bug-gnulib
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

new module 'fstrcmp', generic diff algorithm


From: Bruno Haible
Subject: new module 'fstrcmp', generic diff algorithm
Date: Sun, 1 Jul 2007 13:54:25 +0200
User-agent: KMail/1.5.4

Hi,

More than 4 years ago, I already proposed the addition of an 'fstrcmp'
(fuzzy string compare) module. It computes a measure of similarity
between two strings. It uses the same algorithm as GNU diff for finding
common sets of lines in two files.

The patch was rejected in [1] mainly because the code in GNU diff and the
code in GNU fstrcmp had gone out of sync.

Here is a new patch, that takes into account objections 1 and 3 from [1].
Objection 2, the desire to "have a function that returns an integer
containing the length of the common subsequence of s1 and s2", is not
directly addressed, because:
  - You can get such a function easily, with 20 lines of code: 10 lines of
    code for parametrizing diffseq.h, 10 lines of code for calling
    compareseq.
  - fstrcmp is the main bottleneck of the msgmerge program. I want to avoid
    any extra indirections and calls in this function because they cause
    an immediate slowdown of msgmerge.

The GNU diff source code has been adapted as well. Paul, you find in
the gnulib/lib repository under the branch 'diff-merge' three files:
  lib/diffseq.h   - common to GNU diff and fstrcmp
  lib/analyze.c   - used by GNU diff only
  lib/fstrcmp.c   - used by fstrcmp

For GNU diff, you just need to copy analyze.c and diffseq.h. Ideally
the copying of diffseq.h should be performed by gnulib-tool.

If you don't trust the way the merge was done, take a look at the
'diff-merge' version history of diffseq.h and analyze.c.

diffseq.h may also be useful for wdiff, a program which is getting new
attention [2], and is expected to be used also for 'msgdiff' in GNU gettext.

Objections?

Bruno

[1] http://lists.gnu.org/archive/html/bug-gnulib/2003-01/msg00109.html
[2] http://lists.gnu.org/archive/html/wdiff-bugs/2007-06/threads.html


2007-07-01  Bruno Haible  <address@hidden>

        * modules/fstrcmp: New file, from GNU gettext.
        * lib/diffseq.h: New file, from GNU gettext, extracted from
        GNU gettext's fstrcmp.c and GNU diff's analyze.c.
        * lib/fstrcmp.h: New file, from GNU gettext.
        * lib/fstrcmp.c: New file, from GNU gettext.

============================= modules/fstrcmp ============================
Description:
Fuzzy string comparison.

Files:
lib/fstrcmp.h
lib/fstrcmp.c
lib/diffseq.h

Depends-on:
lock
tls
minmax
xalloc

configure.ac:

Makefile.am:
lib_SOURCES += fstrcmp.h fstrcmp.c

Include:
"fstrcmp.h"

License:
GPL

Maintainer:
Bruno Haible

============================== lib/diffseq.h =============================
/* Analyze differences between two vectors.
   Copyright (C) 1988-1989, 1992-1995, 2001-2004, 2006 Free Software 
Foundation, Inc.

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software Foundation,
   Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.  */


/* The basic idea is to consider two vectors as similar if, when
   transforming the first vector into the second vector through a
   sequence of edits (inserts and deletes of one element each),
   this sequence is short - or equivalently, if the ordered list
   of elements that are untouched by these edits is long.  For a
   good introduction to the subject, read about the "Levenshtein
   distance" in Wikipedia.

   The basic algorithm is described in:
   "An O(ND) Difference Algorithm and its Variations", Eugene Myers,
   Algorithmica Vol. 1 No. 2, 1986, pp. 251-266;
   see especially section 4.2, which describes the variation used below.

   The basic algorithm was independently discovered as described in:
   "Algorithms for Approximate String Matching", E. Ukkonen,
   Information and Control Vol. 64, 1985, pp. 100-118.

