Test decadal case
I've made the changes decided in our last meeting. I make a summary here:
- I use indices to define time selector. The leap year condition is ignored. If the calendar is not 360-day, the time indices are calculated as 365-day calendar.
-
fcst
in recipe should be a number like2021
or a list like[2020,2021]
. The script will find the path (dcppA or dcppB) for each syear. Note that if fcst contains both dcppA and dcppB (e.g., request [2020,2021] in ecearth3-i4), the time attributes is not correct now. startR needs new development to solve this problem. But it is not a critical issue here I think. -
hcst_start
andhcst_end
in recipe can only be dcppA year. I didn't do the same change like for fcst. If it's possible to have dcppB year used in hcst, I can change the script.
To do the test, you can follow these steps:
(1) Create a branch named "user-<user_id>-decadal" (we may change the user testing strategy, but it is what you can do now)
(2) Create your own recipe based on the two decadal ones under https://earth.bsc.es/gitlab/es/auto-s2s/-/tree/master/modules/Loading/testing_recipes
(3) Run this script https://earth.bsc.es/gitlab/es/auto-s2s/-/blob/master/modules/test_victoria.R but change the first two lines
recipe_file <- "<path_to_your_recipe>"
source("modules/Loading/Loading_decadal.R") # under your auto-s2s git repository
(4) If you find problems, please report in this issue, including your recipe. If everything is fine, please let me know which dataset/variables/syear you use as well
It shouldn't take a lot of time to test it since the most complicated part is modifying the recipe. Please let me know if you have any questions. Thanks for your help!
Best,
An-Chi
Ps., here is the definition of recipe https://docs.google.com/spreadsheets/d/1TUQTIZ1yY0bgyg4nFtdydodTMzqD7Xp4D2rnQ194Azk/edit#gid=211865865