First commit of archive example. Archive is a simple storage engine that...

First commit of archive example. Archive is a simple storage engine that handles inserts and selects. 


acconfig.h:
  Adding undef piece for HAVE_ARCHIVE_DB	
acinclude.m4:
  Code needed for --with-archive-storage-engine flag for compile.
configure.in:
  Adding tag for Archive
sql/Makefile.am:
  Source updates to compile ha_archive
sql/examples/ha_archive.cc:
  Class file for archive storage engine. First version.
sql/handler.cc:
  Updates needed for adding archive storage engine.
sql/handler.h:
  ENUM for archive storage engine.
sql/mysql_priv.h:
  Archive show options
sql/mysqld.cc:
  Ifdef foor HAVE_ARCHIVE_DB
sql/examples/ha_archive.h:
  Include file for archive storage engine addition.
BitKeeper/etc/logging_ok:
  Logging to logging@openlogging.org accepted

BitKeeper/etc/logging_ok

@@ -24,6 +24,7 @@ bell@laptop.sanja.is.com.ua
 bell@sanja.is.com.ua
 bk@admin.bk
 bk@mysql.r18.ru
+brian@avenger.(none)
 brian@brian-akers-computer.local
 carsten@tsort.bitbybit.dk
 davida@isil.mysql.com

acconfig.h

@@ -115,6 +115,9 @@
 /* Builds Example DB */
 #undef HAVE_EXAMPLE_DB
 
+/* Builds Archive Storage Engine */
+#undef HAVE_ARCHIVE_DB
+
 /* fp_except from ieeefp.h */
 #undef HAVE_FP_EXCEPT
 

acinclude.m4

@@ -1332,6 +1332,36 @@ dnl ---------------------------------------------------------------------------
 dnl END OF MYSQL_CHECK_EXAMPLE SECTION
 dnl ---------------------------------------------------------------------------
 
+dnl ---------------------------------------------------------------------------
+dnl Macro: MYSQL_CHECK_ARCHIVEDB
+dnl Sets HAVE_ARCHIVE_DB if --with-archive-storage-engine is used
+dnl ---------------------------------------------------------------------------
+AC_DEFUN([MYSQL_CHECK_ARCHIVEDB], [
+  AC_ARG_WITH([archive-storage-engine],
+              [
+  --with-archive-storage-engine
+                          Enable the Archive Storge Engine],
+              [archivedb="$withval"],
+              [archivedb=no])
+  AC_MSG_CHECKING([for archive storage engine])
+
+  case "$archivedb" in
+    yes )
+      AC_DEFINE(HAVE_ARCHIVE_DB)
+      AC_MSG_RESULT([yes])
+      [archivedb=yes]
+      ;;
+    * )
+      AC_MSG_RESULT([no])
+      [archivedb=no]
+      ;;
+  esac
+
+])
+dnl ---------------------------------------------------------------------------
+dnl END OF MYSQL_CHECK_ARCHIVE SECTION
+dnl ---------------------------------------------------------------------------
+
 dnl ---------------------------------------------------------------------------
 dnl Macro: MYSQL_CHECK_NDBCLUSTER
 dnl Sets HAVE_NDBCLUSTER_DB if --with-ndbcluster is used

configure.in

@@ -2619,6 +2619,7 @@ MYSQL_CHECK_ISAM
 MYSQL_CHECK_BDB
 MYSQL_CHECK_INNODB
 MYSQL_CHECK_EXAMPLEDB
+MYSQL_CHECK_ARCHIVEDB
 MYSQL_CHECK_NDBCLUSTER
 
 # If we have threads generate some library functions and test programs

sql/Makefile.am

@@ -58,7 +58,7 @@ noinst_HEADERS =	item.h item_func.h item_sum.h item_cmpfunc.h \
 			log_event.h sql_repl.h slave.h \
 			stacktrace.h sql_sort.h sql_cache.h set_var.h \
 			spatial.h gstream.h client_settings.h \
-			examples/ha_example.h
+			examples/ha_example.h examples/ha_archive.h
 mysqld_SOURCES =	sql_lex.cc sql_handler.cc \
 			item.cc item_sum.cc item_buff.cc item_func.cc \
 			item_cmpfunc.cc item_strfunc.cc item_timefunc.cc \
@@ -88,7 +88,7 @@ mysqld_SOURCES =	sql_lex.cc sql_handler.cc \
 			client.c sql_client.cc mini_client_errors.c pack.c\
 			stacktrace.c repl_failsafe.h repl_failsafe.cc sql_olap.cc\
 			gstream.cc spatial.cc sql_help.cc protocol_cursor.cc \
-			examples/ha_example.cc
+			examples/ha_example.cc examples/ha_archive.cc
 gen_lex_hash_SOURCES =	gen_lex_hash.cc
 gen_lex_hash_LDADD =	$(LDADD) $(CXXLDFLAGS)
 

