Methods not analyzed should not be considered as candidates - Githubissues

ponder-lab / Rejuvenate-Logging-Levels

An Eclipse plug-in research prototype for automatically rejuvenating log statement levels using degree of interest

Eclipse Public License 2.0

2 stars 3 forks source link

Methods not analyzed should not be considered as candidates #231

Closed khatchad closed 5 years ago

khatchad commented 5 years ago

Methods that are not analyzed (because N is too small) should not get a DOI of 0. They should have no DOI. 0 is a valid DOI value, but since we haven't analyzed these methods, we can't say for sure that their DOI is 0. As such, methods with no git history should:

[x] Not be considered for transformation. How can we transform them if we don't know their DOI?
[x] Not be listed as candidates in candidates.csv.
[x] Not be considered in boundary calculations.

khatchad commented 5 years ago

May be related (and contrary) to #123.

yiming-tang-cs commented 5 years ago

Done. 261236683d27c88344dbd1a9ec6ad96fc7677f98

yiming-tang-cs commented 5 years ago

This is the current results with the original decay (N = 24000):

results_original_decay.zip

In the previous results, the candidates' number is 54. For the new results, the candidates' number is 38. Hence, there is 16 enclosing methods which are never analyzed.

khatchad commented 5 years ago

How many commits are there total in the repositorY?

yiming-tang-cs commented 5 years ago

How many commits are there total in the repositorY?

It looks like that there is a bug here.

yiming-tang-cs commented 5 years ago

I tracked one non-candidate method in the set of log invocations. I found that the method was never touched by the mylyngit tool. But I also found some methods with the same name were analyzed by the tool. For now, I am unsure whether the method is missed by the tool or the tool did wrong for file copy and file renaming.

yiming-tang-cs commented 5 years ago

Maybe selenium is too large for debugging. I will try a smaller one.

khatchad commented 5 years ago

@saledouble Please merge this in and close this issue. Then, open a new issue regarding the difference in candidates between having the change and not having the change with the same N covering all commits. You can reference this issue there.