ASSERT-KTH / drr

Tool & data on the correctness of Defects4 patches generated by program repair tools http://arxiv.org/pdf/1909.13694
http://arxiv.org/pdf/1909.13694
Creative Commons Attribution Share Alike 4.0 International
10 stars 6 forks source link

file name and folder conventions #16

Closed martinezmatias closed 6 years ago

martinezmatias commented 6 years ago

(we maintain the convention in this first post)

Patches

Patches folder convention:

patches/<tool_folder>/<project_folder>/<bug_id>/*.patch
eg ACS/Lang/

Patch file naming convention:

1) inside a folder of patches, each patch file has the name: patch<patch ID>-<Project>-<Bugid>-<approach>.patch

2) If there one single patch, theere is also a patch ID, which is "1" patch1-<Project>-<Bugid>-<approach>.patch

Patched version

Patched version folder convention:

patched_programs/<tool_folder>/<project_folder>/<bug_id>/<patch_id>/*.java
(we don't keep the package structure as folders)

Java file naming convention:

The Java file name is the normal class name (as usual in Java)

SophieHYe commented 6 years ago

Hi @martinezmatias “ now it groups by patch id, but it would be better in my opinion to group by Bug id.”

I agree, and actually I even thought about this before. But I think it could be easier to have all patches group by tools and then group by Bug id. In our main-table.tsv, you could search patches with the key word of bug id.

"I would unify both conventions and use the b) (note that when there is only one patch, the name would be patch1-Math-82-jkali"

That would be cool. I will update the other rest of patches to be consistent with the name rule.

Thanks for the jgenpro patches! Great work 👍

monperrus commented 6 years ago

Initial comment by Matias:

Very minor comments, it's not urgent, but it could help the readability of data: 1) inside a folder of patches (i.e., diff) each patch file has the name: patch---.patch. As the parent folder (project name) groups all patches for the project, it could be better to have the following file name for the patches:

---patch-patch. The reason is for visualization on github: now it groups by patch id, but it would be better in my opinion to group by Bug id. 2) The patch files and folder uses two different conventions: a) when an approach has only one patch for a project, the name convention is : patch---.patch. b) when an approach has more than one patch for a project, the name convention is: patch---.patch. I would unify both conventions and use the b) (note that when there is only one patch, the name would be patch1-Math-82-jkali) A similar situation happens with the folders "patch"of the patched programs
monperrus commented 6 years ago

last version of the convention in the first message of this thread