sql/examples/ha_archive.cc

+/* Copyright (C) 2003 MySQL AB
+
+  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 of the License, 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
+
+#ifdef __GNUC__
+#pragma implementation        // gcc: Class implementation
+#endif
+
+#include <mysql_priv.h>
+
+#ifdef HAVE_ARCHIVE_DB
+#include "ha_archive.h"
+
+/*
+  First, if you want to understand storage engines you should look at 
+  ha_example.cc and ha_example.h. 
+  This example was written as a test case for a customer who needed
+  a storage engine without indexes that could compress data very well.
+  So, welcome to a completely compressed storage engine. This storage
+  engine only does inserts. No replace or updates. All reads are 
+  complete table scans. Compression is done through gzip (bzip compresses
+  better, but only marginally, if someone asks I could add support for
+  it too, but beaware that it costs a lot more in CPU time then gzip).
+  
+  We keep a file pointer open for each instance of ha_archive for each read
+  but for writes we keep one open file handle just for that. We flush it
+  only if we have a read occur. gzip handles compressing lots of records
+  at once much better then doing lots of little records between writes.
+  It is possible to not lock on writes but this would then mean we couldn't
+  handle bulk inserts as well (that is if someone was trying to read at
+  the same time since we would want to flush).
+
+  No attempts at durability are made. You can corrupt your data.
+
+  For performance as far as table scans go it is quite fast. I don't have
+  good numbers but locally it has out performed both Innodb and MyISAM. For
+  Innodb the question will be if the table can be fit into the buffer
+  pool. For MyISAM its a question of how much the file system caches the
+  MyISAM file. With enough free memory MyISAM is faster. Its only when the OS
+  doesn't have enough memory to cache entire table that archive turns out 
+  to be any faster. For writes it is always a bit slower then MyISAM. It has no
+  internal limits though for row length.
+
+  TODO:
+   Add bzip optional support.
+   Allow users to set compression level.
+   Add truncate table command.
+   Implement versioning, should be easy.
+   Implement optimize so we can fix broken tables.
+   Allow for errors, find a way to mark bad rows.
+   See if during an optimize you can make the table smaller.
+   Talk to the gzip guys, come up with a writable format so that updates are doable
+     without switching to a block method.
+
+    -Brian
+*/
+
+/* Variables for archive share methods */
+pthread_mutex_t archive_mutex;
+static HASH archive_open_tables;
+static int archive_init= 0;
+
+/* The file extension */
+#define ARZ ".ARZ"
+
+/*
+  Used for hash table that tracks open tables.
+*/
+static byte* archive_get_key(ARCHIVE_SHARE *share,uint *length,
+                             my_bool not_used __attribute__((unused)))
+{
+  *length=share->table_name_length;
+  return (byte*) share->table_name;
+}
+
+
+/*
+  Example of simple lock controls.
+  See ha_example.cc for a description.
+*/
+static ARCHIVE_SHARE *get_share(const char *table_name, TABLE *table)
+{
+  ARCHIVE_SHARE *share;
+  uint length;
+  char *tmp_name;
+
+  if (!archive_init)
+  {
+    /* Hijack a mutex for init'ing the storage engine */
+    pthread_mutex_lock(&LOCK_mysql_create_db);
+    if (!archive_init)
+    {
+      archive_init++;
+      VOID(pthread_mutex_init(&archive_mutex,MY_MUTEX_INIT_FAST));
+      (void) hash_init(&archive_open_tables,system_charset_info,32,0,0,
+                       (hash_get_key) archive_get_key,0,0);
+    }
+    pthread_mutex_unlock(&LOCK_mysql_create_db);
+  }
+  pthread_mutex_lock(&archive_mutex);
+  length=(uint) strlen(table_name);
+
+  if (!(share=(ARCHIVE_SHARE*) hash_search(&archive_open_tables,
+                                           (byte*) table_name,
+                                           length)))
+  {
+    if (!(share=(ARCHIVE_SHARE *)
+          my_multi_malloc(MYF(MY_WME | MY_ZEROFILL),
