[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [PATCH] Speed-up of libsvn_diff using token counts

From: Morten Kloster <morklo_at_gmail.com>
Date: Wed, 18 May 2011 19:31:54 +0200

Thanks, Daniel.

Johan:
I've found your notes trunk/notes/diff-optimizations.txt. As you may
have realized, my patch is a slightly different approach to the
problem you discuss in section I.5, "Discarding non-matching
lines before running the LCS (Longest Common Subsequence)
algorithm." - rather than discarding the lines before calling LCS, I
create count information that lets LCS skip the lines without
counting them as (avoidable) changes. I believe my approach
is easier to implement (not least now, when it's already done ;)
and has less overhead, although the potential speed-up is
somewhat less:

LCS has running time of O( N + P D ), where N is #tokens in
longest source, P is number of mismatches in smallest source
and D is total number of mismatches in both sources.
Discarding the non-matching tokens would replace both P and D
by their discarded versions P' and D', giving a potential speed-up
of the SQUARE of the inverse of the fraction of changed tokens
that are not unique to one source, whereas my approach
effectively only replaces P by P', not D, so you only get one
power, not two (at least I think so; I haven't checked carefully).

There is another possible enhancement that is somewhat
complementary to the one I've implemented: If the number of
changed tokens is smaller than the number of uniquely
matching tokens, and the changes are grouped in well-
separated sections, then whenever you have found more
unique matches than the number of changes you have "used",
you can "restart" the problem from that point: you are
guaranteed that there is no better match for the earlier
part of either source. This reduces the running time from
O( N + P' D ) to something like O( N + sum(P'n Dn)), I think,
where the sum is over the different changed sections. I
haven't worked out the details on how to best implement
this approach, but it should work well with my initial patch.

Morten

On Wed, May 18, 2011 at 11:01 AM, Daniel Shahaf <d.s_at_daniel.shahaf.name> wrote:
> Please don't mix whitespace changes with functional changes.
>
> I'm attaching a version of the patch re-generated with -x-pw.
>
> [[[
> Index: subversion/libsvn_diff/diff.c
> ===================================================================
> --- subversion/libsvn_diff/diff.c       (revision 1104340)
> +++ subversion/libsvn_diff/diff.c       (working copy)
> @@ -105,6 +105,10 @@ svn_diff_diff_2(svn_diff_t **diff,
>  {
>   svn_diff__tree_t *tree;
>   svn_diff__position_t *position_list[2];
> +  apr_uint32_t num_tokens;
> +  apr_uint32_t token_index;
> +  svn_diff__position_t *current;
> +  apr_uint32_t *token_counts[2];
>   svn_diff_datasource_e datasource[] = {svn_diff_datasource_original,
>                                         svn_diff_datasource_modified};
>   svn_diff__lcs_t *lcs;
> @@ -138,6 +142,8 @@ svn_diff_diff_2(svn_diff_t **diff,
>                                prefix_lines,
>                                subpool));
>
> +  num_tokens = svn_diff__get_node_count(tree);
> +
>   /* The cool part is that we don't need the tokens anymore.
>    * Allow the app to clean them up if it wants to.
