topcrashers list: difference compared to previous timeframe and rank change calculated in the wrong direction in time
Categories
(Socorro :: Webapp, defect, P2)
Tracking
(Not tracked)
People
(Reporter: aryx, Assigned: willkg)
Details
Attachments
(2 files)
Topcrasher lists like https://crash-stats.mozilla.org/topcrashers/?product=Firefox&version=89.0a1&process_type=any&_facets_size=200
have the columns Diff and Rank. The higher Diff is, the more Rank is supposed to have lowered - which one would expect the lower Diff is.
The tooltip for Diff suggests the value should be multiplied with -1
. The calculation at https://github.com/mozilla-services/socorro/blob/bf376c159f1c16549b3078cfb202729601687cb8/webapp-django/crashstats/crashstats/utils.py#L122-L129 should be reversed.
Assignee | ||
Comment 1•4 years ago
|
||
Yeah... I think I agree that the order of operands in the calculation should be switched.
I'm also not thrilled about that table. I wonder if anyone sorts on the Diff column. If no one sorts on the Diff column, then it doesn't need to be a separate column and we can merge the % and Diff columns so it's something like:
Rank | Percent | Signature |
---|---|---|
0 | 20.08% (up 2.09%) | IPCError-browser | ShutDownKill | mozilla::ipc::MessagePump::Run |
1 | 6.63% (down 0.76%) | OOM | small |
That reduces some of the horizontal space usage in that table which seems prudent. Especially if we use arrow symbols instead of "up/down".
Anyhow, I'll take this and reverse the operands for this bug and we can think about improving this table in another bug.
Assignee | ||
Comment 2•4 years ago
|
||
Sebastian: Does this look better/clearer? I added an up/down arrow figuring that makes it easier to skim without having to look at the hover text in addition to switching the operands.
Reporter | ||
Comment 3•4 years ago
|
||
(In reply to Will Kahn-Greene [:willkg] ET needinfo? me from comment #1)
I'm also not thrilled about that table. I wonder if anyone sorts on the Diff column. If no one sorts on the Diff column
Sorting the Diff column rarely yields useful information because
- a different topcrashers could have increased the amount of crashes and reduced the share of the other signatures
- it tracks the difference as percentage points and an increase of a crash with a 0.1% share by 1.0 percentage point is more 'interesting' than a 10% share increasing by 1.1 percentage points. Showing how much it has change is still meaningful. Rounding(100 * diff/previous value) might be the better metric.
(In reply to Will Kahn-Greene [:willkg] ET needinfo? me from comment #2)
Created attachment 9215833 [details]
possible adjustment to percent diff in topcrashersSebastian: Does this look better/clearer? I added an up/down arrow figuring that makes it easier to skim without having to look at the hover text in addition to switching the operands.
Yes, the difference is as expected and old and new value are available. Thank you.
Assignee | ||
Comment 4•4 years ago
|
||
Ok. I'm going to make these changes and then tell #stability on Matrix and the stability mailing list so no one is surprised.
Assignee | ||
Comment 5•4 years ago
|
||
Assignee | ||
Comment 6•4 years ago
|
||
Assignee | ||
Comment 7•4 years ago
|
||
This was deployed to prod in bug #1706647. Marking as FIXED.
Description
•