Data Cleansing Manual

[masud-technope]

Phase I: Steps to be followed

1. Select an issue report like this one
2. Determine whether it discusses about a bug or a new feature.
3. For example, this is a new feature, but this is a bug.
4. Take a spreadsheet for each subject system, and mark its issue IDs as either new feature (NF) or bug (B).
5. We have 8 subject systems. So, please create separate files for the systems.
6. We have 2,885 issue reports from 8 systems. This should take a few days I guess.

Phase II: Steps to be followed

1. Check the changed files for each bug report or feature request. For example, this is the changeset for this bug report.

2. Count the number of changed Java files in each change set.

3. If it is only one, then it is a valid changeset.

4. If the count<=5, take a close look at the files. Are they really related to the bug report/feature request? Do they look related to the bug report or feature request? If yes, mark it as a valid changeset. You should spend at most 3 minutes for this.

5. If the count>5, take a closer look at the changeset. Are they really related to the bug report/feature request?. According to my experience, they could have changed files which are not related to the bug fix or the feature implementation. If you find that a changeset contains one or more files unrelated to either the bug fix or the feature implementation, mark the changeset as bloated changeset. For example, this is definitely a bloated changeset for the issue report #263537. You should spend at most 6 minutes for this. If you cannot decide within 6 minutes, just mark it as bloated changeset.

6. The output format for each issue report entry:
   BugID, #ChangedFiles, ** #Valid/Bloated**
   Please create separate files for individual subject systems.

7. You can use code such as Valid Changeset=VC and Bloated Changeset=BC