Re: Anything more powerful than diff's -I...?

[Bill Rugolsky Jr.]

On Fri, Sep 14, 2007 at 01:05:27PM -0700, Paul Eggert wrote:
> Paul Smith <address@hidden> writes:
> 
> > I guess I could do something like remove all the const's from the file
> > then compare that, and if it was the same I would skip it and if it was
> > different I could run diff on the original files.
> 
> I find this strategy useful in other circumstances too.  Typically I
> don't bother to run diff on the original files; I just run it on their
> transformed versions.  It'd be nice if "diff" supported this sort of
> thing better.  (I don't know of other tools that do.)

Transform "don't care input patterns" into a unique whitespace pattern on
input [*], do an ignore-whitespace compare [diff -b or --ignore-all-space],
and transform it back on output.

[*] You can encode a list of patterns using a variable-length "binary"
    (e.g., Huffman-like) representation via the little-used vertical tab
    "\v" and form-feed "\f" patterns.  Be sure to start it off with "\v",
    since form-feed sometimes appears in source code for pagination.

I once had a hackish (sh+awk) word-oriented diff script that applied this
and other transformations, but I lost the script long ago (on QIC tape).
Something similar is easy to reproduce though, and there are word-oriented
diff utils available today.

Regards,

        Bill Rugolsky