+                          &share, sizeof(*share),
+                          &tmp_name, length+1,
+                          NullS))) 
+    {
+      pthread_mutex_unlock(&archive_mutex);
+      return NULL;
+    }
+
+    share->use_count=0;
+    share->table_name_length=length;
+    share->table_name=tmp_name;
+    fn_format(share->data_file_name,table_name,"",ARZ,MY_REPLACE_EXT|MY_UNPACK_FILENAME);
+    strmov(share->table_name,table_name);
+    if (my_hash_insert(&archive_open_tables, (byte*) share))
+      goto error;
+    /* 
+      It is expensive to open and close the data files and since you can'thave
+      a gzip file that can be both read and written we keep two files open
+      that are shared amoung all open tables.
+    */
+    if ((share->archive_write = gzopen(share->data_file_name, "ab")) == NULL)
+      goto error;
+    thr_lock_init(&share->lock);
+    if (pthread_mutex_init(&share->mutex,MY_MUTEX_INIT_FAST))
+      goto error2;
+  }
+  share->use_count++;
+  pthread_mutex_unlock(&archive_mutex);
+
+  return share;
+
+error2:
+  thr_lock_delete(&share->lock);
+  /* We close, but ignore errors since we already have errors */
+  (void)gzclose(share->archive_write);
+error:
+  pthread_mutex_unlock(&archive_mutex);
+  my_free((gptr) share, MYF(0));
+
+  return NULL;
+}
+
+
+/* 
+  Free lock controls.
+  See ha_example.cc for a description.
+*/
+static int free_share(ARCHIVE_SHARE *share)
+{
+  int rc= 0;
+  pthread_mutex_lock(&archive_mutex);
+  if (!--share->use_count)
+  {
+    hash_delete(&archive_open_tables, (byte*) share);
+    thr_lock_delete(&share->lock);
+    pthread_mutex_destroy(&share->mutex);
+    my_free((gptr) share, MYF(0));
+    if (gzclose(share->archive_write) == Z_ERRNO)
+      rc = -1;
+  }
+  pthread_mutex_unlock(&archive_mutex);
+
+  return rc;
+}
+
+
+/* 
+  We just implement one additional file extension.
+*/
+const char **ha_archive::bas_ext() const
+{ static const char *ext[]= { ARZ, NullS }; return ext; }
+
+
+/* 
+  When opening a file we:
+  Create/get our shared structure.
+  Init out lock.
+  We open the file we will read from.
+  Set the size of ref_length.
+*/
+int ha_archive::open(const char *name, int mode, uint test_if_locked)
+{
+  DBUG_ENTER("ha_archive::open");
+
+  if (!(share = get_share(name, table)))
+    DBUG_RETURN(1);
+  thr_lock_data_init(&share->lock,&lock,NULL);
+
+  if ((archive = gzopen(share->data_file_name, "rb")) == NULL)
+    DBUG_RETURN(-1);
+
+  DBUG_RETURN(0);
+}
+
+
+/*
+  Closes the file. We first close this storage engines file handle to the
+  archive and then remove our referece count to the table (and possibly
+  free it as well).
+  */
+int ha_archive::close(void)
+{
+  DBUG_ENTER("ha_archive::close");
+  int rc= 0;
+  if (gzclose(archive) == Z_ERRNO)
+    rc =-1;
+  rc |= free_share(share);
+  DBUG_RETURN();
+}
+
+
+/*
+  We create our data file here. The format is pretty simple. The first bytes in
+  any file are the version number. Currently we do nothing with this, but in
+  the future this gives us the ability to figure out version if we change the
+  format at all. After the version we starting writing our rows. Unlike other
+  storage engines we do not "pack" our data. Since we are about to do a general
+  compression, packing would just be a waste of CPU time. If the table has blobs
+  they are written after the row in the order of creation. 
+  So to read a row we:
+    Read the version
+    Read the record and copy it into buf
+    Loop through any blobs and read them
+  */
+int ha_archive::create(const char *name, TABLE *table_arg, HA_CREATE_INFO *create_info)
+{
+  File create_file;
+  char name_buff[FN_REFLEN];
+  size_t written;
+  DBUG_ENTER("ha_archive::create");
+
+  if ((create_file = my_create(fn_format(name_buff,name,"",ARZ,MY_REPLACE_EXT|MY_UNPACK_FILENAME),0,
+                               O_RDWR | O_TRUNC,MYF(MY_WME))) < 0)
+    DBUG_RETURN(-1);
+  if ((archive = gzdopen(create_file, "ab")) == NULL)
+    DBUG_RETURN(-1);
+  version = ARCHIVE_VERSION;
+  written = gzwrite(archive, &version, sizeof(version));
+  if (written == 0 || written != sizeof(version))
