It's very good API.

Could you tell me about how to integrate to GWT.

Although this is nor the place to ask user questions, nor the place to answer them (Neil, feel free to delete both): @lopmuj: just call it using JSNI - something like this will do the trick:

    private native String getDiffString(String original, String amendment)/*-{
        var dmp = new $wnd.diff_match_patch(); 
        var d = dmp.diff_main(original, amendment);
        dmp.diff_cleanupSemantic(d);
        return dmp.diff_prettyHtml(d);
    }-*/;

But this is so obvious that I'm not even sure if you tried finding it out by yourself in the first place ..

This is an incredible piece of work, nice job Neil!

This is completely what I was looking for after getting fed up with SVN&Co. bogus APIs Thanx Neil!

Excellent work. Thanks!

Simply awesome!

I just built a wrapper around your calls, customize it according to my needs! I was initially tempted to mimic the GNU "diff" behavior (dumping everything into the stdout), but when I saw the results in HTML....I was "...wwoooowed! this is great, it's really easy to spot the differences (Inserts/Deletes), as they are coded: "Insertions" show up in the HTML with a green background and underlined font whilst "Deletions" show up with a red background and Strike-through font...

Great job!

Thanks!

This is a great tool. However, I wasn't able to understand the matching algorithm. Is there a way to ignore the location and match based just on the pattern in match_main. Basically, I'm trying to identify if a particular word or phrase (or its variant (like a missing hyphen etc.)) appears in the given text.

Neil, Excellent!

One question: If I diff a big document against a small one, I get a big "DELETE" block in the resulting diff. Is this really nessesary? I think a "delete XX chars" would be enough?

Bob

jjknopf, how you format/display/visualize the diff is entirely up to you. diff_prettyHtml(diffs) is one way to render a diff, it's a trivial function, you can roll your own. Take a look at diff_toDelta(diff), it's not listed in the above API, but it is very compact and may be similar to what you are looking for.

I'm trying to match a piece of text against several "candidates'. Is there some way to compare the results of two matches to see which is "best"?

Thanks!

I've added a new function to all language versions called diff_levenshtein(diff) which takes a diff and reports the number of character insertions, deletions or substitutions it represents. Thus one can diff your target against several options and compute the Levenshtein distance for each to find the option that's closest to the target.

Hello all,

I would like to know. based on this implementation, how can I create a list of diffs to simulate something like this : http://www.caffeinated.me.uk/kompare/

What I need is the following : - be able to know "inserted" block of text - be able to know "deleted" block of text - be able to know "changed" block of text

- if the block changed, I would like to know if the block is completely changed on where is the change in the block.

Is it possible with this implementation ? Thank you very much !

Jimmy

Jimmy this would better be asked in the newsgroup: http://groups.google.com/group/diff-match-patch

But the simple answer is run diff_main(text1, text2), then step through the resulting array of diffs. They will be insertions, deletions or equalities. Obviously a 'change' is a deletion and an insertion next to each other.

Nice work. However, how hard would it be add unpatch(), that is, reversepatch()? The second issue is that I need to port this to php ...

Unpatching would just be to loop through the diff, swapping DIFF_INSERT with DIFF_DELETE, then apply the patch. As for PHP, there's a partial translation which someone wrote, email me and I'll send you a copy.

great job, this is what I'm looking for! thanks a lot.

A terrific work! Thanks a lot!

good stuff, just what i was looking for.

I need the PHP implementation as well. Neil -- I don't have your email but could you email me it at mathews.kyle@gmail.com? Thanks, that'd be really helpful.

Great Stuff! Please send me the php-implementation as well. I will send it back if i extend it.

E-Mail: dmp@shwups.ch

Great job!

Hi, but what about patch conflicts? In standard GNU patch it is implemented. How we can recognize patching conflict occured?

Thanks

Tom, look at the results from patch_apply. There are two returned values, one is the patched text, and the other is a list of booleans indicating which patches were applied and which failed.

I used a "Demo of Patch" example (http://neil.fraser.name/software/diff_match_patch/svn/trunk/demos/demo_patch.html). 1. I put "Old Version" text like this: <appSettings>

<add key="AccessAllowedIPAddresses" value="127.0.0.1" />

<setting name="SqlServerDataPath" serializeAs="String">

<value>C:\DATA_PATH</value>

</setting>

2. I put "New Version" text like this: <appSettings>

<add key="AccessAllowedIPAddresses" value="127.0.0.1" /> <add key="SqlServerDataPath" value="C:\data_path" />

3. I computed a patch.

4. I put "Old Version" text like this: <appSettings>

<add key="AccessAllowedIPAddresses" value="127.0.0.1" />

<setting name="SqlServerDataPath" serializeAs="String">

<value>C:\REAL_DATA_PATH</value>

</setting>

5. I applied a patch. No conflicts were found. I got text like "New Version" text. Why weren't patch conflicts found?

What a remarkable piece of work! I'm impressed. :)

Would it be possible to add functionality to work with files, not just strings?

