Diff to actually totally ignore whitespace

Question

I am looking for an application that will diff two files, and actually ignore all whitespace, so for example:

class foo { 
  bar
  spaz 
}

is equally equivilant to

class foo{bar spaz}

or, as well

classfoo { 
  barspaz}

but NOT

classfoo { 
  spazspaz
}

i.e. it would show me that spaz in the previous example has taken the place of bar in any of the other examples. It only needs to compare 2 files.

• It can be a windows or linux/unix/posix-compatible utility
• I've tried the lin/unix diff -w command, it only ignores whitespace if the difference per-line is whitespace. I don't see an option to totally ignore whitespace.
• I also tried UECompare or Ultracompare, a non-free comparison utility for windows.

Answer 1

Are you looking for something like the tr command? Here are the manpages. It's included with msysgit, cygwin, and gnuwin32 tools as far as I can tell.

So you can remove all the whitespace prior to diffing by doing something like:

tr --delete '[:space:]' <filename.txt

You can then feed the output of that command to diff and have it work without having any whitespace.

For example, I have a file named HelloWorldApp.java. Let me show you how tr processes it:

C:\temp>cat HelloWorldApp.java
class HelloWorldApp {
    public static void main(String[] args) {
        System.out.println("Hello World!"); // Display the string.
    }
}
C:\temp>tr -d '[:space:]' <HelloWorldApp.java
classHelloWorldApp{publicstaticvoidmain(String[]args){System.out.println("HelloWorld!");//Displaythestring.}}

Answer 2

Despite being line oriented, diff could be made to do what's being requested. Roughly, you could do the diff -w and then post process the diff output. It could look at the diff output and join various pairs of lines to see if they (now) match the line in the other file.

Seems like it might be O(n^2) or other nasty thing, but it would still be very helpful if it limited it self to (say) no join, join 2 lines, join 3 lines on file 1 X no join, join 2 lines, join 3 lines on file 2 on a moving "window" of the diff output. (where X is ~ "cross product")

In fact, seems like a job that a Perl diff output post-processing script could do with few hours of work (depending on your (Perl) programming skill)

Maybe when I have some task that I don't really want to be doing at work ... :-)