   Unless the 'find_minimal' flag is set, this code uses the TOO_EXPENSIVE
   heuristic, by Paul Eggert, to limit the cost to O(N**1.5 log N)
   at the price of producing suboptimal output for large inputs with
   many differences.  */

/* Before including this file, you need to define:
     ELEMENT                 The element type of the vectors being compared.
     EQUAL                   A two-argument macro that tests two elements for
                             equality.
     OFFSET                  A signed integer type sufficient to hold the
                             difference between two indices. Usually
                             something like ssize_t.
     EXTRA_CONTEXT_FIELDS    Declarations of fields for 'struct context'.
     NOTE_DELETE(ctxt, xoff) Record the removal of the object xvec[xoff].
     NOTE_INSERT(ctxt, yoff) Record the insertion of the object yvec[yoff].
     USE_HEURISTIC           (Optional) Define if you want to support the
                             heuristic for large vectors.  */

/* Maximum value of type OFFSET.  */
#define OFFSET_MAX \
  ((((OFFSET)1 << (sizeof (OFFSET) * CHAR_BIT - 2)) - 1) * 2 + 1)

/* Use this to suppress gcc's `...may be used before initialized' warnings. */
#ifndef IF_LINT
# ifdef lint
#  define IF_LINT(Code) Code
# else
#  define IF_LINT(Code) /* empty */
# endif
#endif

/*
 * Context of comparison operation.
 */
struct context
{
  /* Vectors being compared.  */
  const ELEMENT *xvec;
  const ELEMENT *yvec;

  /* Extra fields.  */
  EXTRA_CONTEXT_FIELDS

  /* Vector, indexed by diagonal, containing 1 + the X coordinate of the point
     furthest along the given diagonal in the forward search of the edit
     matrix.  */
  OFFSET *fdiag;

  /* Vector, indexed by diagonal, containing the X coordinate of the point
     furthest along the given diagonal in the backward search of the edit
     matrix.  */
  OFFSET *bdiag;

  #ifdef USE_HEURISTIC
  /* This corresponds to the diff -H flag.  With this heuristic, for
     vectors with a constant small density of changes, the algorithm is
     linear in the vectors size.  */
  bool heuristic;
  #endif

  /* Edit scripts longer than this are too expensive to compute.  */
  OFFSET too_expensive;

  /* Snakes bigger than this are considered `big'.  */
  #define SNAKE_LIMIT 20
};

struct partition
{
  /* Midpoints of this partition.  */
  OFFSET xmid;
  OFFSET ymid;

  /* True if low half will be analyzed minimally.  */
  bool lo_minimal;

  /* Likewise for high half.  */
  bool hi_minimal;
};


/* Find the midpoint of the shortest edit script for a specified portion
   of the two vectors.

   Scan from the beginnings of the vectors, and simultaneously from the ends,
   doing a breadth-first search through the space of edit-sequence.
   When the two searches meet, we have found the midpoint of the shortest
   edit sequence.

   If FIND_MINIMAL is true, find the minimal edit script regardless of
   expense.  Otherwise, if the search is too expensive, use heuristics to
   stop the search and report a suboptimal answer.

   Set PART->(xmid,ymid) to the midpoint (XMID,YMID).  The diagonal number
   XMID - YMID equals the number of inserted elements minus the number
   of deleted elements (counting only elements before the midpoint).

   Set PART->lo_minimal to true iff the minimal edit script for the
   left half of the partition is known; similarly for PART->hi_minimal.

   This function assumes that the first elements of the specified portions
   of the two vectors do not match, and likewise that the last elements do not
   match.  The caller must trim matching elements from the beginning and end
   of the portions it is going to specify.

   If we return the "wrong" partitions, the worst this can do is cause
   suboptimal diff output.  It cannot cause incorrect diff output.  */

static void
diag (OFFSET xoff, OFFSET xlim, OFFSET yoff, OFFSET ylim, bool find_minimal,
      struct partition *part, struct context *ctxt)
{
  OFFSET *const fd = ctxt->fdiag;       /* Give the compiler a chance. */
  OFFSET *const bd = ctxt->bdiag;       /* Additional help for the compiler. */
  const ELEMENT *const xv = ctxt->xvec; /* Still more help for the compiler. */
  const ELEMENT *const yv = ctxt->yvec; /* And more and more . . . */
  const OFFSET dmin = xoff - ylim;      /* Minimum valid diagonal. */
  const OFFSET dmax = xlim - yoff;      /* Maximum valid diagonal. */
  const OFFSET fmid = xoff - yoff;      /* Center diagonal of top-down search. 
*/
  const OFFSET bmid = xlim - ylim;      /* Center diagonal of bottom-up search. 
*/
  OFFSET fmin = fmid;
  OFFSET fmax = fmid;           /* Limits of top-down search. */
  OFFSET bmin = bmid;
  OFFSET bmax = bmid;           /* Limits of bottom-up search. */
  OFFSET c;                     /* Cost. */
  bool odd = (fmid - bmid) & 1; /* True if southeast corner is on an odd
                                   diagonal with respect to the northwest. */

