year to date pull statistics

Brad Roberts via Digitalmars-d digitalmars-d at puremagic.com
Mon Mar 7 19:06:34 PST 2016


A topic that rolls around periodically is the number of open pulls and or the frequency at which a 
pull is 'ignored' (no, it's not ignored, it's just lost in the noise).  I've dug up some rate of 
change statistics for the year to date for pulls to the D-P-L master branches of the dmd, druntime, 
and phobos repositories:

                        created  closed  delta
2016-03-06 - today          12      10     -2
2016-02-28 - 2016-03-05     54      48     -6
2016-02-21 - 2016-02-27     29      20     -9
2016-02-14 - 2016-02-20     32      36     +4
2016-02-07 - 2016-02-13     52      52      0
2016-01-31 - 2016-02-06     54      61     +7
2016-01-24 - 2016-01-30     40      37     -3
2016-01-17 - 2016-01-23     31      21    -10
2016-01-10 - 2016-01-16     39      42     +3
2016-01-03 - 2016-01-09     26      33     +7
2016-01-01 - 2016-01-02      2       5     +3
                            ---     ---    ---
                            371     365     -6

Not bad at all; lots of activity going on pretty much constantly.  The problem is that there's so 
much traffic with such a large backlog:

    total open: 272
    created since 2016-01-01 and still open: 86

that unless something brings a pull request to the top of the queue then it's visibility is shot.  A 
great illustration of that effect is in the age (in number of days) histogram of closed pulls for 
all pulls created since 2016-01-01:

+------+----------+
| age  | count(*) | acc% of closed
+------+----------+
|    0 |      120 | 42%
|    1 |       64 | 65%
|    2 |       25 | 73%
|    3 |       12 | 78%
|    4 |        5 | 79%
|    5 |        5 | 81%
|    6 |        6 | 83%
|    7 |        6 | 85%
|    8 |        5 | 87%
|    9 |        5 | 89%
|   10 |        2 | 89%
|   11 |        2 | 90%
|   12 |        4 |
|   13 |        1 |
|   14 |        2 |
|   15 |        1 |
|   16 |        1 |
|   17 |        1 |
|   18 |        1 |
|   20 |        1 |
|   21 |        1 |
|   22 |        1 |
|   23 |        3 |
|   24 |        1 |
|   26 |        1 |
|   28 |        1 |
|   29 |        1 |
|   37 |        1 |
|   39 |        1 |
|   41 |        2 |
|   42 |        1 |
|   46 |        1 |
|   47 |        1 |
+------+----------+
          sum: 285

So, I guess one point that comes from this data, and is already anecdotally well known, if you have 
a pull request that you believe is ready to be merged and nothing has progressed in a couple days or 
maybe a week, please ping the PR.  Chances are good that it has simply fallen off the active radar.

To see a live view of the above data:
     https://auto-tester.puremagic.com/chart.ghtml?projectid=1

The top graph is a histogram of the current number of open pull requests in monthly sized buckets. 
The next two are the same data as the first chart above broken down by week for the last year.  The 
last two are similar but looking at the issue data rather than the pull data.


Later,
Brad


More information about the Digitalmars-d mailing list