search

SOM-Research / gila

Label analysis of GitHub projects
https://som-research.uoc.edu/research-tools/
17 stars 4 forks source link

Verify visualization 3 (at least for the reddit/reddit project) #19

Closed jcabot closed 10 years ago

jcabot commented 10 years ago

It doesn´t make any sense to me (0% still open?, 0 hours on average?, 1800 hours before anybody comments on average?)

belenrolandi commented 10 years ago

The problem of the percentages was solved by 587740bf25be5814a4c088558ec54c1a171b7351. The high average comment time, however, seems to be ok according to the information provided by the dataset. Some issues receive a very fast response, but within the same label there are other issues that take a long time to be answered, thus increasing notably the average.

jcabot commented 10 years ago

Just to be sure I interpret the figure the way I should, could you please describe me in words what the figure is describing for the reddit project?

Jordi

On Tue, Aug 12, 2014 at 4:47 PM, belenrolandi notifications@github.com wrote:

The problem of the percentages was solved by 587740b https://github.com/atlanmod/gila/commit/587740bf25be5814a4c088558ec54c1a171b7351. The high average comment time, however, seems to be ok according to the information provided by the dataset. Some issues receive a very fast response, but within the same label there are other issues that take a long time to be answered, thus increasing notably the average.

— Reply to this email directly or view it on GitHub https://github.com/atlanmod/gila/issues/19#issuecomment-51924126.

AtlanMod Research Team INRIA - École des Mines de Nantes Home page: http://jordicabot.com Software modeling languages portal: http://modeling-languages.com

belenrolandi commented 10 years ago

Taking a label from the Reddit project, let's say 'frontend', it goes like this: the first stopping point tells that in average it takes 1796.03hs until an issue with this label gets the first comment (from any contributor either user or administrator), then the second stopping point states that the first response from an administrator comes after 2306.35hs in average. The 'closing' branch indicates that 62.64% of the total number of issues for the label follow this path, while the end point for this branch shows that in average it takes 2929.25hs to perform a close action on an issue with this label. Likewise, the 'Still open' branch tells that 37.36% of the issues are pending, while the average age (branch end point) of those issues, is 15785.21hs. In this case there are no issues for the label with an associated 'merge event' but if there were, the interpretation would be the same as the other two cases (for an example you can look at the akka project). The numbers for all the stopping points are obtained taking the value of each individual issue (tagged with the selected label) and then calculating the arithmetic mean for this set, the presence of extreme values (i.e. issues that have long waiting times) may cause the resulting average to have the high values often seen in the diagram.

belenrolandi commented 10 years ago

I forgot to mention that the issues included in the closing path, are those issues that are rejected (i.e. closed without merging) so the three sets of issues (merged, closed and pending) are mutually exclusive. If an issue has a 'merge' and a 'close' action associated (merged then closed), the merge action wins and the issue is included in the 'merged' set.

jcabot commented 10 years ago

This comment about the distinction between close and merge should be added to the description.

The fact that the closing branch splits from the main line before the first contribution comment has no semantic at all, right? I mean the branching point is just a graphical issue, it does no have any special semantics. Right?

Should we be showing the numbers and days and not hours?

On Wed, Aug 13, 2014 at 11:12 AM, belenrolandi notifications@github.com wrote:

I forgot to mention that the issues included in the closing path, are those issues that are rejected (i.e. closed without merging) so the three sets of issues (merged, closed and pending) are mutually exclusive. If an issue has a 'merge' and a 'close' action associated (merged then closed), the merge action wins and the issue is included in the 'merged' set.

— Reply to this email directly or view it on GitHub https://github.com/atlanmod/gila/issues/19#issuecomment-52025773.

AtlanMod Research Team INRIA - École des Mines de Nantes Home page: http://jordicabot.com Software modeling languages portal: http://modeling-languages.com

belenrolandi commented 10 years ago

The comment about closed and merged issues was added to the description text in rq3 (624a226ba15c13a22370a1409e0b741ab5b44fd6) I think the split point has no special meaning, but it's arranged in such a way as to allow the end points be ordered by time. We changed the tooltip to show the information in days whenever the number of hours is greater than 24, otherwise we show the time in hours. This change was implemented by 9e8950a73eabfa79fbc6fdf4205f88917f5334f4