional context
xr.infer_freq always returns : AS-JAN, QS-OCT, QS-NOV or QS-DEC. compute_indicators will usually use infer_freq to find the frequency to use in the catalog. Thus, if one passes freq='YS', there's an inconsistency between the output catalog and the input yaml. Only fixing the annual case is easy, but it becomes inelegant when we want to fix the seasonal cases too.
Contribution
[X] I would be willing/able to open a Pull Request to contribute this feature.
Addressing a Problem?
Searching in a catalog for
xrfreq='YS'
will not match the entries where that frequency is written asAS-JAN
. Same thing forQS
vsQS-APR
vsQS-JUL
vsQS-OCT
(and the other 2 groups).Potential Solution
I see three solutions:
search
, expand the requested list with all equivalentxrfreq
.Number 2 is my favorite, but the issue is that we don't control all the ways to add entries to the catalog. I guess the normalization could be run upon loading the catalog, as well as in
update
.Additional context
xr.infer_freq
always returns :AS-JAN
,QS-OCT
,QS-NOV
orQS-DEC
.compute_indicators
will usually useinfer_freq
to find the frequency to use in the catalog. Thus, if one passesfreq='YS'
, there's an inconsistency between the output catalog and the input yaml. Only fixing the annual case is easy, but it becomes inelegant when we want to fix the seasonal cases too.Contribution