I compare 2 large texts and it gives me one big delete and one big insert even with "No cleanup". If I make texts smaller it shows differences nicely. Something is wrong.

student00x, nothing is wrong, your diff exceeded the time limit and it was forced to return a valid, but non-optimal diff. Try increasing diff_match_patch.Diff_Timeout?, or setting it to zero.

I have two documents (.txt) with accented characters (è, ù). After taking the patches, I try to use the function "patch_toText(patches)" but I obtain a segmentation fault. Is there a solution?. I'm working with C language.

Impressing !

Nice work. I'm evaluating this for use in a larger project requiring client-side text operations. The project is built on the GWT-platform. Are there any recommendations as to: 1. Use the supplied javascript library via JSNI (GWT). 2. Compile the supplied java library to javascript using the GWT-compiler.

I'm considering option 2 to be the most feasible. I might have to verify the correctness of the result but i will get highly optimized, crosscompiled, javascript. And i don't have to include any JSNI-code (which should be avoided when possible). Any thoughts anyone?

Very Cool, Is it possible to make it an option to have the comparison unit be a word instead of a char? An example where the output is confusing (to me at least) is Text1 = $200.00 abc $100.00 Text2 = $205.00 abc $102.50

Output is ... $20<strike>0</strike>5.00 abc $10<strike>0.0</strike>2.50

This is much less intuitive that <strike>200.00</strike>205.00 abc <strike>100.0</strike>102.50

Very nice and perfectly simple to use!

Very impressive piece of work. Thanks for sharing it with the world!

I too would like the php version, whatever state it's in.

Can i use this Library for generating diffs of binary files.

Yes Mike, DMP works great on binary files. When stress its suitability for 'text', that's meant as opposed to structured content such as XML; there's no problem with binary.

Hey Neil,

Great job on this package. Truly remarkable work. :) I have a question; I'd like to develop an editing app (non realtime) that behaves like GNU diff3 (compare docs A and B to older, parent copy C). I want to graphically represent the successful merges and and conflicting segments. I was looking at using patch_apply for this but I noticed your comment in the function description:

"...A more informative API is in development..."

I'm assuming I probably won't be able to build a conflict resolver built on the output of this api until that API improvement you mentioned is ready?

Thanks again!

-Mike

Nice job. One correction should be done in the Java-version though: The Diff-subclass implements a equals()-method but NO hashcode()-method. A hashCode is needed if you e.g. put Diff-instances into a HashMap? to count occurencies of certain Diffs.

Nice Work! Would it be possible to provide a line-level diff? I'm looking for a tool that would provide a line-level diff between two text docs that can be stored in db and later recreate a particular version by either applying/removing patches.

for eg:

Original:
This simple strategy is relatively fragile.
Minor changes between the two base documents can result in failure.
In the above example if the function name had been changed from "maxbits" to "max_bits" the patch would have failed to match on the first pass, but would have matched on the second pass (once the first line of the prefix context was ignored)
However, if the "b++;" had been changed to "b=b+1;", the patch would fail, despite the fact that all the other context lines match.



Revised:
This simple strategy is relatively faster.
Minor changes between the two base documents can result in failure.
In the above example if the function name had been changed from "maxbits" to "max_bits" the patch would have failed to match on the first pass, but would have matched on the second pass (once the first line of the prefix context was ignored).

--- O
+++ R
@@ -1,1 +1,1 @@
-This simple strategy is relatively fragile.
+This simple strategy is relatively faster.
@@ -3,2 +3,1 @@
-In the above example if the function name had been changed from "maxbits" to "max_bits" the patch would have failed to match on the first pass, but would have matched on the second pass (once the first line of the prefix context was ignored).
-However, if the "b++;" had been changed to "b=b+1;", the patch would fail, despite the fact that all the other context lines match.
+In the above example if the function name had been changed from "maxbits" to "max_bits" the patch would have failed to match on the first pass, but would have matched on the second pass (once the first line of the prefix context was ignored).

Thanks

Since the question of line and word level diffs is frequently asked, I've created a page describing how to do this:

http://code.google.com/p/google-diff-match-patch/wiki/LineOrWordDiffs

Thanks Neil!

