braincoke / hexviewer Goto Github PK

Whenever a diff has to be generated the process is the following:

• Divide the files in chunk of x KB (it can be 10KB or 100KB it is a variable)
• Obtains the chunks r1, r2, r3, ... rN for the reference file, and the chunks c1, c2, ... cM for the compared file
• Generate the diff for r1 <-> c1 , then for r2 <-> c2 ...

The problem with that approach is that it generates false positives is there was an insertion or a deletion earlier in the files as everything will be shifted according to the insertion / deletion size.

The application should be able to take into account those shifts to generate better diffs. That means being able to find parts that were modified instead of just finding "Deleted" parts in the reference file and "Inserted" parts in the compared file.