Reduce complexity of "possible_ancestors" from quadratic to linear.
"Possible ancestors" are all the leaf revisions with a position less than any
of the missing revisions. Previously, for every leaf revision we potentially
checked every missing revision. And with missing revisions list being sorted
from smallest to largest, that meant often traversing most of the missing
revisions list. In general, it meant performing O(R * M) operations, where R is
the number of leaf revision, and M is the number of missing revisions.
The optimization is instead of comparing against every single missing revision
we can compare only against the highest one. If a leaf revision is less than
highest missing revison, then it is already a possible ancestor, so we can stop
checking. Thus, we can go from a quadratic O(R * M) to an O(R) + O(M)
complexity, which, when merging large conflicting documents could give us a
nice performance boost.
diff --git a/src/chttpd/test/eunit/chttpd_revs_diff_tests.erl b/src/chttpd/test/eunit/chttpd_revs_diff_tests.erl
index 9a9bd25..7cdb81f 100644
--- a/src/chttpd/test/eunit/chttpd_revs_diff_tests.erl
+++ b/src/chttpd/test/eunit/chttpd_revs_diff_tests.erl
@@ -124,18 +124,23 @@
},
{Code, Res} = req(post, Top ++ Db ++ "/_revs_diff", Body),
?assertEqual(200, Code),
- ?assertEqual(
- #{
- ?DOC1 => #{
- <<"missing">> => [<<"1-xyz">>, <<"2-def">>, <<"3-klm">>],
- <<"possible_ancestors">> => [<<"2-revb">>, <<"2-revc">>]
- },
- ?DOC2 => #{
- <<"missing">> => [<<"1-pqr">>]
- }
- },
- Res
- ).
+
+ ?assert(maps:is_key(?DOC1, Res)),
+ ?assert(maps:is_key(?DOC2, Res)),
+ #{?DOC1 := Doc1, ?DOC2 := Doc2} = Res,
+
+ ?assert(maps:is_key(<<"missing">>, Doc1)),
+ ?assert(maps:is_key(<<"missing">>, Doc2)),
+ #{<<"missing">> := Missing1} = Doc1,
+ #{<<"missing">> := Missing2} = Doc2,
+ ?assertEqual([<<"1-xyz">>, <<"2-def">>, <<"3-klm">>], lists:sort(Missing1)),
+ ?assertEqual([<<"1-pqr">>], Missing2),
+
+ ?assert(maps:is_key(<<"possible_ancestors">>, Doc1)),
+ ?assert(not maps:is_key(<<"possible_ancestors">>, Doc2)),
+
+ #{<<"possible_ancestors">> := PAs1} = Doc1,
+ ?assertEqual([<<"2-revb">>, <<"2-revc">>], lists:sort(PAs1)).
t_empty_missing_revs({Top, Db}) ->
{Code, Res} = req(post, Top ++ Db ++ "/_missing_revs", #{}),
diff --git a/src/couch/src/couch_db.erl b/src/couch/src/couch_db.erl
index 5727462..70ba1c2 100644
--- a/src/couch/src/couch_db.erl
+++ b/src/couch/src/couch_db.erl
@@ -2102,28 +2102,13 @@
possible_ancestors(FullInfo, MissingRevs) ->
#doc_info{revs = RevsInfo} = couch_doc:to_doc_info(FullInfo),
LeafRevs = [Rev || #rev_info{rev = Rev} <- RevsInfo],
- % Find the revs that are possible parents of this rev
- lists:foldl(
- fun({LeafPos, LeafRevId}, Acc) ->
- % this leaf is a "possible ancenstor" of the missing
- % revs if this LeafPos lessthan any of the missing revs
- case
- lists:any(
- fun({MissingPos, _}) ->
- LeafPos < MissingPos
- end,
- MissingRevs
- )
- of
- true ->
- [{LeafPos, LeafRevId} | Acc];
- false ->
- Acc
- end
- end,
- [],
- LeafRevs
- ).
+ % Find the revs that are possible ancestors of this rev. A leaf is
+ % a possible ancestor if its position is less than any of the
+ % missing revs, and if it is less than any, it means it is also
+ % less than the maximum missing rev, so we just compare against
+ % that.
+ {MaxMissingPos, _} = lists:max(MissingRevs),
+ lists:filter(fun({Pos, _}) -> Pos < MaxMissingPos end, LeafRevs).
-ifdef(TEST).
-include_lib("eunit/include/eunit.hrl").