>    */
> @@ -147,8 +153,35 @@ svn_diff_diff_2(svn_diff_t **diff,
>   /* We don't need the nodes in the tree either anymore, nor the tree itself */
>   svn_pool_destroy(treepool);
>
> +  token_counts[0] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[1] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  for (token_index = 0; token_index < num_tokens; token_index++)
> +    token_counts[0][token_index] = token_counts[1][token_index] = 0;
> +
> +  current = position_list[0];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[0][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[0]);
> +    }
> +  current = position_list[1];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[1][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[1]);
> +    }
> +
>   /* Get the lcs */
> -  lcs = svn_diff__lcs(position_list[0], position_list[1], prefix_lines,
> +  lcs = svn_diff__lcs(position_list[0], position_list[1], token_counts[0],
> +                         token_counts[1], num_tokens, prefix_lines,
>                       suffix_lines, subpool);
>
>   /* Produce the diff */
> Index: subversion/libsvn_diff/diff.h
> ===================================================================
> --- subversion/libsvn_diff/diff.h       (revision 1104340)
> +++ subversion/libsvn_diff/diff.h       (working copy)
> @@ -62,7 +62,7 @@ struct svn_diff_t {
>  struct svn_diff__position_t
>  {
>   svn_diff__position_t *next;
> -  svn_diff__node_t     *node;
> +  apr_int32_t          node;
>   apr_off_t             offset;
>  };
>
> @@ -91,12 +91,21 @@ typedef enum svn_diff__normalize_state_t
>  svn_diff__lcs_t *
>  svn_diff__lcs(svn_diff__position_t *position_list1, /* pointer to tail (ring) */
>               svn_diff__position_t *position_list2, /* pointer to tail (ring) */
> +              apr_uint32_t *token_counts_list1, /* array of counts */
> +              apr_uint32_t *token_counts_list2, /* array of counts */
> +              apr_uint32_t num_tokens,
>               apr_off_t prefix_lines,
>               apr_off_t suffix_lines,
>               apr_pool_t *pool);
>
>
>  /*
> + * Returns number of tokens in a tree
> + */
> +apr_uint32_t
> +svn_diff__get_node_count(svn_diff__tree_t *tree);
> +
> +/*
>  * Support functions to build a tree of token positions
>  */
>  void
> @@ -128,6 +137,7 @@ void
>  svn_diff__resolve_conflict(svn_diff_t *hunk,
>                            svn_diff__position_t **position_list1,
>                            svn_diff__position_t **position_list2,
> +                           apr_uint32_t num_tokens,
>                            apr_pool_t *pool);
>
>
> Index: subversion/libsvn_diff/token.c
> ===================================================================
> --- subversion/libsvn_diff/token.c      (revision 1104340)
> +++ subversion/libsvn_diff/token.c      (working copy)
> @@ -46,6 +46,7 @@ struct svn_diff__node_t
>   svn_diff__node_t     *right;
>
>   apr_uint32_t          hash;
> +  apr_int32_t           index;
>   void                 *token;
>  };
>
> @@ -53,10 +54,20 @@ struct svn_diff__tree_t
>  {
>   svn_diff__node_t     *root[SVN_DIFF__HASH_SIZE];
>   apr_pool_t           *pool;
> +  apr_uint32_t          node_count;
>  };
>
>
>  /*
> + * Returns number of tokens in a tree
> + */
> +apr_uint32_t
> +svn_diff__get_node_count(svn_diff__tree_t *tree)
> +{
> +  return tree->node_count;
> +}
> +
> +/*
>  * Support functions to build a tree of token positions
>  */
>
> @@ -65,6 +76,7 @@ svn_diff__tree_create(svn_diff__tree_t **tree, apr
>  {
>   *tree = apr_pcalloc(pool, sizeof(**tree));
>   (*tree)->pool = pool;
> +  (*tree)->node_count = 0;
>  }
>
>
> @@ -122,6 +134,7 @@ tree_insert_token(svn_diff__node_t **node, svn_dif
>   new_node->right = NULL;
>   new_node->hash = hash;
>   new_node->token = token;
> +  new_node->index = tree->node_count++;
>
>   *node = *node_ref = new_node;
>
> @@ -168,7 +181,7 @@ svn_diff__get_tokens(svn_diff__position_t **positi
>       /* Create a new position */
>       position = apr_palloc(pool, sizeof(*position));
>       position->next = NULL;
> -      position->node = node;
> +      position->node = node->index;
>       position->offset = offset;
>
>       *position_ref = position;
> Index: subversion/libsvn_diff/lcs.c
> ===================================================================
> --- subversion/libsvn_diff/lcs.c        (revision 1104340)
> +++ subversion/libsvn_diff/lcs.c        (working copy)
> @@ -51,6 +51,7 @@ static APR_INLINE void
>  svn_diff__snake(apr_off_t k,
>                 svn_diff__snake_t *fp,
>                 int idx,
> +                               apr_uint32_t *token_counts[2],
>                 svn_diff__lcs_t **freelist,
>                 apr_pool_t *pool)
>  {
> @@ -92,6 +93,11 @@ svn_diff__snake(apr_off_t k,
>     }
>
>
> +  if (previous_lcs)
> +    {
> +      previous_lcs->refcount++;
> +    }
> +
>   /* ### Optimization, skip all positions that don't have matchpoints
>    * ### anyway. Beware of the sentinel, don't skip it!