+    DBUG_RETURN(-1);
+  gzclose(archive);
+  (void)my_close(create_file,MYF(0));
+
+  DBUG_RETURN(0);
+}
+
+/* 
+  Looop at ha_archive::open() for an explanation of the row format.
+  Here we just write out the row.
+*/
+int ha_archive::write_row(byte * buf)
+{
+  char *pos;
+  z_off_t written;
+  DBUG_ENTER("ha_archive::write_row");
+
+  statistic_increment(ha_write_count,&LOCK_status);
+  if (table->timestamp_default_now)
+    update_timestamp(record+table->timestamp_default_now-1);
+  written = gzwrite(share->archive_write, buf, table->reclength);
+  share->dirty= true;
+  if (written == 0 || written != table->reclength)
+    DBUG_RETURN(-1);
+
+  for (Field_blob **field=table->blob_field ; *field ; field++)
+  {
+    char *ptr;
+    uint32 size= (*field)->get_length();
+
+    (*field)->get_ptr(&ptr);
+    written = gzwrite(share->archive_write, ptr, (unsigned)size);
+    if (written == 0 || written != size)
+      DBUG_RETURN(-1);
+  }
+
+  DBUG_RETURN(0);
+}
+
+
+/*
+  All calls that need to scan the table start with this method. If we are told
+  that it is a table scan we rewind the file to the beginning, otherwise
+  we assume the position will be set.
+*/
+int ha_archive::rnd_init(bool scan)
+{
+  DBUG_ENTER("ha_archive::rnd_init");
+  int read; // gzread() returns int, and we use this to check the header
+  /* We rewind the file so that we can read from the beginning if scan */
+  if(scan)
+    if (gzrewind(archive))
+      DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
+  /* 
+    If dirty, we lock, and then reset/flush the data.
+    I found that just calling gzflush() doesn't always work.
+  */
+  if (share->dirty == true)
+  {
+    pthread_mutex_lock(&share->mutex);
+    if (share->dirty == true)
+    {
+      gzclose(share->archive_write);
+      if ((share->archive_write = gzopen(share->data_file_name, "ab")) == NULL)
+      {
+        pthread_mutex_unlock(&share->mutex);
+        DBUG_RETURN(-1);
+      }
+      share->dirty= false;
+    }
+    pthread_mutex_unlock(&share->mutex);
+  }
+  
+  /* 
+    At the moment we just check the size of version to make sure the header is 
+    intact.
+  */
+  read= gzread(archive, &version, sizeof(version));
+  if (written == 0 || written != sizeof(version))
+    DBUG_RETURN(-1);
+  records = 0;
+  DBUG_RETURN(0);
+}
+
+
+/*
+  This is the method that is used to read a row. It assumes that the row is 
+  positioned where you want it.
+*/
+int ha_archive::read_row(byte *buf)
+{
+  int read; // Bytes read, gzread() returns int
+  char *last;
+  size_t total_blob_length= 0;
+  DBUG_ENTER("ha_archive::read_row");
+
+  read = gzread(archive, buf, table->reclength);
+
+  /* If we read nothing we are at the end of the file */
+  if (read == 0)
+    DBUG_RETURN(HA_ERR_END_OF_FILE);
+
+  /* If the record is the wrong size, the file is probably damaged */
+  if (read != table->reclength)
+    DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
+
+  /* Calculate blob length, we use this for our buffer */
+  for (Field_blob **field=table->blob_field; *field ; field++)
+    total_blob_length += (*field)->get_length();
+
+  /* Adjust our row buffer if we need be */
+  buffer.alloc(total_blob_length);
+  last = (char *)buffer.ptr();
+
+  /* Loopp through our blobs and read them */
+  for (Field_blob **field=table->blob_field; *field ; field++)
+  {
+    /* Need to setup buffer tomorrow so that it is sued to contain all blobs */
+    size_t size= (*field)->get_length();
+    read = gzread(archive, last, size);
+    if (read == 0 || read != size)
+      DBUG_RETURN(HA_ERR_CRASHED_ON_USAGE);
+    (*field)->set_ptr(size, last);
+    last += size;
+  }
+  DBUG_RETURN(0);
+}
+
+/* 
+  Called during ORDER BY. Its position is either from being called sequentially
+  or by having had ha_archive::rnd_pos() called before it is called.
+*/
+int ha_archive::rnd_next(byte *buf)
+{
+  DBUG_ENTER("ha_archive::rnd_next");
+  int rc;
+
+  statistic_increment(ha_read_rnd_next_count,&LOCK_status);
+  current_position = gztell(archive);
+  rc = read_row(buf);
