[PATCH 8 of 9 "] discovery: cache the children mapping used during each discovery

[Yuya Nishihara]

On Tue, 05 Mar 2019 18:39:19 +0100, Pierre-Yves David wrote:
> # HG changeset patch
> # User Pierre-Yves David <pierre-yves.david at octobus.net>
> # Date 1551314900 -3600
> #      Thu Feb 28 01:48:20 2019 +0100
> # Node ID 6127533daac7b331ba980d70040d55ee938ce106
> # Parent  fe57cbd053432d10bccd3955072a4c6c581e3355
> # EXP-Topic discovery-speedup
> # Available At https://bitbucket.org/octobus/mercurial-devel/
> #              hg pull https://bitbucket.org/octobus/mercurial-devel/ -r 6127533daac7
> discovery: cache the children mapping used during each discovery
> 
> During discovery, the `undecided` set keep shrinking. Therefore, the map
> computed for an iteration N will be valid for iteration N+1. Instead of
> computing the same data over and over we cache it the first time.

Can you copy this to the function doc or inline-comment?

>      def addcommons(self, commons):
>          """registrer nodes known as common"""
> @@ -173,15 +174,14 @@ class partialdiscovery(object):
>  
>      def _childrengetter(self, revs):
>  
> +        if self._childrenmap is not None:
> +            return self._childrenmap.__getitem__

Without the comment above, this code would look invalid since our cache
strategy ignores the revs argument (or relying on the fact that undecided
set would never grow.)