>    */
> @@ -99,6 +105,8 @@ svn_diff__snake(apr_off_t k,
>   position[0] = start_position[0];
>   position[1] = start_position[1];
>
> +  while (1)
> +    {
>   while (position[0]->node == position[1]->node)
>     {
>       position[0] = position[0]->next;
> @@ -122,18 +130,19 @@ svn_diff__snake(apr_off_t k,
>       lcs->length = position[1]->offset - start_position[1]->offset;
>       lcs->next = previous_lcs;
>       lcs->refcount = 1;
> -      fp[k].lcs = lcs;
> +          previous_lcs = lcs;
> +             start_position[0] = position[0];
> +             start_position[1] = position[1];
>     }
> +      if (position[0]->node >= 0 && token_counts[1][position[0]->node] == 0)
> +        start_position[0] = position[0] = position[0]->next;
> +      else if (position[1]->node >= 0 && token_counts[0][position[1]->node] == 0)
> +        start_position[1] = position[1] = position[1]->next;
>   else
> -    {
> -      fp[k].lcs = previous_lcs;
> +        break;
>     }
>
> -  if (previous_lcs)
> -    {
> -      previous_lcs->refcount++;
> -    }
> -
> +  fp[k].lcs = previous_lcs;
>   fp[k].position[0] = position[0];
>   fp[k].position[1] = position[1];
>
> @@ -188,12 +197,18 @@ prepend_lcs(svn_diff__lcs_t *lcs, apr_off_t lines,
>  svn_diff__lcs_t *
>  svn_diff__lcs(svn_diff__position_t *position_list1, /* pointer to tail (ring) */
>               svn_diff__position_t *position_list2, /* pointer to tail (ring) */
> +              apr_uint32_t *token_counts_list1, /* array of counts */
> +              apr_uint32_t *token_counts_list2, /* array of counts */
> +              apr_uint32_t num_tokens,
>               apr_off_t prefix_lines,
>               apr_off_t suffix_lines,
>               apr_pool_t *pool)
>  {
>   int idx;
>   apr_off_t length[2];
> +  apr_uint32_t *token_counts[2];
> +  apr_uint32_t unique_count[2];
> +  apr_uint32_t token_index;
>   svn_diff__snake_t *fp;
>   apr_off_t d;
>   apr_off_t k;
> @@ -231,10 +246,19 @@ svn_diff__lcs(svn_diff__position_t *position_list1
>       return lcs;
>     }
>
> +  unique_count[1] = unique_count[0] = 0;
> +  for (token_index = 0; token_index < num_tokens; token_index++)
> +    {
> +      if (token_counts_list1[token_index] == 0)
> +        unique_count[1] += token_counts_list2[token_index];
> +      if (token_counts_list2[token_index] == 0)
> +        unique_count[0] += token_counts_list1[token_index];
> +    }
> +
>   /* Calculate length of both sequences to be compared */
>   length[0] = position_list1->offset - position_list1->next->offset + 1;
>   length[1] = position_list2->offset - position_list2->next->offset + 1;
> -  idx = length[0] > length[1] ? 1 : 0;
> +  idx = length[0] - unique_count[0] > length[1] - unique_count[1] ? 1 : 0;
>
>   /* strikerXXX: here we allocate the furthest point array, which is
>    * strikerXXX: sized M + N + 3 (!)
