Small change in the behavior of `verdi group list`

[ramirezfranciscof]

After using groups to organize my nodes, I find that the verdi group list may be improved by the following small changes:

• List order by label instead of pk (as this has a better chance of listing related group together).
• List the count of nodes by default (I often find this considerable more relevant than the User or Type string) Could be slow in large databases.
• Allow to include the "begining" of the label as an argument and filter results accordingly (ease of just listing the groups you are interested in, specially if you have a lot of groups).

Example:

$ verdi group list
  PK  Label                              Type string    User                         Node count
----  --------------------------------   -------------  -------------------------  ------------
   1  structures/3ddb_2020/all           core           myname@mail.com                   24901
   3  structures/3ddb_2020/electrides    core           myname@mail.com                      45
   2  workflows/restartgen/0started      core           myname@mail.com                       0
   4  workflows/restartgen/finished      core           myname@mail.com                       0

$ verdi group list struct
  PK  Label                              Type string    User                         Node count
----  --------------------------------   -------------  -------------------------  ------------
   1  structures/3ddb_2020/all           core           myname@mail.com                   24901
   3  structures/3ddb_2020/electrides    core           myname@mail.com                      45

EDIT: Also, enable autocomplete as in verdi group show

[chrisjsewell]

Allow to include the "begining" of the label as an argument and filter results accordingly (ease of just listing the groups you are interested in, specially if you have a lot of groups).

Especially for this, but also for the others, I would also look at verdi group path ls (also linking to #3914)

[chrisjsewell]

perhaps we even just merge them in some way

[chrisjsewell]

List the count of nodes by default

I'd say not by default, because it would be slower, but indeed it would be helpful 👍

[sphuber]

List the count of nodes by default

I would be careful with this. For bigger projects with groups with many nodes, this can give considerable slow down that make the command very annoying to use

[ramirezfranciscof]

Especially for this, but also for the others, I would also look at verdi group path ls (also linking to #3914)

perhaps we even just merge them in some way

I believe it is still valuable to have both, and as long as these are just user accessibility options with not too much maintainance required, I don't see any drawback (contrary to maintainig two database backends, for example). Although I see how verdi group path ls would be better for very long projects or if you have many projects or groups, for the ammount of groups I am currently handling it is still more confortable to just list the full path with one command rather than needing to do the ls through a folder-like structure.

I'd say not by default, because it would be slower, but indeed it would be helpful 👍

I would be careful with this. For bigger projects with groups with many nodes, this can give considerable slow down that make the command very annoying to use

Ah, good point, I didn't think of performance because it always shows it quickly for me (although my database is still under the 1MM nodes). Yeah, maybe that could go away.

[sphuber]

I would note that the last feature request is already there, but one should use the -s/--startswith option. That leaves essentially ordering by label instead of pk by default. Even though there is the -O/--order-by option, I think the label is a more sensible default.