+  if (!(HA_ERR_END_OF_FILE == rc))
+    records++;
+
+  DBUG_RETURN(rc);
+}
+
+
+/* 
+  Thanks to the table flag HA_REC_NOT_IN_SEQ this will be called after
+  each call to ha_archive::rnd_next() if an ordering of the rows is
+  needed.
+*/
+void ha_archive::position(const byte *record)
+{
+  DBUG_ENTER("ha_archive::position");
+  ha_store_ptr(ref, ref_length, current_position);
+  DBUG_VOID_RETURN;
+}
+
+
+/*
+  This is called after a table scan for each row if the results of the scan need
+  to be ordered. It will take *pos and use it to move the cursor in the file so
+  that the next row that is called is the correctly ordered row.
+*/
+int ha_archive::rnd_pos(byte * buf, byte *pos)
+{
+  DBUG_ENTER("ha_archive::rnd_pos");
+  statistic_increment(ha_read_rnd_count,&LOCK_status);
+  current_position = ha_get_ptr(pos, ref_length);
+  z_off_t seek= gzseek(archive, current_position, SEEK_SET);
+
+  DBUG_RETURN(read_row(buf));
+}
+
+/******************************************************************************
+
+  Everything below here is default, please look at ha_example.cc for 
+  descriptions.
+
+ ******************************************************************************/
+
+int ha_archive::update_row(const byte * old_data, byte * new_data)
+{
+
+  DBUG_ENTER("ha_archive::update_row");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+int ha_archive::delete_row(const byte * buf)
+{
+  DBUG_ENTER("ha_archive::delete_row");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+int ha_archive::index_read(byte * buf, const byte * key,
+                           uint key_len __attribute__((unused)),
+                           enum ha_rkey_function find_flag
+                           __attribute__((unused)))
+{
+  DBUG_ENTER("ha_archive::index_read");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+int ha_archive::index_read_idx(byte * buf, uint index, const byte * key,
+                               uint key_len __attribute__((unused)),
+                               enum ha_rkey_function find_flag
+                               __attribute__((unused)))
+{
+  DBUG_ENTER("ha_archive::index_read_idx");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+
+int ha_archive::index_next(byte * buf)
+{
+  DBUG_ENTER("ha_archive::index_next");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+int ha_archive::index_prev(byte * buf)
+{
+  DBUG_ENTER("ha_archive::index_prev");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+int ha_archive::index_first(byte * buf)
+{
+  DBUG_ENTER("ha_archive::index_first");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+int ha_archive::index_last(byte * buf)
+{
+  DBUG_ENTER("ha_archive::index_last");
+  DBUG_RETURN(HA_ERR_NOT_IMPLEMENTED);
+}
+
+
+void ha_archive::info(uint flag)
+{
+  DBUG_ENTER("ha_archive::info");
+  /* This is a lie, but you don't want the optimizer to see zero or 1 */
+  if (records < 2) 
+    records = 2;
+  DBUG_VOID_RETURN;
+}
+
+int ha_archive::extra(enum ha_extra_function operation)
+{
+  DBUG_ENTER("ha_archive::extra");
+  DBUG_RETURN(0);
+}
+
+int ha_archive::reset(void)
+{
+  DBUG_ENTER("ha_archive::reset");
+  DBUG_RETURN(0);
+}
+
+
+int ha_archive::external_lock(THD *thd, int lock_type)
+{
+  DBUG_ENTER("ha_archive::external_lock");
+  DBUG_RETURN(0);
+}
+
+THR_LOCK_DATA **ha_archive::store_lock(THD *thd,
+                                       THR_LOCK_DATA **to,
+                                       enum thr_lock_type lock_type)
+{
+  if (lock_type != TL_IGNORE && lock.type == TL_UNLOCK)
+    lock.type=lock_type;
+  *to++= &lock;
+  return to;
+}
+
+ha_rows ha_archive::records_in_range(int inx,
+                                     const byte *start_key,uint start_key_len,
+                                     enum ha_rkey_function start_search_flag,
+                                     const byte *end_key,uint end_key_len,
+                                     enum ha_rkey_function end_search_flag)
+{
+  DBUG_ENTER("ha_archive::records_in_range ");
+  DBUG_RETURN(records); // HA_ERR_NOT_IMPLEMENTED 
+}
+#endif /* HAVE_ARCHIVE_DB */