> @@ -246,18 +270,20 @@ svn_diff__lcs(svn_diff__position_t *position_list1
>   sentinel_position[idx].next = position_list1->next;
>   position_list1->next = &sentinel_position[idx];
>   sentinel_position[idx].offset = position_list1->offset + 1;
> +  token_counts[idx] = token_counts_list1;
>
>   sentinel_position[abs(1 - idx)].next = position_list2->next;
>   position_list2->next = &sentinel_position[abs(1 - idx)];
>   sentinel_position[abs(1 - idx)].offset = position_list2->offset + 1;
> +  token_counts[abs(1 - idx)] = token_counts_list2;
>
> -  /* These are never dereferenced, only compared by value, so we
> -   * can safely fake these up and the void* cast is OK.
> +  /* Negative indices will not be used elsewhere
>    */
> -  sentinel_position[0].node = (void*)&sentinel_position[0];
> -  sentinel_position[1].node = (void*)&sentinel_position[1];
> +  sentinel_position[0].node = -1;
> +  sentinel_position[1].node = -2;
>
> -  d = length[abs(1 - idx)] - length[idx];
> +  d = length[abs(1 - idx)] - unique_count[abs(1 - idx)]
> +      - (length[idx] - unique_count[idx]);
>
>   /* k = -1 will be the first to be used to get previous
>    * position information from, make sure it holds sane
> @@ -272,12 +298,12 @@ svn_diff__lcs(svn_diff__position_t *position_list1
>       /* Forward */
>       for (k = -p; k < d; k++)
>         {
> -          svn_diff__snake(k, fp, idx, &lcs_freelist, pool);
> +          svn_diff__snake(k, fp, idx, token_counts, &lcs_freelist, pool);
>         }
>
>       for (k = d + p; k >= d; k--)
>         {
> -          svn_diff__snake(k, fp, idx, &lcs_freelist, pool);
> +          svn_diff__snake(k, fp, idx, token_counts, &lcs_freelist, pool);
>         }
>
>       p++;
> Index: subversion/libsvn_diff/diff3.c
> ===================================================================
> --- subversion/libsvn_diff/diff3.c      (revision 1104340)
> +++ subversion/libsvn_diff/diff3.c      (working copy)
> @@ -38,6 +38,7 @@ void
>  svn_diff__resolve_conflict(svn_diff_t *hunk,
>                            svn_diff__position_t **position_list1,
>                            svn_diff__position_t **position_list2,
> +                           apr_uint32_t num_tokens,
>                            apr_pool_t *pool)
>  {
>     apr_off_t modified_start = hunk->modified_start + 1;
> @@ -47,6 +48,9 @@ svn_diff__resolve_conflict(svn_diff_t *hunk,
>     apr_off_t latest_length = hunk->latest_length;
>     svn_diff__position_t *start_position[2];
>     svn_diff__position_t *position[2];
> +  apr_uint32_t token_index;
> +  svn_diff__position_t *current;
> +  apr_uint32_t *token_counts[2];
>     svn_diff__lcs_t *lcs = NULL;
>     svn_diff__lcs_t **lcs_ref = &lcs;
>     svn_diff_t **diff_ref = &hunk->resolved_diff;
> @@ -173,9 +177,29 @@ svn_diff__resolve_conflict(svn_diff_t *hunk,
>         position[1]->next = start_position[1];
>       }
>
> -    *lcs_ref = svn_diff__lcs(position[0], position[1], 0, 0,
> -                             subpool);
> +  token_counts[0] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[1] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  for (token_index = 0; token_index < num_tokens; token_index++)
> +    token_counts[0][token_index] = token_counts[1][token_index] = 0;
>
> +  current = position[0];
> +  do
> +    {
> +      token_counts[0][current->node]++;
> +      current = current->next;
> +    }
> +  while (current != position[0]);
> +  current = position[1];
> +  do
> +    {
> +      token_counts[1][current->node]++;
> +      current = current->next;
> +    }
> +  while (current != position[1]);
> +
> +  *lcs_ref = svn_diff__lcs(position[0], position[1], token_counts[0],
> +                           token_counts[1], num_tokens, 0, 0, subpool);
> +
>     /* Fix up the EOF lcs element in case one of
>      * the two sequences was NULL.
