Use teams for ownership

[benbc]

Built on #141.

[benbc]

@iaindillingham flattening nested lists is a case where I think the list comprehension alternative is pretty intolerable. Although here my case is again damaged by the Python stdlib with its iterator-maximalism and terrible naming of standard functional operations (see my alias at the end of the module, which you might reasonably object to).

[iaindillingham]

Just to check my understanding, you prefer the above to...

tech_repo_names = [
    name
    for names in (query.team_repos(client, org, team) for team in _TECH_TEAMS)
    for name in names
]

I'm uneasy about the alias, although it is a better name. No objections, though.

[benbc]

Yes. I think the comprehension version is horrible. Once we start splitting list comprehensions over multiple lines I actually prefer for loops. I'd be happy to consider the for loop alternative if you prefer it to flatten(). Or I'd be happy to write flatten() as a helper function, rather than a straight alias, which would allow me to absorb the list() into it.

[benbc]

def tech_repos(client, org):
    tech_repo_names = []
    for team in _TECH_TEAMS:
        tech_repo_names.extend(query.team_repos(client, org, team))

    return [r for r in _active_repos(client, org) if r.name in tech_repo_names]

or

def tech_repos(client, org):
    tech_repo_names = flatten(query.team_repos(client, org, team) for team in _TECH_TEAMS)

    return [r for r in _active_repos(client, org) if r.name in tech_repo_names]

def flatten(xss):
    return [x for xs in xss for x in xs]

Thoughts?

[benbc]

Oh, of course, black won't stand for that and unnecessarily splits the flatten() over multiple lines:

def tech_repos(client, org):
    tech_repo_names = flatten(
        query.team_repos(client, org, team) for team in _TECH_TEAMS
    )

    return [r for r in _active_repos(client, org) if r.name in tech_repo_names]

[benbc]

And my for-loop version had a bug, now fixed above. I think I now prefer the for-loop.

[iaindillingham]

My thoughts are very much "You have bad taste, Iain". To my eye, the nested list comprehension has a superficial appeal: it looks tidy. But looking tidy and being tidy are different, as the mountain of junk behind my sofa testifies. The nested list comprehension is hard to parse (start in the middle and work from right to left, go to the bottom, go to the top?).

I'm drawn to the for loop alternative. I don't think the flatten helper adds much; I'd have to read the implementation to remind myself that it only flattens lists-in-lists.

[benbc]

Every time I see one of those nested for list comps I have to remind my self that it's equivalent to nested for loops in the same order as the list comp. At that point I think "wouldn't this be fewer lines if it was just nested for loops?" and realize that the answer is "yes".

I can forgive a very simple one like my rewritten flatten(). But once they get split over multiple lines I usually abandon them.

I do like them in principle. Python just isn't a terse enough language to carry it off, I think.

[iaindillingham]

Just to check my understanding, you prefer the above to...

tech_repo_names = [
    name
    for names in (query.team_repos(client, org, team) for team in _TECH_TEAMS)
    for name in names
]

I'm uneasy about the alias, although it is a better name. No objections, though.