sql/examples/ha_archive.h

+/* Copyright (C) 2003 MySQL AB
+
+  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 of the License, 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA */
+
+#include <zlib.h>
+
+/*
+  Please read ha_archive.cc first. If you are looking for more general
+  answers on how storage engines work, look at ha_example.cc and 
+  ha_example.h.
+*/
+
+typedef struct st_archive_share {
+  char *table_name;
+  char data_file_name[FN_REFLEN];
+  uint table_name_length,use_count;
+  pthread_mutex_t mutex;
+  THR_LOCK lock;
+  gzFile archive_write;             /* Archive file we are working with */
+  bool dirty;                       /* Flag for if a flush should occur */
+} ARCHIVE_SHARE;
+
+/* 
+  Version for file format.
+  1 - Initial Version
+*/
+#define ARCHIVE_VERSION 1
+
+class ha_archive: public handler
+{
+  THR_LOCK_DATA lock;        /* MySQL lock */
+  ARCHIVE_SHARE *share;      /* Shared lock info */
+  gzFile archive;            /* Archive file we are working with */
+  z_off_t current_position;  /* The position of the row we just read */
+  byte byte_buffer[IO_SIZE]; /* Initial buffer for our string */
+  String buffer;             /* Buffer used for blob storage */
+  unsigned int version;      /* Used for recording version */
+
+public:
+  ha_archive(TABLE *table): handler(table)
+  {
+    /* Set our original buffer from pre-allocated memory */
+    buffer.set(byte_buffer, IO_SIZE, system_charset_info);
+
+    /* The size of the offset value we will use for position() */
+    ref_length = sizeof(z_off_t);
+  }
+  ~ha_archive() 
+  {
+  }
+  const char *table_type() const { return "ARCHIVE"; }
+  const char *index_type(uint inx) { return "NONE"; }
+  const char **bas_ext() const;
+  ulong table_flags() const
+  {
+    return (HA_REC_NOT_IN_SEQ | HA_NOT_EXACT_COUNT | HA_NO_WRITE_DELAYED |
+            HA_NO_AUTO_INCREMENT );
+  }
+  ulong index_flags(uint inx) const
+  {
+    return 0;
+  }
+  /* 
+    This is just a default, there is no real limit as far as
+    archive is concerned.
+  */
+  uint max_record_length() const { return HA_MAX_REC_LENGTH; }
+  uint max_keys()          const { return 0; }
+  uint max_key_parts()     const { return 0; }
+  uint max_key_length()    const { return 0; }
+  /*
+    Called in test_quick_select to determine if indexes should be used.
+  */
+  virtual double scan_time() { return (double) (records+deleted) / 20.0+10; }
+  /* The next method will never be called */
+  virtual double read_time(ha_rows rows) { return (double) rows /  20.0+1; }
+  virtual bool fast_key_read() { return 1;}
+
+  int open(const char *name, int mode, uint test_if_locked);
+  int close(void);
+  int write_row(byte * buf);
+  int update_row(const byte * old_data, byte * new_data);
+  int delete_row(const byte * buf);
+  int index_read(byte * buf, const byte * key,
+                 uint key_len, enum ha_rkey_function find_flag);
+  int index_read_idx(byte * buf, uint idx, const byte * key,
+                     uint key_len, enum ha_rkey_function find_flag);
+  int index_next(byte * buf);
+  int index_prev(byte * buf);
+  int index_first(byte * buf);
+  int index_last(byte * buf);
+  int rnd_init(bool scan=1);
+  int rnd_next(byte *buf);
+  int rnd_pos(byte * buf, byte *pos);
+  int ha_archive::read_row(byte *buf);
+  void position(const byte *record);
+  void info(uint);
+  int extra(enum ha_extra_function operation);
+  int reset(void);
+  int external_lock(THD *thd, int lock_type);
+  ha_rows records_in_range(int inx, const byte *start_key,uint start_key_len,
+                           enum ha_rkey_function start_search_flag,
+                           const byte *end_key,uint end_key_len,
+                           enum ha_rkey_function end_search_flag);
+  int create(const char *name, TABLE *form, HA_CREATE_INFO *create_info);
+
+  THR_LOCK_DATA **store_lock(THD *thd, THR_LOCK_DATA **to,
+                             enum thr_lock_type lock_type);
+};