>      */
> @@ -251,6 +275,10 @@ svn_diff_diff3_2(svn_diff_t **diff,
>  {
>   svn_diff__tree_t *tree;
>   svn_diff__position_t *position_list[3];
> +  apr_uint32_t num_tokens;
> +  apr_uint32_t token_index;
> +  svn_diff__position_t *current;
> +  apr_uint32_t *token_counts[3];
>   svn_diff_datasource_e datasource[] = {svn_diff_datasource_original,
>                                         svn_diff_datasource_modified,
>                                         svn_diff_datasource_latest};
> @@ -292,6 +320,8 @@ svn_diff_diff3_2(svn_diff_t **diff,
>                                prefix_lines,
>                                subpool));
>
> +  num_tokens = svn_diff__get_node_count(tree);
> +
>   /* Get rid of the tokens, we don't need them to calc the diff */
>   if (vtable->token_discard_all != NULL)
>     vtable->token_discard_all(diff_baton);
> @@ -299,10 +329,52 @@ svn_diff_diff3_2(svn_diff_t **diff,
>   /* We don't need the nodes in the tree either anymore, nor the tree itself */
>   svn_pool_destroy(treepool);
>
> +  token_counts[0] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[1] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[2] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  for (token_index = 0; token_index < num_tokens; token_index++)
> +    {
> +      token_counts[0][token_index] = token_counts[1][token_index] = 0;
> +      token_counts[2][token_index] = 0;
> +    }
> +
> +  current = position_list[0];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[0][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[0]);
> +    }
> +  current = position_list[1];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[1][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[1]);
> +    }
> +  current = position_list[2];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[2][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[2]);
> +    }
> +
>   /* Get the lcs for original-modified and original-latest */
> -  lcs_om = svn_diff__lcs(position_list[0], position_list[1], prefix_lines,
> +  lcs_om = svn_diff__lcs(position_list[0], position_list[1], token_counts[0],
> +                         token_counts[1], num_tokens, prefix_lines,
>                          suffix_lines, subpool);
> -  lcs_ol = svn_diff__lcs(position_list[0], position_list[2], prefix_lines,
> +  lcs_ol = svn_diff__lcs(position_list[0], position_list[2], token_counts[0],
> +                         token_counts[2], num_tokens, prefix_lines,
>                          suffix_lines, subpool);
>
>   /* Produce a merged diff */
> @@ -435,6 +507,7 @@ svn_diff_diff3_2(svn_diff_t **diff,
>                 svn_diff__resolve_conflict(*diff_ref,
>                                            &position_list[1],
>                                            &position_list[2],
> +                                           num_tokens,
>                                            pool);
>               }
>             else if (is_modified)
> Index: subversion/libsvn_diff/diff4.c
> ===================================================================
> --- subversion/libsvn_diff/diff4.c      (revision 1104340)
> +++ subversion/libsvn_diff/diff4.c      (working copy)
> @@ -173,6 +173,10 @@ svn_diff_diff4_2(svn_diff_t **diff,
>  {
>   svn_diff__tree_t *tree;
>   svn_diff__position_t *position_list[4];
> +  apr_uint32_t num_tokens;
> +  apr_uint32_t token_index;
> +  svn_diff__position_t *current;
> +  apr_uint32_t *token_counts[4];
>   svn_diff_datasource_e datasource[] = {svn_diff_datasource_original,
>                                         svn_diff_datasource_modified,
>                                         svn_diff_datasource_latest,
> @@ -227,6 +231,8 @@ svn_diff_diff4_2(svn_diff_t **diff,
>                                prefix_lines,
>                                subpool2));
>
> +  num_tokens = svn_diff__get_node_count(tree);
> +
>   /* Get rid of the tokens, we don't need them to calc the diff */
>   if (vtable->token_discard_all != NULL)
