[PATCH 06 of 15] commitctx: return a richer object from _prepare_files

[Yuya Nishihara]

On Wed, 29 Jul 2020 18:57:36 +0200, Pierre-Yves David wrote:
> # HG changeset patch
> # User Pierre-Yves David <pierre-yves.david at octobus.net>
> # Date 1595684952 -7200
> #      Sat Jul 25 15:49:12 2020 +0200
> # Node ID 88cc2b7a810243e8c101933fd99778ce772ac316
> # Parent  6b4ada6dca3ff5591c5dea32ffbe2fa22f4e5ed7
> # EXP-Topic files-change
> # Available At https://foss.heptapod.net/octobus/mercurial-devel/
> #              hg pull https://foss.heptapod.net/octobus/mercurial-devel/ -r 88cc2b7a8102
> commitctx: return a richer object from _prepare_files

Queued the remainder, thanks.

> +class ChangingFiles(object):
> +    """A class recording the changes made to a file by a revision
> +    """
> +
> +    def __init__(
> +        self, touched=(), added=(), removed=(), p1_copies=(), p2_copies=(),
> +    ):
> +        self._added = set(added)
> +        self._removed = set(removed)
> +        self._touched = set(touched)
> +        self._touched.update(self._added)
> +        self._touched.update(self._removed)
> +        self._p1_copies = dict(p1_copies)
> +        self._p2_copies = dict(p2_copies)
> +
> +    @property
> +    def added(self):
> +        return frozenset(self._added)

I generally avoid using @property function that does non-trivial work per call.
Does this frozenset business really matter?

> +    def mark_added(self, filename):
> +        self._added.add(filename)
> +        self._touched.add(filename)

> +    def mark_copied_from_p1(self, source, dest):
> +        self._p1_copies[dest] = source

It doesn't make sense that mark_added() updates both added and touched,
but mark_copied() doesn't.

Since ChangingFiles is basically a data object, I think sanity check can be
split into a function, and called occasionally e.g. before serializing the
metadata.