sql/handler.cc

@@ -35,6 +35,9 @@
 #ifdef HAVE_EXAMPLE_DB
 #include "examples/ha_example.h"
 #endif
+#ifdef HAVE_ARCHIVE_DB
+#include "examples/ha_archive.h"
+#endif
 #ifdef HAVE_INNOBASE_DB
 #include "ha_innodb.h"
 #else
@@ -81,6 +84,8 @@ struct show_table_type_st sys_table_types[]=
    "Alias for BDB", DB_TYPE_BERKELEY_DB},
   {"EXAMPLE",&have_example_db,
    "Example storage engine", DB_TYPE_EXAMPLE_DB},
+  {"ARCHIVE",&have_archive_db,
+   "Archive storage engine", DB_TYPE_ARCHIVE_DB},
   {NullS, NULL, NullS, DB_TYPE_UNKNOWN}
 };
 
@@ -180,6 +185,10 @@ handler *get_new_handler(TABLE *table, enum db_type db_type)
 #ifdef HAVE_EXAMPLE_DB
   case DB_TYPE_EXAMPLE_DB:
     return new ha_example(table);
+#endif
+#ifdef HAVE_ARCHIVE_DB
+  case DB_TYPE_ARCHIVE_DB:
+    return new ha_archive(table);
 #endif
   case DB_TYPE_HEAP:
     return new ha_heap(table);

sql/handler.h

@@ -146,7 +146,7 @@ enum db_type { DB_TYPE_UNKNOWN=0,DB_TYPE_DIAB_ISAM=1,
 	       DB_TYPE_RMS_ISAM, DB_TYPE_HEAP, DB_TYPE_ISAM,
 	       DB_TYPE_MRG_ISAM, DB_TYPE_MYISAM, DB_TYPE_MRG_MYISAM,
 	       DB_TYPE_BERKELEY_DB, DB_TYPE_INNODB, DB_TYPE_GEMINI,
-	       DB_TYPE_EXAMPLE_DB, DB_TYPE_DEFAULT };
+	       DB_TYPE_EXAMPLE_DB, DB_TYPE_ARCHIVE_DB, DB_TYPE_DEFAULT };
 
 struct show_table_type_st {
   const char *type;

sql/mysql_priv.h

@@ -908,7 +908,8 @@ extern struct my_option my_long_options[];
 
 /* optional things, have_* variables */
 
-extern SHOW_COMP_OPTION have_isam, have_innodb, have_berkeley_db, have_example_db;
+extern SHOW_COMP_OPTION have_isam, have_innodb, have_berkeley_db;
+extern SHOW_COMP_OPTION have_example_db, have_archive_db;
 extern SHOW_COMP_OPTION have_raid, have_openssl, have_symlink;
 extern SHOW_COMP_OPTION have_query_cache, have_berkeley_db, have_innodb;
 extern SHOW_COMP_OPTION have_crypt;

sql/mysqld.cc

@@ -370,7 +370,7 @@ KEY_CACHE *sql_key_cache;
 CHARSET_INFO *system_charset_info, *files_charset_info ;
 CHARSET_INFO *national_charset_info, *table_alias_charset;
 
-SHOW_COMP_OPTION have_berkeley_db, have_innodb, have_isam, have_example_db;
+SHOW_COMP_OPTION have_berkeley_db, have_innodb, have_isam, have_example_db, have_archive_db;
 SHOW_COMP_OPTION have_raid, have_openssl, have_symlink, have_query_cache;
 SHOW_COMP_OPTION have_crypt, have_compress;
 
@@ -5126,6 +5126,11 @@ static void mysql_init_variables(void)
 #else
   have_example_db= SHOW_OPTION_NO;
 #endif
+#ifdef HAVE_ARCHIVE_DB
+  have_archive_db= SHOW_OPTION_YES;
+#else
+  have_archive_db= SHOW_OPTION_NO;
+#endif
 #ifdef USE_RAID
   have_raid=SHOW_OPTION_YES;
 #else