  fd[fmid] = xoff;
  bd[bmid] = xlim;

  for (c = 1;; ++c)
    {
      OFFSET d;                 /* Active diagonal. */
      bool big_snake = false;

      /* Extend the top-down search by an edit step in each diagonal. */
      if (fmin > dmin)
        fd[--fmin - 1] = -1;
      else
        ++fmin;
      if (fmax < dmax)
        fd[++fmax + 1] = -1;
      else
        --fmax;
      for (d = fmax; d >= fmin; d -= 2)
        {
          OFFSET x;
          OFFSET y;
          OFFSET oldx;
          OFFSET tlo = fd[d - 1];
          OFFSET thi = fd[d + 1];

          if (tlo >= thi)
            x = tlo + 1;
          else
            x = thi;
          oldx = x;
          y = x - d;
          while (x < xlim && y < ylim && EQUAL (xv[x], yv[y]))
            {
              ++x;
              ++y;
            }
          if (x - oldx > SNAKE_LIMIT)
            big_snake = true;
          fd[d] = x;
          if (odd && bmin <= d && d <= bmax && bd[d] <= x)
            {
              part->xmid = x;
              part->ymid = y;
              part->lo_minimal = part->hi_minimal = true;
              return;
            }
        }

      /* Similarly extend the bottom-up search.  */
      if (bmin > dmin)
        bd[--bmin - 1] = OFFSET_MAX;
      else
        ++bmin;
      if (bmax < dmax)
        bd[++bmax + 1] = OFFSET_MAX;
      else
        --bmax;
      for (d = bmax; d >= bmin; d -= 2)
        {
          OFFSET x;
          OFFSET y;
          OFFSET oldx;
          OFFSET tlo = bd[d - 1];
          OFFSET thi = bd[d + 1];

          if (tlo < thi)
            x = tlo;
          else
            x = thi - 1;
          oldx = x;
          y = x - d;
          while (x > xoff && y > yoff && EQUAL (xv[x - 1], yv[y - 1]))
            {
              --x;
              --y;
            }
          if (oldx - x > SNAKE_LIMIT)
            big_snake = true;
          bd[d] = x;
          if (!odd && fmin <= d && d <= fmax && x <= fd[d])
            {
              part->xmid = x;
              part->ymid = y;
              part->lo_minimal = part->hi_minimal = true;
              return;
            }
        }

      if (find_minimal)
        continue;

#ifdef USE_HEURISTIC
      /* Heuristic: check occasionally for a diagonal that has made lots
         of progress compared with the edit distance.  If we have any
         such, find the one that has made the most progress and return it
         as if it had succeeded.

         With this heuristic, for vectors with a constant small density
         of changes, the algorithm is linear in the vector size.  */

      if (c > 200 && big_snake && ctxt->heuristic)
        {
          OFFSET best;

          best = 0;
          for (d = fmax; d >= fmin; d -= 2)
            {
              OFFSET dd = d - fmid;
              OFFSET x = fd[d];
              OFFSET y = x - d;
              OFFSET v = (x - xoff) * 2 - dd;

              if (v > 12 * (c + (dd < 0 ? -dd : dd)))
                {
                  if (v > best
                      && xoff + SNAKE_LIMIT <= x && x < xlim
                      && yoff + SNAKE_LIMIT <= y && y < ylim)
                    {
                      /* We have a good enough best diagonal; now insist
                         that it end with a significant snake.  */
                      int k;