>     vtable->token_discard_all(diff_baton);
> @@ -234,8 +240,61 @@ svn_diff_diff4_2(svn_diff_t **diff,
>   /* We don't need the nodes in the tree either anymore, nor the tree itself */
>   svn_pool_clear(subpool3);
>
> +  token_counts[0] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[1] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[2] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  token_counts[3] = apr_palloc(subpool, num_tokens * sizeof(**token_counts));
> +  for (token_index = 0; token_index < num_tokens; token_index++)
> +    {
> +      token_counts[0][token_index] = token_counts[1][token_index] = 0;
> +      token_counts[2][token_index] = token_counts[3][token_index] = 0;
> +    }
> +
> +  current = position_list[0];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[0][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[0]);
> +    }
> +  current = position_list[1];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[1][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[1]);
> +    }
> +  current = position_list[2];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[2][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[2]);
> +    }
> +  current = position_list[3];
> +  if (current != NULL)
> +    {
> +      do
> +        {
> +          token_counts[3][current->node]++;
> +          current = current->next;
> +        }
> +      while (current != position_list[3]);
> +    }
> +
>   /* Get the lcs for original - latest */
> -  lcs_ol = svn_diff__lcs(position_list[0], position_list[2], prefix_lines,
> +  lcs_ol = svn_diff__lcs(position_list[0], position_list[2],
> +                         token_counts[0], token_counts[2],
> +                         num_tokens, prefix_lines,
>                          suffix_lines, subpool3);
>   diff_ol = svn_diff__diff(lcs_ol, 1, 1, TRUE, pool);
>
> @@ -257,7 +316,9 @@ svn_diff_diff4_2(svn_diff_t **diff,
>   /* Get the lcs for common ancestor - original
>    * Do reverse adjustements
>    */
> -  lcs_adjust = svn_diff__lcs(position_list[3], position_list[2], prefix_lines,
> +  lcs_adjust = svn_diff__lcs(position_list[3], position_list[2],
> +                             token_counts[3], token_counts[2],
> +                             num_tokens, prefix_lines,
>                              suffix_lines, subpool3);
>   diff_adjust = svn_diff__diff(lcs_adjust, 1, 1, FALSE, subpool3);
>   adjust_diff(diff_ol, diff_adjust);
> @@ -267,7 +328,9 @@ svn_diff_diff4_2(svn_diff_t **diff,
>   /* Get the lcs for modified - common ancestor
>    * Do forward adjustments
>    */
> -  lcs_adjust = svn_diff__lcs(position_list[1], position_list[3], prefix_lines,
> +  lcs_adjust = svn_diff__lcs(position_list[1], position_list[3],
> +                             token_counts[1], token_counts[3],
> +                             num_tokens, prefix_lines,
>                              suffix_lines, subpool3);
>   diff_adjust = svn_diff__diff(lcs_adjust, 1, 1, FALSE, subpool3);
>   adjust_diff(diff_ol, diff_adjust);
> @@ -283,7 +346,7 @@ svn_diff_diff4_2(svn_diff_t **diff,
>       if (hunk->type == svn_diff__type_conflict)
>         {
>           svn_diff__resolve_conflict(hunk, &position_list[1],
> -                                     &position_list[2], pool);
> +                                     &position_list[2], num_tokens, pool);
>         }
>     }
>
> ]]]
>
> Morten Kloster wrote on Wed, May 18, 2011 at 10:31:13 +0200:
>> On Wed, May 18, 2011 at 8:33 AM, Johan Corveleyn <jcorvel_at_gmail.com> wrote:
>> > On Wed, May 18, 2011 at 1:56 AM, Morten Kloster <morklo_at_gmail.com> wrote:
>> >> Log message:
>> >>
>> >> Speed-up of libsvn_diff using token counts
>> >> By using indices, not node pointers, to refer to tokens, and counting
>> >> the number of each token, the longest common subsequence (lcs)
>> >> algorithm at the heart of libsvn_diff becomes much faster in many
>> >> situations by skipping tokens that are unique to one source or the other.
