Skip secondary indexes during replay #1918
Conversation
|
Measured a speedup of 2% for replaying 40M blocks. |
pmconrad
force-pushed
the
452-683-secondary-indexes
branch
from
b45a4b2 to
8d263e1
Compare
|
Rebased to resolve conflict |
tests/common/database_fixture.cpp
Outdated
| auto current_test_suite_id = boost::unit_test::framework::current_test_case().p_parent_id; | ||
| const auto current_test_name = boost::unit_test::framework::current_test_case().p_name.value; | ||
| const auto current_test_suite_id = boost::unit_test::framework::current_test_case().p_parent_id; | ||
| const auto current_suite_name = boost::unit_test::framework::get<boost::unit_test::test_suite>(current_test_suite_id) |
There was a problem hiding this comment.
Unfortunately this line is still too long, ideally at maximum 117 characters per line.
There was a problem hiding this comment.
Hm, now I see only 109 chars in the github diff view. ISTR that I used to see more.
Also it seems to depend on font size now (smaller font means more chars), but I'm sure I tried that before too and it didn't have an effect.
There was a problem hiding this comment.
Github is great and is always improving. There is a feature request: https://github.community/t5/How-to-use-Git-and-GitHub/Feature-Request-Increase-Unified-Diff-Width-to-120-Characters/m-p/30157
BTW there are tools to change the width in individual browsers, e.g. Wide Github for Chrome, Stylish for Firefox and Fathub for running in web dev console.
| @@ -59,6 +59,8 @@ namespace graphene { namespace chain { | |||
| using std::cout; | |||
| using std::cerr; | |||
|
|
|||
| namespace buf = boost::unit_test::framework; | |||
There was a problem hiding this comment.
buf is a commonly used abbreviation of buffer, which usually appears as a local variable or a function argument. I hope we don't use it in tests.
pmconrad
modified the milestones:
4.1.0 - Feature Release,
4.0.0 - Protocol Upgrade Release
Fixes #452
Fixes #683
The indexes handled here are not used by consensus code. There's no need to update them during replay.
Two indexes have been moved into
api_helper_indexesplugin.The
grouped_ordersplugin has been modified to add its secondary index after replay, in the same way thatapi_helper_indexesdoes it.I believe the risk I mentioned in #683 (comment) is sufficiently low.
Haven't measured speedup yet.