                      for (k = 1; EQUAL (xv[x - k], yv[y - k]); k++)
                        if (k == SNAKE_LIMIT)
                          {
                            best = v;
                            part->xmid = x;
                            part->ymid = y;
                            break;
                          }
                    }
                }
            }
          if (best > 0)
            {
              part->lo_minimal = true;
              part->hi_minimal = false;
              return;
            }

          best = 0;
          for (d = bmax; d >= bmin; d -= 2)
            {
              OFFSET dd = d - bmid;
              OFFSET x = bd[d];
              OFFSET y = x - d;
              OFFSET v = (xlim - x) * 2 + dd;

              if (v > 12 * (c + (dd < 0 ? -dd : dd)))
                {
                  if (v > best
                      && xoff < x && x <= xlim - SNAKE_LIMIT
                      && yoff < y && y <= ylim - SNAKE_LIMIT)
                    {
                      /* We have a good enough best diagonal; now insist
                         that it end with a significant snake.  */
                      int k;

                      for (k = 0; EQUAL (xv[x + k], yv[y + k]); k++)
                        if (k == SNAKE_LIMIT - 1)
                          {
                            best = v;
                            part->xmid = x;
                            part->ymid = y;
                            break;
                          }
                    }
                }
            }
          if (best > 0)
            {
              part->lo_minimal = false;
              part->hi_minimal = true;
              return;
            }
        }
#endif /* USE_HEURISTIC */

      /* Heuristic: if we've gone well beyond the call of duty, give up
         and report halfway between our best results so far.  */
      if (c >= ctxt->too_expensive)
        {
          OFFSET fxybest;
          OFFSET fxbest IF_LINT (= 0);
          OFFSET bxybest;
          OFFSET bxbest IF_LINT (= 0);

          /* Find forward diagonal that maximizes X + Y.  */
          fxybest = -1;
          for (d = fmax; d >= fmin; d -= 2)
            {
              OFFSET x;
              OFFSET y;

              x = MIN (fd[d], xlim);
              y = x - d;
              if (ylim < y)
                {
                  x = ylim + d;
                  y = ylim;
                }
              if (fxybest < x + y)
                {
                  fxybest = x + y;
                  fxbest = x;
                }
            }

          /* Find backward diagonal that minimizes X + Y.  */
          bxybest = OFFSET_MAX;
          for (d = bmax; d >= bmin; d -= 2)
            {
              OFFSET x;
              OFFSET y;

              x = MAX (xoff, bd[d]);
              y = x - d;
              if (y < yoff)
                {
                  x = yoff + d;
                  y = yoff;
                }
              if (x + y < bxybest)
                {
                  bxybest = x + y;
                  bxbest = x;
                }
            }

          /* Use the better of the two diagonals.  */
          if ((xlim + ylim) - bxybest < fxybest - (xoff + yoff))
            {
              part->xmid = fxbest;
              part->ymid = fxybest - fxbest;
              part->lo_minimal = true;
              part->hi_minimal = false;
            }
          else
            {
              part->xmid = bxbest;
              part->ymid = bxybest - bxbest;
              part->lo_minimal = false;
              part->hi_minimal = true;
            }
          return;
        }
    }
}


/* Compare in detail contiguous subsequences of the two vectors
   which are known, as a whole, to match each other.

   The subsequence of vector 0 is [XOFF, XLIM) and likewise for vector 1.

   Note that XLIM, YLIM are exclusive bounds.  All indices into the vectors
   are origin-0.

   If FIND_MINIMAL, find a minimal difference no matter how
   expensive it is.

   The results are recorded by invoking NOTE_DELETE and NOTE_INSERT.  */

static void
compareseq (OFFSET xoff, OFFSET xlim, OFFSET yoff, OFFSET ylim,
            bool find_minimal, struct context *ctxt)
{
  const ELEMENT *const xv = ctxt->xvec; /* Help the compiler.  */
  const ELEMENT *const yv = ctxt->yvec;

  /* Slide down the bottom initial diagonal. */
  while (xoff < xlim && yoff < ylim && EQUAL (xv[xoff], yv[yoff]))
    {
      ++xoff;
      ++yoff;
    }

