Add tag all_years: True

[sloosvel]

Description

This pull request adds the option to load all years in an experiment. To be merged after #771, as I started from a commit in there because the changes were already there. I will update the branch after the merge of #771 to make these changes more clear.

Closes #1120

Link to documentation:

---

Before you get started

• ☝ Create an issue to discuss what you are going to do

Checklist

It is the responsibility of the author to make sure the pull request is ready to review. The icons indicate whether the item will be subject to the 🛠 Technical or 🧪 Scientific review.

• 🧪 The new functionality is relevant and scientifically sound
• 🛠 This pull request has a descriptive title and labels
• 🛠 Code is written according to the code quality guidelines
• 🧪 and 🛠 Documentation is available
• 🛠 Unit tests have been added
• 🛠 Changes are backward compatible
• 🛠 Any changed dependencies have been added or removed correctly
• 🛠 The list of authors is up to date
• 🛠 All checks below this pull request were successful

---

To help with the number pull requests:

• 🙏 We kindly ask you to review two other open pull requests in this repository

[valeriupredoi]

I think a more user-friendly (and introducing less overhead with less optional arguments) is to leave out either start_year or end_year or both - no start_year means the code finds all the data before end_year, no end_year means it finds all the data after start_year and neither start nor end years meaning ALL the data available. Bit of a headache to implement but much nicer and with more scope IMHO 🍺

[zklaus]

I think a more user-friendly (and introducing less overhead with less optional arguments) is to leave out either start_year or end_year or both - no start_year means the code finds all the data before end_year, no end_year means it finds all the data after start_year and neither start nor end years meaning ALL the data available. Bit of a headache to implement but much nicer and with more scope IMHO beer

I am not a fan of this implicit approach. It makes the recipe depend on the available data and thus reduces the reproducibility. But I think there is an opportunity here: Other fairly common requests are things like using the first/last 20 years of an experiment (useful for monitoring among other things). Could we have something like

timeselection_mode: normal/first 20yr/last 20yr/all_experiment/all_available

?

Not sure about the exact API here. I know this is urgent for @sloosvel and @jvegasbsc, so I would also be ok with something simpler now, but if you can come up with something easy and generic, that would be great.

[sloosvel]

I think @zklaus idea is more in line with the work that is already done, because the pull request already finds all data available and it would just be a matter of subsetting it.

[valeriupredoi]

@zklaus @sloosvel sounds good to me! Sorry, I didn't realize this was urgent and don't want to sandbag it. The implicit configuration is useful (maybe not so implicit) since I've heard a lot of them scientists making noises they want to use all the available data on a particular site, they're not concerned how much it is available, but rather they just want to grab it all, and am honestly not a big fan of more configuration arguments/keys/definitions. But let's discuss this in the future, after this PR has been merged - Saskia, I can review it now if you ready!

[sloosvel]

Let me give it a go at the timeselection range, and if it's too much trouble maybe we can do with this?

[sloosvel]

I'm closing this and opening a new PR.