Merge with git+ssh://cvs.ucw.cz/projects/sherlock/GIT/sherlock.git

[libucw.git] / lib / lizard.c

diff --git a/lib/lizard.c b/lib/lizard.c
index 3b0026a0fa59ac7185d605ef9369c682401cbb11..e5035e05698827fb0af13e0d3503110bbe79fbd8 100644 (file)
--- a/lib/lizard.c
+++ b/lib/lizard.c
@@ -6,7 +6,7 @@
  *     This software may be freely distributed and used according to the terms
  *     of the GNU Lesser General Public License.
  *
  *     This software may be freely distributed and used according to the terms
  *     of the GNU Lesser General Public License.
  *

- *     The file format is based on LZO1X and 
+ *     The file format is based on LZO1X and

  *     the compression method is based on zlib.
  */
 
  *     the compression method is based on zlib.
  */
 


@@ -35,7 +35,7 @@ hashf(byte *string)
 }
 
 static inline byte *
 }
 
 static inline byte *

-locate_string(byte *string, uns record_id, uns head)
+locate_string(byte *string, int record_id, int head)

   /* The strings are recorded into the hash-table regularly, hence there is no
    * need to store the pointer there.  */
 {
   /* The strings are recorded into the hash-table regularly, hence there is no
    * need to store the pointer there.  */
 {


@@ -177,7 +177,7 @@ lizard_compress(byte *in, uns in_len, byte *out)
   while (in < in_end)
   {
     uns hash = hashf(in);
   while (in < in_end)
   {
     uns hash = hashf(in);

-    byte *best;
+    byte *best = NULL;

     uns len = find_match(hash_tab[hash], hash_rec, in, in_end, &best, head);
     if (len < 3)
 #if 0                  // TODO: now, our routine does not detect matches of length 2
     uns len = find_match(hash_tab[hash], hash_rec, in, in_end, &best, head);
     if (len < 3)
 #if 0                  // TODO: now, our routine does not detect matches of length 2


@@ -380,7 +380,8 @@ lizard_decompress(byte *in, byte *out)
     else
     {                                          /* overlapping */
       for (; len-- > 0; out++)
     else
     {                                          /* overlapping */
       for (; len-- > 0; out++)

-       *out = out[-pos];
+       *out = *(out-pos);
+      /* It's tempting to use out[-pos] above, but unfortunately it's not the same */

     }
     /* extract the copy-bits */
     len = in[-2] & 0x3;
     }
     /* extract the copy-bits */
     len = in[-2] & 0x3;


@@ -415,18 +416,22 @@ perform_copy_command:
 Description of the LZO1X format :
 =================================
 
 Description of the LZO1X format :
 =================================
 

+The meaning of the commands depends on the current mode. It can be either
+the compressed mode or the copy mode. In some cases, the compressed mode
+also distinguishes whether we just left the copy mode or not.
+

 Beginning of file:
 ------------------
 
 Beginning of file:
 ------------------
 

-If the first byte is 00010001, it means probably EOF (empty file), so switch
-to the compressed mode.  If it is bigger, subtract 17 and copy this number of
-the following characters to the ouput and switch to the compressed mode.  If
-it is smaller, go to the copy mode.
+Start in copy mode. If the first byte is 00010001, it means probably EOF (empty file),
+so switch to the compressed mode.  If it is bigger, subtract 17 and copy this number of
+the following characters to the output and switch to the compressed mode.
+If it is smaller, interpret it as a regular copy mode command.

 
 

-Compressed mode :
------------------
+Compressed mode:
+----------------

 
 

-Read the first byte of the sequence and determine the type of bit-encoding by
+Read the first byte of the sequence and determine the type of bit encoding by

 looking at the most significant bits.  The sequence is always at least 2 bytes
 long.  Decode sequences of these types until the EOF or END marker is read.
 
 looking at the most significant bits.  The sequence is always at least 2 bytes
 long.  Decode sequences of these types until the EOF or END marker is read.
 


@@ -445,25 +450,23 @@ long.  Decode sequences of these types until the EOF or END marker is read.
     C following characters will be copied from the compressed text to the
     output.  The number CC is always stored in the 2 least significant bits of
     the second last byte of the sequence.
     C following characters will be copied from the compressed text to the
     output.  The number CC is always stored in the 2 least significant bits of
     the second last byte of the sequence.

-    
+

     If END is read, the algorithm switches to the copy mode.
 
 pattern                                        length          position
 
     If END is read, the algorithm switches to the copy mode.
 
 pattern                                        length          position
 

-0000ppCC                pppppppp       2               10 bits (*)
-0001pLLL L*    ppppppCC pppppppp       3..9 + extend   15 bits + EOF
+0000ppCC                pppppppp       2               10 bits         [default interpretation]
+0000ppCC                pppppppp       3               10 bits + 2048  [just after return from copy mode]
+0001pLLL L*    ppppppCC pppppppp       3..9 + extend   15 bits         [pos 0 interpreted as EOF]

 001LLLLL L*    ppppppCC pppppppp       3..33 + extend  14 bits
 001LLLLL L*    ppppppCC pppppppp       3..33 + extend  14 bits

-01\
-10 \
-11  \
-LLLpppCC                pppppppp       3..8            11 bits
+LLLpppCC                pppppppp       3..8            11 bits         [LLL >= 010]

 
 

-Copy mode :
------------
+Copy mode:
+----------

 
 Read the first byte and, if the most significant bits are 0000, perform the
 following command, otherwise switch to the compressed mode (and evaluate the
 
 Read the first byte and, if the most significant bits are 0000, perform the
 following command, otherwise switch to the compressed mode (and evaluate the

-command).
+command there).

 
 pattern                                        length          position
 
 
 pattern                                        length          position
 


@@ -472,9 +475,4 @@ pattern                                     length          position
   Copy L characters from the compressed text to the output.  The overhead for
   incompressible strings is only roughly 1/256 + epsilon.
 
   Copy L characters from the compressed text to the output.  The overhead for
   incompressible strings is only roughly 1/256 + epsilon.
 

-(*) After reading one copy command, switch to the compressed mode with the
-following "optimisation": the pattern 0000ppCC expands to length 3 instead of 2
-and 2048 is added to the position (now it is slightly more than 11 bits),
-because a sequence of length 2 would never be used.
-

 */
 */