  /* Slide up the top initial diagonal. */
  while (xlim > xoff && ylim > yoff && EQUAL (xv[xlim - 1], yv[ylim - 1]))
    {
      --xlim;
      --ylim;
    }

  /* Handle simple cases. */
  if (xoff == xlim)
    while (yoff < ylim)
      {
        NOTE_INSERT (ctxt, yoff);
        yoff++;
      }
  else if (yoff == ylim)
    while (xoff < xlim)
      {
        NOTE_DELETE (ctxt, xoff);
        xoff++;
      }
  else
    {
      struct partition part;

      /* Find a point of correspondence in the middle of the vectors.  */
      diag (xoff, xlim, yoff, ylim, find_minimal, &part, ctxt);

      /* Use the partitions to split this problem into subproblems.  */
      compareseq (xoff, part.xmid, yoff, part.ymid, part.lo_minimal, ctxt);
      compareseq (part.xmid, xlim, part.ymid, ylim, part.hi_minimal, ctxt);
    }
}

#undef ELEMENT
#undef EQUAL
#undef OFFSET
#undef EXTRA_CONTEXT_FIELDS
#undef NOTE_DELETE
#undef NOTE_INSERT
#undef USE_HEURISTIC
#undef OFFSET_MAX
============================== lib/fstrcmp.h =============================
/* Fuzzy string comparison.
   Copyright (C) 1995, 2000, 2002-2003, 2006 Free Software Foundation, Inc.

   This file was written by Peter Miller <address@hidden>

This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2, or (at your option)
any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software Foundation,
Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.  */

#ifndef _FSTRCMP_H
#define _FSTRCMP_H

#ifdef __cplusplus
extern "C" {
#endif

/* Fuzzy compare of S1 and S2.  Return a measure for the similarity of S1
   and S1.  The higher the result, the more similar the strings are.  */
extern double fstrcmp (const char *s1, const char *s2);

#ifdef __cplusplus
}
#endif

#endif
============================== lib/fstrcmp.c =============================
/* Functions to make fuzzy comparisons between strings
   Copyright (C) 1988-1989, 1992-1993, 1995, 2001-2003, 2006 Free Software 
Foundation, Inc.

   This program is free software; you can redistribute it and/or modify
   it under the terms of the GNU General Public License as published by
   the Free Software Foundation; either version 2, or (at your option)
   any later version.

   This program is distributed in the hope that it will be useful,
   but WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
   GNU General Public License for more details.

   You should have received a copy of the GNU General Public License
   along with this program; if not, write to the Free Software Foundation,
   Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.


   Derived from GNU diff 2.7, analyze.c et al.

   The basic idea is to consider two vectors as similar if, when
   transforming the first vector into the second vector through a
   sequence of edits (inserts and deletes of one element each),
   this sequence is short - or equivalently, if the ordered list
   of elements that are untouched by these edits is long.  For a
   good introduction to the subject, read about the "Levenshtein
   distance" in Wikipedia.

   The basic algorithm is described in:
   "An O(ND) Difference Algorithm and its Variations", Eugene Myers,
   Algorithmica Vol. 1 No. 2, 1986, pp. 251-266;
   see especially section 4.2, which describes the variation used below.

   The basic algorithm was independently discovered as described in:
   "Algorithms for Approximate String Matching", E. Ukkonen,
   Information and Control Vol. 64, 1985, pp. 100-118.

   Unless the 'find_minimal' flag is set, this code uses the TOO_EXPENSIVE
   heuristic, by Paul Eggert, to limit the cost to O(N**1.5 log N)
   at the price of producing suboptimal output for large inputs with
   many differences.  */

#include <config.h>

/* Specification.  */
#include "fstrcmp.h"

#include <string.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <limits.h>

#include "lock.h"
#include "tls.h"
#include "minmax.h"
#include "xalloc.h"

#ifndef uintptr_t
# define uintptr_t unsigned long
#endif


#define ELEMENT char
#define EQUAL(x,y) ((x) == (y))
#define OFFSET int
#define EXTRA_CONTEXT_FIELDS \
  /* The number of elements inserted or deleted. */ \
  int xvec_edit_count; \
  int yvec_edit_count;
#define NOTE_DELETE(ctxt, xoff) ctxt->xvec_edit_count++
#define NOTE_INSERT(ctxt, yoff) ctxt->yvec_edit_count++
/* We don't need USE_HEURISTIC, since it is unlikely in typical uses of
   fstrcmp().  */
#include "diffseq.h"


