Compare two text lines and highlight difference

[avk]

Here's a Myers diff unit:

Code: Pascal  [Select][+][-]

1. unit uMyersDiff;
2. ...
3.  

@Thaddy, it is suspected that the authorship of the uMyersDiff unit belongs to ChatGPT.

[440bx]

@Thaddy, it is suspected that the authorship of the uMyersDiff unit belongs to ChatGPT.

Now, THAT is an interesting statement.

These A.I things only produce derived works, which means it is highly dubious they could claim ownership of anything but, it's difficult not to wonder if someone or some company will eventually attempt to claim ownership of something that is the product of A.I.

"It's mine, I stole it fair and square" ... reminds me of a company who tried to pocket the concept of a GUI for itself.

[simone]

compilation with FPC-3.2.2 requires some changes in the source code, despite the claimed compatibility with Lazarus;

In February, I submitted patches to the author, who accepted them in the following commit:

https://github.com/rickard67/TextDiff/commit/97f0bce070658c6206616bbe5041284c06008851

The changes I proposed fixed the compatibility issues with newer versions of Lazarus/Fpc.

Now, I've noticed that, unfortunately, new incompatibilities have been introduced in subsequent commits, which I'll try to fix as soon as I can.

I reported the compilation issues on Lazarus 4.0 / Fpc 3.2.2 to the author and submitted some small changes that fix them:

https://github.com/rickard67/TextDiff/issues/8

@avk: In the past, when I used this library, I too had doubts about the correctness of the results obtained with these implementations of the lcs algorithms

[Thaddy]

Here's a Myers diff unit:

Code: Pascal  [Select][+][-]

1. unit uMyersDiff;
2. ...
3.  

@Thaddy, it is suspected that the authorship of the uMyersDiff unit belongs to ChatGPT.

No. it was verified for correctness by DeepSeek, and Copilot, though. Hence you might get the same or about the same answer.
Happens all the time.
As you can see from my other answers, if I use AI to write code or partial code I always add  the source.

[avk]

...
"It's mine, I stole it fair and square"
...

Isn't that how all business works?
But I just meant that the code was produced by some so-called AI.

...
@Thaddy, it is suspected that the authorship of the uMyersDiff unit belongs to ChatGPT.

No. it was verified for correctness by DeepSeek, and Copilot, though.
...

However, the code seems to have a some small but very distinctive feature: it doesn't work.

[Thaddy]

Where doesn't it work?
Because it works:

Code: Bash  [Select][+][-]

1. - This is the third line
2.   This is the third line
3. - This is the fourth line
4.   This is the fourth line
5.  
6. Detailed differences:
7. Operation 1: Equal (1 lines)
8. Old index: 2, New index: 1
9. Operation 2: Add (1 lines)
10. Old index: 3, New index: 3
11. Operation 3: Delete (1 lines)
12. Old index: 2, New index: 1
13. Operation 4: Equal (1 lines)
14. Old index: 2, New index: 1
15. Operation 5: Delete (1 lines)
16. Old index: 3, New index: 3
17. Operation 6: Equal (1 lines)
18. Old index: 3, New index: 3

[avk]

Where doesn't it work?
Because it works:
...

Judge for yourself, this simple program

Code: Pascal  [Select][+][-]

1. program test;
2.  
3. {$mode objfpc}{$h+}
4.  
5. uses
6.   SysUtils, uMyersDiff;
7.  
8. var
9.   a: TStringArray = ('111',      '222','333',      '444', '555');
10.   b: TStringArray = ('111','666','222','333','777','444');
11.   I: Integer = 1;
12.   op: TDiffOp;
13.   s: string;
14. const
15.   Fmt = '%d. op: %s,'#9'old_index: %d, new_index: %d, length: %d';
16. begin
17.   for op in ComputeMyersDiff(a, b) do begin
18.     WriteStr(s, op.Kind);
19.     WriteLn(Format(Fmt, [I, s, op.OldIndex, op.NewIndex, op.Length]));
20.     Inc(I);
21.   end;
22. end.
23.  

prints as result

Code: [Select]

1. op: dEqual,  old_index: 2, new_index: 1, length: 2
2. op: dDelete, old_index: 4, new_index: 4, length: 1
3. op: dEqual,  old_index: 4, new_index: 4, length: 1
4. op: dDelete, old_index: 5, new_index: 6, length: 1
5. op: dAdd,    old_index: 4, new_index: 6, length: 1


[dbannon]

@Davo: You should have a 'LookSee' in 'StrUtils', IIRC there's an implementation hidden in there 
  FindMatchesBoyerMooreCaseSensitive

Thanks Benny. I suspect Thaddy made a mistake (yes, I know, very unlikely) when he mentioned Boyer Moore, a search model. He has quietly switched to Myers Diff, a matching thing.

But my tests of Thaddy's Myers Diff reproduce AVK's results, apparently unrelated to input. I tried Thaddy's outputting method too, also not good. The code is beyond me, lots of I, J and K variables...

Davo

[Aruna]

Hi!

Is there any simple solution for compare two text lines and highlight difference with synedit?

If no, where am I start?

Thanks!

The attached ZIP has a fully working demo using TRichMemo (not TSynEdit). You can use it to guide you if you decide to modify or extend TSynEdit yourself. Hope this will help and good luck!

[avk]

...
The code is beyond me, lots of I, J and K variables...

Davo

The code is not particularly interesting, the algorithm behind it has space complexity O((N+M)*D).
But if it is really necessary, then I believe it is quite possible to fix it.

[dbannon]

But if it is really necessary, then I believe it is quite possible to fix it.

Definitely not needed for me AVK, I have a system that works. Its just sometimes Thaddy comes up with quite nice solutions and I wondered if this was such a case. My solution, posted above is a whole lot shorter and easier to read, certainly slower but OK.

(eg, Thaddy convinced me of the superiority of "for in..." over the conventional "for i := 0...." recently. If only he was no quite so rude.)

Davo