Which water resnames would you use to detect water?

Also, note that

ag = u.select_atoms("protein and water")

will likely give you an empty atomgroup because I don't know of any molecules that are water and protein — did you mean or?

Which water resnames would you use to detect water?

From the ones that I saw mostly it would be
'H2O', 'HOH', 'OH2', 'TIP', 'TIP2', 'TIP3', 'TIP4', 'T3P', 'T4P', 'T5P', 'SOL', 'WAT', 'HHO', 'OHH'

at least those are the ones that come up quite often, although I didn't encounter 'HHO' or 'OHH' that often myself, some
tools have them like in this tool:

https://github.com/KurtzmanLab/SSTMap/blob/master/sstmap/water_analysis.py

Other resname suggestions are also welcome, if I didn't cover any important ones :)

Also, note that

ag = u.select_atoms("protein and water")

will likely give you an empty atomgroup because I don't know of any molecules that are water and protein — did you mean or?

Yes my bad, would be or, didn't notice that mistake, my bad :(

So I will start by saying that I recognise this to be a useful thing for folks that aren't power users of MDAnalysis.

Generally I don't like how these keywords obfuscate what actually happens. I.e. I know plenty of time I've sat down with a student and gone "are you sure your residues actually got selected with the protein keyword", just for them to realise that maybe they had a special residue that wasn't in the standard list.

My personal take is that, like guessing by default, I would love to see keywords go away because they enable bad mistakes. That being said, I'll be honest and say that my opinion really isn't that strong. If it helps folks do what they need then I could live with it.

So I will start by saying that I recognise this to be a useful thing for folks that aren't power users of MDAnalysis.

Generally I don't like how these keywords obfuscate what actually happens. I.e. I know plenty of time I've sat down with a student and gone "are you sure your residues actually got selected with the protein keyword", just for them to realise that maybe they had a special residue that wasn't in the standard list.

My personal take is that, like guessing by default, I would love to see keywords go away because they enable bad mistakes. That being said, I'll be honest and say that my opinion really isn't that strong. If it helps folks do what they need then I could live with it.

True, I agree that it can lead to errors, since the user would rely more on the token, similar as you mention the cases with
the protein token.

Basically it would be correct vs practical and both have good points, where the correct way would be not having them, so that such errors do not happen and practical is with the protein token currently, which makes it more easier for a non power user, but would lead to potential mistakes at some point.

Although I could imagine moving completly away from keywords now would be tricky, since MDAnalysis users got used to the protein token, so it could be tricky to remove it from them.

I will then let you @IAlibay and @orbeckst and the other coredevs decide what the best option is :)

I am in favor of convenience selectors — I'd absolutely hate having to select proteins by index or chain; a selection keyword for water (and perhaps lipid #2082 (PR #4302) and carbohydrate #4563) seem reasonable. I feel that this is one of the few occasions where convenience is quite important.

Generally I don't like how these keywords obfuscate what actually happens.

That's fair, I'd just like to have my cake and eat it, too. 🍰

Perhaps we should be thinking along guesser contexts to set what these selections match. The downside is that MDA will behave differently depending on the context in which a universe was loaded but I think that will already be effectively happening with guessers.

I am in favor of convenience selectors — I'd absolutely hate having to select proteins by index or chain; a selection keyword for water (and perhaps lipid #2082 (PR #4302) and carbohydrate #4563) seem reasonable. I feel that this is one of the few occasions where convenience is quite important.

Generally I don't like how these keywords obfuscate what actually happens.

That's fair, I'd just like to have my cake and eat it, too. 🍰

Perhaps we should be thinking along guesser contexts to set what these selections match. The downside is that MDA will behave differently depending on the context in which a universe was loaded but I think that will already be effectively happening with guessers.

True, than I will start working on the PR.

@orbeckst As for #4563 there I mentioned the possible problem in PR #4790 that due to it having a lot of different resnames
it can lead to the case where a ligand would be recognized through the token, this is the reason why it is in the
draft phase, since I am not sure if as @IAlibay mentioned, such a token that covers that many resnames could
potentially lead to an increase of errors, so there I would require also an opinion if it is reasonable to have the resnames,
since currently they cover all of the possible resnames.

Now that I see since I already started doing keywords with the water and the carbohydrates I could also do #2082,
since I am not sure if PR #4302 is still still active 🤔