From: Johannes Schindelin <johannes.schindelin@gmx.de>
Date: Wed, 19 Nov 2008 00:53:45 GMT
Subject: Re: [PATCH] Fix deletion of last character in levenshtein distance
Message-ID: <alpine.DEB.1.00.0811190151000.30769@pacific.mpi-cbg.de>
In-Reply-To: <20081118185326.12721.71576.stgit@arrakis.enst.fr>

Hi,

On Tue, 18 Nov 2008, Samuel Tardieu wrote:

> diff --git a/levenshtein.c b/levenshtein.c
> index db52f2c..98fea72 100644
> --- a/levenshtein.c
> +++ b/levenshtein.c
> @@ -25,7 +25,7 @@ int levenshtein(const char *string1, const char *string2,
>  					row2[j + 1] > row0[j - 1] + w)
>  				row2[j + 1] = row0[j - 1] + w;
>  			/* deletion */
> -			if (j + 1 < len2 && row2[j + 1] > row1[j + 1] + d)
> +			if (row2[j + 1] > row1[j + 1] + d)

I do not understand: does row2 have more entries than len2?  In any case, 
you will _have_ to guard against accessing elements outside the reserved 
memory.

You'll have to be more convincing to make me agree that this is a good 
change, and I am pretty certain that other people are less familiar with 
that particular part of Git's source code than me.

Ciao,
Dscho