/* Because fstrcmp is typically called multiple times, attempt to minimize
   the number of memory allocations performed.  Thus, let a call reuse the
   memory already allocated by the previous call, if it is sufficient.
   To make it multithread-safe, without need for a lock that protects the
   already allocated memory, store the allocated memory per thread.  Free
   it only when the thread exits.  */

static gl_tls_key_t buffer_key; /* TLS key for a 'int *' */
static gl_tls_key_t bufmax_key; /* TLS key for a 'size_t' */

static void
keys_init (void)
{
  gl_tls_key_init (buffer_key, free);
  gl_tls_key_init (bufmax_key, NULL);
  /* The per-thread initial values are NULL and 0, respectively.  */
}

/* Ensure that keys_init is called once only.  */
gl_once_define(static, keys_init_once)


/* NAME
        fstrcmp - fuzzy string compare

   SYNOPSIS
        double fstrcmp(const char *, const char *);

   DESCRIPTION
        The fstrcmp function may be used to compare two string for
        similarity.  It is very useful in reducing "cascade" or
        "secondary" errors in compilers or other situations where
        symbol tables occur.

   RETURNS
        double; 0 if the strings are entirly dissimilar, 1 if the
        strings are identical, and a number in between if they are
        similar.  */

double
fstrcmp (const char *string1, const char *string2)
{
  struct context ctxt;
  int xvec_length;
  int yvec_length;
  int i;

  size_t fdiag_len;
  int *buffer;
  size_t bufmax;

  /* set the info for each string.  */
  ctxt.xvec = string1;
  xvec_length = strlen (string1);
  ctxt.yvec = string2;
  yvec_length = strlen (string2);

  /* short-circuit obvious comparisons */
  if (xvec_length == 0 && yvec_length == 0)
    return 1.0;
  if (xvec_length == 0 || yvec_length == 0)
    return 0.0;

  /* Set TOO_EXPENSIVE to be approximate square root of input size,
     bounded below by 256.  */
  ctxt.too_expensive = 1;
  for (i = xvec_length + yvec_length;
       i != 0;
       i >>= 2)
    ctxt.too_expensive <<= 1;
  if (ctxt.too_expensive < 256)
    ctxt.too_expensive = 256;

  /* Allocate memory for fdiag and bdiag from a thread-local pool.  */
  fdiag_len = xvec_length + yvec_length + 3;
  gl_once (keys_init_once, keys_init);
  buffer = (int *) gl_tls_get (buffer_key);
  bufmax = (size_t) (uintptr_t) gl_tls_get (bufmax_key);
  if (fdiag_len > bufmax)
    {
      /* Need more memory.  */
      bufmax = 2 * bufmax;
      if (fdiag_len > bufmax)
        bufmax = fdiag_len;
      /* Calling xrealloc would be a waste: buffer's contents does not need
         to be preserved.  */
      if (buffer != NULL)
        free (buffer);
      buffer = (int *) xnmalloc (bufmax, 2 * sizeof (int));
      gl_tls_set (buffer_key, buffer);
      gl_tls_set (bufmax_key, (void *) (uintptr_t) bufmax);
    }
  ctxt.fdiag = buffer + yvec_length + 1;
  ctxt.bdiag = ctxt.fdiag + fdiag_len;

  /* Now do the main comparison algorithm */
  ctxt.xvec_edit_count = 0;
  ctxt.yvec_edit_count = 0;
  compareseq (0, xvec_length, 0, yvec_length, 0,
              &ctxt);

  /* The result is
        ((number of chars in common) / (average length of the strings)).
     This is admittedly biased towards finding that the strings are
     similar, however it does produce meaningful results.  */
  return ((double) (xvec_length + yvec_length
                    - ctxt.yvec_edit_count - ctxt.xvec_edit_count)
          / (xvec_length + yvec_length));
}
==========================================================================





reply via email to

[Prev in Thread] Current Thread [Next in Thread]