Is there an option to unpatch patches from the latest text in order to obtain the original?

I don't see one in the api list.

	@Test
	public void testDiffLineMode(){
		dmp = new diff_match_patch();
		String text1 = "The quick brown fox jumps over the lazy dog.";
		String text2 = "That quick brown fox jumped over a lazy dog.";
		LinesToCharsResult a = dmp.diff_linesToChars(text1, text2);   
		List<String> lines = a.lineArray;
		
		//create the patches
		LinkedList<diff_match_patch.Diff> diffs = dmp.diff_main(a.chars1,a.chars2);
		dmp.diff_charsToLines(diffs, lines);		
		LinkedList<Patch> patches = dmp.patch_make(diffs);
		
		//apply patches to text1 to get text2
		Object[] results = dmp.patch_apply(patches, text1);
		Assert.assertEquals("That quick brown fox jumped over a lazy dog.", results[0]);
		
		boolean[] boolArray = (boolean[]) results[1];
		Assert.assertTrue(boolArray[0]);
		  
		//remove patches from text2 to get text1
		....something like this ...
		results = dmp.patch_remove(patches, text2);  --> remove the patch(es) from text2 to obtain text1				
	}	

Unpatching can be done by just looping through the diff, swapping DIFF_INSERT with DIFF_DELETE, then applying the patch.

This page is for documentation of the API. meena, could you move this discussion to the newsgroup? I'll delete these posts once you have done so. Thanks.

Please go ahead and remove them (including this one). Going forward I'll use the newsgroup. thanks

can i use it to compare two HTML files

Thanks Vishal

Grate job Thank you but in my working i want to escape html tag example text1 = "<span class='some_class'>This is my Text</span>"; text2 = "<div>This is my Text</div>"; it equals text(not check html tag <span>,<div>) how i use diff_match_patch please suggest

Neil, is it possible to see how php version works? farhat.aminov@gmail.com

Hello!

One question. It is possible to know the line number where an INSERT diff appear?

Thanks for all!

ANN: A Ruby implementation is available as a GEM. Gem can be installed with RubyGems?, "gem install diff_match_patch" Source Code here https://github.com/kalmbach/diff_match_patch

Big thx for this great work! It's just what I was looking for :)

Hi , I have used the following program to compare data of 2 text files , got to check whether first file data is present in second file , comparing row wise . .......................................... import java.io.BufferedReader?; import java.io.File; import java.io.FileNotFoundException?; import java.io.FileReader?; import java.io.IOException; import java.util.ArrayList?; import java.util.List; public class ComparingTextFiles? {

private BufferedReader? mReader = null; private static String firstFileRow; private static String secondFileRow; private static ArrayList? firstList = new ArrayList?(); private static ArrayList? secondList = new ArrayList?();

private static ArrayList? differentDataRows = new ArrayList?();

public ComparingTextFiles?(String aFileName) throws java.io.FileNotFoundException? {

mReader = new BufferedReader?(new FileReader?(aFileName));

} public static void main(String args) {

//compareFiles("D:/first.txt","D:/second.txt");

compareFiles("D:/TPDealsNashFinchGreatLakes20120212.txt","D:/VendorNumber?_TPDealsNashFinchGreatLakes20120212.txt"); }

public static void compareFiles(String firstFile,String secondFile) {

try {

ComparingTextFiles? comparingTextFiles1 = new ComparingTextFiles?(firstFile);

while ((firstFileRow = comparingTextFiles1.readLine()) != null) {

firstList.add(firstFileRow);

}

ComparingTextFiles? comparingTextFiles2 = new ComparingTextFiles?(secondFile);

while ((secondFileRow = comparingTextFiles2.readLine()) != null) {

secondList.add(secondFileRow);

}

} catch (Exception e) {

e.printStackTrace();

}

resultOfComparingFiles();

}

public static void resultOfComparingFiles() {

for(Object rowFromFirstFile : firstList)

{

String row = (String)rowFromFirstFile;

if(!(secondList.contains(row))) {

int index = firstList.indexOf(row); int atRowNo = index++;

System.out.println(" ... This row in first file is not present in second file ... ");

System.out.println(row);

System.out.println(" at row no "+atRowNo); System.out.println(".....................");

}

}

} public String readLine() throws java.io.IOException {

if (null == mReader)

throw new java.io.IOException();

String lLine = mReader.readLine(); if (null == lLine)

this.close();

return lLine;

} public void close() {

if (null == mReader)

return;

try {

mReader.close(); mReader = null;

} catch (Exception e) { }

}

}

..................................

its working .

Can you unapply a patch?