[Emacs-diffs] trunk r116775: Fix bug #16830 with slow search for newlines in forward-line.

[Eli Zaretskii]

------------------------------------------------------------
revno: 116775
revision-id: address@hidden
parent: address@hidden
fixes bug: http://debbugs.gnu.org/16830
committer: Eli Zaretskii <address@hidden>
branch nick: trunk
timestamp: Sun 2014-03-16 18:28:34 +0200
message:
  Fix bug #16830 with slow search for newlines in forward-line.
  
   src/search.c (find_newline): Speed up the function when using the
   newline cache, by halving the number of calls to
   region_cache_forward and region_cache_backward.
modified:
  src/ChangeLog                  changelog-20091113204419-o5vbwnq5f7feedwu-1438
  src/search.c                   search.c-20091113204419-o5vbwnq5f7feedwu-473

=== modified file 'src/ChangeLog'
--- a/src/ChangeLog     2014-03-15 11:16:12 +0000
+++ b/src/ChangeLog     2014-03-16 16:28:34 +0000
@@ -1,3 +1,9 @@
+2014-03-16  Eli Zaretskii  <address@hidden>
+
+       * search.c (find_newline): Speed up the function when using the
+       newline cache, by halving the number of call to
+       region_cache_forward and region_cache_backward.  (Bug#16830)
+
 2014-03-15  Juanma Barranquero  <address@hidden>
 
        * buffer.c (Fset_buffer): Document return value (bug#17015).

=== modified file 'src/search.c'
--- a/src/search.c      2014-01-01 17:44:48 +0000
+++ b/src/search.c      2014-03-16 16:28:34 +0000
@@ -715,19 +715,62 @@
            examine.  */
        ptrdiff_t tem, ceiling_byte = end_byte - 1;
 
-        /* If we're looking for a newline, consult the newline cache
-           to see where we can avoid some scanning.  */
+        /* If we're using the newline cache, consult it to see whether
+           we can avoid some scanning.  */
         if (newline_cache)
           {
             ptrdiff_t next_change;
+           int result = 1;
+
             immediate_quit = 0;
-            while (region_cache_forward
-                   (cache_buffer, newline_cache, start, &next_change))
-              start = next_change;
+            while (start < end && result)
+             {
+               ptrdiff_t lim1;
+
+               result = region_cache_forward (cache_buffer, newline_cache,
+                                              start, &next_change);
+               if (result)
+                 {
+                   start = next_change;
+                   lim1 = next_change = end;
+                 }
+               else
+                 lim1 = min (next_change, end);
+
+               /* The cache returned zero for this region; see if
+                  this is because the region is known and includes
+                  only newlines.  While at that, count any newlines
+                  we bump into, and exit if we found enough off them.  */
+               start_byte = CHAR_TO_BYTE (start);
+               while (start < lim1
+                      && FETCH_BYTE (start_byte) == '\n')
+                 {
+                   start_byte++;
+                   start++;
+                   if (--count == 0)
+                     {
+                       if (bytepos)
+                         *bytepos = start_byte;
+                       return start;
+                     }
+                 }
+               /* If we found a non-newline character before hitting
+                  position where the cache will again return non-zero
+                  (i.e. no newlines beyond that position), it means
+                  this region is not yet known to the cache, and we
+                  must resort to the "dumb loop" method.  */
+               if (start < next_change && !result)
+                 break;
+               result = 1;
+             }
+           if (start >= end)
+             {
+               start = end;
+               start_byte = end_byte;
+               break;
+             }
             immediate_quit = allow_quit;
 
-           start_byte = CHAR_TO_BYTE (start);
-
             /* START should never be after END.  */
             if (start_byte > ceiling_byte)
               start_byte = ceiling_byte;
@@ -762,9 +805,9 @@
              unsigned char *nl = memchr (lim_addr + cursor, '\n', - cursor);
              next = nl ? nl - lim_addr : 0;
 
-              /* If we're looking for newlines, cache the fact that
-                 this line's region is free of them. */
-              if (newline_cache)
+              /* If we're using the newline cache, cache the fact that
+                 the region we just traversed is free of newlines. */
+              if (newline_cache && cursor != next)
                {
                  know_region_cache (cache_buffer, newline_cache,
                                     BYTE_TO_CHAR (lim_byte + cursor),
@@ -800,14 +843,47 @@
         if (newline_cache)
           {
             ptrdiff_t next_change;
+           int result = 1;
+
             immediate_quit = 0;
-            while (region_cache_backward
-                   (cache_buffer, newline_cache, start, &next_change))
-              start = next_change;
+            while (start > end && result)
+             {
+               ptrdiff_t lim1;
+
+               result = region_cache_backward (cache_buffer, newline_cache,
+                                               start, &next_change);
+               if (result)
+                 {
+                   start = next_change;
+                   lim1 = next_change = end;
+                 }
+               else
+                 lim1 = max (next_change, end);
+               start_byte = CHAR_TO_BYTE (start);
+               while (start > lim1
+                      && FETCH_BYTE (start_byte - 1) == '\n')
+                 {
+                   if (++count == 0)
+                     {
+                       if (bytepos)
+                         *bytepos = start_byte;
+                       return start;
+                     }
+                   start_byte--;
+                   start--;
+                 }
+               if (start > next_change && !result)
+                 break;
+               result = 1;
+             }
+           if (start <= end)
+             {
+               start = end;
+               start_byte = end_byte;
+               break;
+             }
             immediate_quit = allow_quit;
 
-           start_byte = CHAR_TO_BYTE (start);
-
             /* Start should never be at or before end.  */
             if (start_byte <= ceiling_byte)
               start_byte = ceiling_byte + 1;
@@ -840,7 +916,7 @@
 
               /* If we're looking for newlines, cache the fact that
                  this line's region is free of them. */
-              if (newline_cache)
+              if (newline_cache && cursor != prev + 1)
                {
                  know_region_cache (cache_buffer, newline_cache,
                                     BYTE_TO_CHAR (ceiling_byte + prev + 1),