The Delivery Documentation is available on-line at:
Aparentlly the link did not work but here it is again, it tests great with fork.c which is the file Bob first gave us!!!
Sincerely,
Daniel Fernandez
Well kids this is it, this code removes the stopwords, expands and substitutes for function names. It's here to stay, and its fully integrated just run!!
Sincerely,
Daniel Fernandez
Download file
I think this works but java gives me an out of memory error!!!
Hey kids -- here is the midterm presentation we can use as a template for the final presentatation.
Enjoy....
Hi folks, I just finnished with my part. The file is called filter.java and it assumes function names and comments are placed inside vectors like the program that Matt wrote indicated.
It pulls out the stopwords first, then it expands those function names that need to be expanded and finally it replaces the words in the domain dictionary file. The files stopwords.txt and domaindic.txt are attached, see you guys tomorrow!!
Also includes "flow.txt" be sure to look at for specific breakdowns on who's doing what.
OK, so first we discussed the final things needed to make our SCDM bulletproof, then we spent some brainstorming time on each segment.
First issue is the stop word removal:
we are placing the stop words in a Collection, then using Java's vector removeAll method to remove them all from the function comments vector:
func.comments.removeAll(stopWords);
Next issue is the gcc preprocessing
Using a simple:
gcc -E -C file.c
The gcc -E option performs only preprocessing. The -C option leaves all comments intact. This creates an extremely large file. We now have lots of extra comments that are not ours (from each .h included, etc.), so we need to rehash this idea, and implement it as an option, to execute preprocessor commands. We may need to simply skip many of the generated preprocessor segments, this may be easy to do since they are labelled as they are inserted.
File directories
Needs to be bulletproof - it is not creating the new directories, and is file system dependent.
Integration
We have started integrating many of the methods, however there is excess code for directory structures manipulation. We will correct this next week.