Smart Reprs for Groups

[richardjgowers]

In PR #1222 we fix the reprto match old behaviour, but this makes Residue/SegmentGroup representation a little long as it will always print the repr of each member:

u = mda.Universe(PSF, DCD)

rg = u.residues[:10]

<ResidueGroup [<Residue MET, 1>, <Residue ARG, 2>, <Residue ILE, 3>, <Residue ILE, 4>, <Residue LEU, 5>, <Residue LEU, 6>, <Residue GLY, 7>, <Residue ALA, 8>, <Residue PRO, 9>, <Residue GLY, 10>]>

TODO:

• Implement both __str__ and __repr__ for AtomGroup, ResidueGroup, and SegmentGroup. Ideally this is done in a generic way in GroupBase
• __repr__ is to be short and terse (similar to how AtomGroup is now) ie '<ResidueGroup with 5 Residues>'
• __str__ is to be long, pretty and informational, similar to how ResidueGroup is currently, so AtomGroup needs a similar treatment.
• the str representation needs to truncate itself if the Group is over 10 members, and instead only show the first and last three members in a verbose way. This is similar to how long numpy arrays truncate the repr of themselves:

np.arange(10000)
array([   0,    1,    2, ..., 9997, 9998, 9999])

[orbeckst]

I like your proposal

• ≤ 10 items: show full
• > 10 items: show first three and last three

For AtomGroup: for consistency showing Atom reprs in the same way would be good although I also like the quick overview over how many items there are.

In fact I would also be fine with the following:

• For all Group instances, just show a summary repr such as <AtomGroup with N atoms> or <ResidueGroup with N residues>.
• list(group) gives a full list of representations, np.array(group) gives an object array that is then formatted according to the numpy pretty printing style.

This would keep things compact (Dr @dotsdl would probably be happier) but still allow sufficient detail for anyone who wants to look (like myself).

[dotsdl]

@orbeckst can we do __repr__ giving the summary like <AtomGroup with N atoms> and __str__ giving the list of representations with compacting? This way one can always do print(ag) to get a more detailed picture.

list(atomgroup) should do what list(<iterable>) would be expected to do: give a list of individual Atom objects in order represented in the AtomGroup. Not sure what we should expect np.array(atomgroup) to do, but I guess it would do something similar but give a numpy array instead.

[orbeckst]

Compact repr and smart extended str sounds good to me. If list() always work (i.e., the items have good repr) then I am happy. array was an afterthought (mainly so that we don't have to think about smart representations...).

[richardjgowers]

Yeah I like the use of repr being terse and like current AG, and str being longwinded like the current RG. Single item repr can be long too