Re: sdiff Enhancement

[Bruno Haible]

Hi,

Grant Stevens wrote:
> $ diff -W 43 -Y file1 file2 # (-Y = -y option, enhanced for tighter matching)
> line1                   line1
>                      >  line2
> close but not exact  |  close but inexact
>                      >  line3
>                      >  line4
> another not exact    |  another inexact
> last line               last line

Two remarks:
1) I think your "tighter matching" mode would also be useful for unified
   diff (diff -u). The problem with unified diffs is when a 30-line chunk
   is reindented or otherwise slightly changed, the resulting output is
   not understandable any more. Your patch has the potential to fix this.
2) I don't know how you detect similarity between
   "close but not exact" and "close but inexact" - your functions
   insert_matchup, find_best_match1, figure_score, matching_chars lack
   a comment describing what they do -, but a good metrics for this kind
   of question is the Levenshtein distance between strings [1], which is
   implemented in the fstrcmp() function in gnulib [2].

Bruno

[1] http://en.wikipedia.org/wiki/Levenshtein_distance
[2] http://www.gnu.org/software/gnulib/MODULES.html#module=fstrcmp