Rename similarity scoring for file pair

3 messages, 2 authors, 2016-06-15 · open the first message on its own page

Rename similarity scoring for file pair

From: Debayan Banerjee <hidden>
Date: 2016-06-15 22:48:36

I was going through this thread
<http://markmail.org/message/mge2spy7uqle5ghy#query:diffcore-rename.c%20algorithm+page:1+mid:ji7jkzypzqpml2xn+state:results>
trying to understand how the similarity scoring for modified files
works. I see that the algorithm uses two hashes, one for each file,
and proceeds to compare the 2 hashes to determine percentage
similarity.
I think I have an algorithm which uses only one hash and has a worst
case space complexity of O(N) where N is the number of lines
inserted/deleted/moved.
The algorithm is as follows:

def similarity(leftFileContent, rightFileContent)
{
	int lineCounter = 0;
	struct HashElement  {
	HashElement(): leftFileCount(0), rightFileCount(0) // One of these
counters will remain 0 for its lifetime. The other will be >=0
		int leftFileCount; // Increment if line is from leftFile
		int rightFileCount; // Increment if line is from right file.
	};
	Hash<lineContent, HashElement> hash; //One hash for both files
	int similartyCounter = 0;
	while(true)  {
		leftLline = leftFileContent[lineCounter];
		rightLine = rightFileContent[lineCounter];
		if (hash.contains(leftLine))  {
			HashElement h = hash.value(leftLine);
			if (h.leftFileCount > 0)   {//The previous occurence has come from
the left file
				h.leftFileCount++;  //This line is another occurence of the same
line in the same file. So we increment the counter.
			}
			else  {
				h.leftFileCount--;
				similarityCounter++;  //Hey, this line matched an earlier
occurence from the other file!
			}
			if (h.leftFileCount == 0 && h.rightFileCount == 0)
				hash.remove(leftLine);			
		}
		else  { // this line is being encountered for the first time, so add
it to the hash
			HashElement h;
			h.leftFileCount++;
			hash.add(leftLine, h);
		}
		
		if (hash.contains(rightLine))  {
			HashElement h = hash.value(rightLine);
			if (h.rightFileCount > 0)   {//The previous occurence has come from
the right file
				h.rightFileCount++;  //This line is another occurence of the same
line in the same file. So we increment the counter.
			}
			else  {
				h.rightFileCount--;
				similarityCounter++;  //Hey, this line matched an earlier
occurence from the other file!
			}
			if (h.leftFileCount == 0 && h.rightFileCount == 0)
				hash.remove(line2);			
		}
		else  { // this line is being encountered for the first time, so add
it to the hash
			HashElement h;
			h.rightFileCount++;
			hash.add(line2, h);			
		}
		lineCounter++;
		if (lineCounter == leftFileContent.size() || lineCounter ==
rightFileContent.size())
			break;
	}
	/* Upto this point we have processed an equal number of lines in both
the files simultaneously. Now we may have a file which is
	larger in size than the other. We need to process the remanant of this file*/
	bool leftFileIsLonger;
	FileContent fileContent;
	int size;
	if (leftFileContent.size() == rightFileContent.size())		
		return ((similarityCounter*100) / rightFileContent.size());
	if (leftFileContent.size() >rightFileContent.size())  {
		leftFileIsLonger = true;
	        fileContent = leftFileContent; //fileContent should really be
 a pointer to leftFileContent
		size = leftFileContent.size();
        }
	else  {
		leftFileIsLonger = false;
		fileContent = rightFileContent; //fileContent should really be  a
pointer to rightFileContent
		size = rightFileContent.size();
	}
	for (i = lineCounter + 1;  i < size; i++ )  {
		line = fileContent[i];
		if (hash.contains(line))  {
			Hash h = hash.value(line);
			if (h.leftLineCount > 0)  { //Line in hash came from left file
				if (leftFileIsLonger)
					h.leftFileCount++; //Even this line came from the same file, so
its just a repeated line.
				else  {
					h.leftFileCount--;
					similarityCounter++;  //This line came from the right file, so
increase similarityCounter and decrease the leftCounter.
				}
			}
			else  { //Line in hash came from right file
				if (leftFileIsLonger)  {
					h.rightFileCount--;
					similarityCounter++;
				}
				else
					h.rightFileCount++;
			}
			if (h.leftFileCount == 0 && h.rightFileCount == 0)
					hash.remove(line);
		}
	}
    return ((similarityCounter*100) / size); //Formula for calculating
similarity is same as git uses
}

--
Debayan Banerjee
Software Engineer

Re: Rename similarity scoring for file pair

From: Debayan Banerjee <hidden>
Date: 2016-06-15 22:48:36

On 12 April 2010 23:42, Debayan Banerjee [off-list ref] wrote:
I was going through this thread
Sorry about that. Sent it prematurely, and by mistake. Find the
attached file for the algorithm.

-- 
Debayan Banerjee
Software Engineer

Re: Rename similarity scoring for file pair

From: Jeff King <hidden>
Date: 2016-06-15 22:48:39

On Mon, Apr 12, 2010 at 11:42:59PM +0530, Debayan Banerjee wrote:
I was going through this thread
<http://markmail.org/message/mge2spy7uqle5ghy#query:diffcore-rename.c%20algorithm+page:1+mid:ji7jkzypzqpml2xn+state:results>
trying to understand how the similarity scoring for modified files
works. I see that the algorithm uses two hashes, one for each file,
and proceeds to compare the 2 hashes to determine percentage
similarity.
I think I have an algorithm which uses only one hash and has a worst
case space complexity of O(N) where N is the number of lines
inserted/deleted/moved.
That looks pretty similar to the algorithm that Andy Chup proposed. I
did some work towards integrating it, but never completed it. See these
threads for details:

  http://thread.gmane.org/gmane.comp.version-control.git/61975

  http://thread.gmane.org/gmane.comp.version-control.git/62655

So yes, it's an interesting way to go. But it needs somebody to:

  1. Implement it in C and integrate it into diffcore-rename (I did
     that, but there were some problems, see below).

  2. Confirm that the quality of results is similar to the current
     algorithm. In my case, the results were different, and I suggested
     some tweaks to improve them (see the second thread above). I don't
     remember if I ever followed up.

  3. Confirm that we really do get a speed improvement on the slow
     cases, and don't slow down or significantly bloat memory usage for
     the fast cases.

If you're interested in working on it, I'd be happy to help with
discussion and code review.

-Peff
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help