>> >>
>> >> * subversion/libsvn_diff/token.c
>> >>  (svn_diff__node_t): Added 'index' member.
>> >>  (svn_diff__tree_t): Added 'node_count' member.
>> >>  (svn_diff__get_token_count): New method.
>> >>  (tree_insert_token): Assigns indices to nodes.
>> >>  (svn_diff__get_tokens): Uses index, not pointer,
>> >>   to identify node in position struct.
>> >>
>> >> * subversion/libsvn_diff/lcs.c
>> >>  (svn_diff__snake): Takes token counts as additional argument.
>> >>   Skips tokens that are unique to one or the other file.
>> >>  (svn_diff__lcs): Takes token counts and number of different tokens
>> >>   as additional arguments. Calculates number of tokens unique to
>> >>   each source and uses this to compute effective length difference.
>> >>
>> >> * subversion/libsvn_diff/diff.h
>> >>  (svn_diff__position_t): Changed node member from being
>> >>   a pointer to the node to an integer index for the node.
>> >>  Updated method declarations.
>> >>
>> >> * subversion/libsvn_diff/diff.c
>> >> * subversion/libsvn_diff/diff3.c
>> >> * subversion/libsvn_diff/diff4.c
>> >>  (svn_diff_diff_2, svn_diff_diff_3, svn_diff_diff_4): Count the
>> >>   number of times each token appears in each source.
>> >>  (svn_diff__resolve_conflict): Takes number of different tokens
>> >>   as additional argument. Counts the number of times
>> >>   each token appears in each (partial) source.
>> >>
>> >>
>> >>
>> >>
>> >>
>> >>
>> >> Comments:
>> >> This patch can reduce the time required for a diff by a large factor
>> >> when comparing long files with large differences. As an extreme case,
>> >> splitting sqlite.c into two roughly equal parts and diffing them against
>> >> each other took about a minute on my laptop with the original code,
>> >> and 7-8 seconds with my patch. The speed-up is gained only if a
>> >> large fraction of the changed tokens are unique to one source or the
>> >> other. I can not see that the change would cause a significant
>> >> performance degradation in any reasonable case.
>> >>
>> >> Counting the number of times each token appears is also useful for
>> >> potential additional features in the future, such as identifying moved
>> >> blocks outside of the longest common subsequence.
>> >>
>> >> I have tested only svn_diff_diff_2 (that it gives identical results as
>> >> before), but the changes to svn_diff_diff3 and svn_diff_diff4 are
>> >> completely analogous. Tested on a Windows 7 64-bit machine,
>> >> TortoiseSVN build script.
>> >>
>> >>
>> >> Morten Kloster (first patch)
>> >
>> > Interesting! The patch seems to be missing though. The list software
>> > tends to strip off attachments if they don't have a text/plain
>> > mime-type. Can you try sending it again with a .txt extension?
>> >
>> > --
>> > Johan
>> >
>>
>> Sorry, here's the patch with .txt extension. I found a bug in the
>> first version anyway, so it was just as well. :)
>>
>> And of course I meant sqlite3.c for the stress test.
>>
>> Morten
>
>
Received on 2011-05-18 19:32:23 CEST

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.