Charmap work
[idzebra-moved-to-github.git] / index / main.c
index 17fdbb4..8c18e9f 100644 (file)
@@ -1,10 +1,75 @@
 /*
- * Copyright (C) 1994-1995, Index Data I/S 
+ * Copyright (C) 1994-1996, Index Data I/S 
  * All rights reserved.
  * Sebastian Hammer, Adam Dickmeiss
  *
  * $Log: main.c,v $
- * Revision 1.25  1995-12-01 16:24:39  adam
+ * Revision 1.42  1996-05-31 09:07:01  quinn
+ * Work on character-set handling
+ *
+ * Revision 1.41  1996/05/14  14:04:34  adam
+ * In zebraidx, the 'stat' command is improved. Statistics about ISAM/DICT
+ * is collected.
+ *
+ * Revision 1.40  1996/04/26  10:00:23  adam
+ * Added option -V to zebraidx to display version information.
+ * Removed stupid warnings from file update.
+ *
+ * Revision 1.39  1996/04/09  10:05:21  adam
+ * Bug fix: prev_name buffer possibly too small; allocated in key_file_init.
+ *
+ * Revision 1.38  1996/03/26  16:01:14  adam
+ * New setting lockPath: directory of various lock files.
+ *
+ * Revision 1.37  1996/03/19  12:43:26  adam
+ * Bug fix: File update traversal didn't handle trailing slashes correctly.
+ * Bug fix: Update of sub directory groups wasn't handled correctly.
+ *
+ * Revision 1.36  1996/02/12  18:45:37  adam
+ * New fileVerboseFlag in record group control.
+ *
+ * Revision 1.35  1996/02/12  15:56:11  adam
+ * New code command: unread.
+ *
+ * Revision 1.34  1996/02/07  14:06:39  adam
+ * Better progress report during register merge.
+ * New command: clean - removes temporary shadow files.
+ *
+ * Revision 1.33  1996/02/06  17:11:18  adam
+ * Minor changes.
+ *
+ * Revision 1.32  1996/02/01  20:50:04  adam
+ * Bug fix: zebraIndexUnlock was always called even though zebraIndexLock
+ * was never called - happens when no commands are specified.
+ *
+ * Revision 1.31  1996/01/08  19:15:46  adam
+ * New input filter that works!
+ *
+ * Revision 1.30  1995/12/12  16:00:59  adam
+ * System call sync(2) used after update/commit.
+ * Locking (based on fcntl) uses F_EXLCK and F_SHLCK instead of F_WRLCK
+ * and F_RDLCK.
+ *
+ * Revision 1.29  1995/12/11  11:43:30  adam
+ * Locking based on fcntl instead of flock.
+ * Setting commitEnable removed. Command line option -n can be used to
+ * prevent commit if commit setting is defined in the configuration file.
+ *
+ * Revision 1.28  1995/12/08  16:22:56  adam
+ * Work on update while servers are running. Three lock files introduced.
+ * The servers reload their registers when necessary, but they don't
+ * reestablish result sets yet.
+ *
+ * Revision 1.27  1995/12/07  17:38:47  adam
+ * Work locking mechanisms for concurrent updates/commit.
+ *
+ * Revision 1.26  1995/12/06  12:41:23  adam
+ * New command 'stat' for the index program.
+ * Filenames can be read from stdin by specifying '-'.
+ * Bug fix/enhancement of the transformation from terms to regular
+ * expressons in the search engine.
+ *
+ * Revision 1.25  1995/12/01  16:24:39  adam
  * Commit files use separate meta file area.
  *
  * Revision 1.24  1995/11/30  17:01:38  adam
@@ -102,15 +167,20 @@ char *prog;
 size_t mem_max = 4*1024*1024;
 extern char *data1_tabpath;
 
+static void abort_func (int level, const char *msg, void *info)
+{
+    if (level & LOG_FATAL)
+        abort ();
+}
+
 int main (int argc, char **argv)
 {
-    int commit_at_end = 0;
     int ret;
     int cmd = 0;
     char *arg;
     char *configName = NULL;
     int nsections;
-    int key_open_flag = 0;
+    int disableCommit = 0;
 
     struct recordGroup rGroupDef;
     
@@ -121,6 +191,7 @@ int main (int argc, char **argv)
     rGroupDef.recordType = NULL;
     rGroupDef.flagStoreData = -1;
     rGroupDef.flagStoreKeys = -1;
+    rGroupDef.fileVerboseFlag = 1;
 
     prog = *argv;
     if (argc < 2)
@@ -130,24 +201,30 @@ int main (int argc, char **argv)
         " update <dir>  Update index with files below <dir>.\n"
        "               If <dir> is empty filenames are read from stdin.\n"
         " delete <dir>  Delete index with files below <dir>.\n"
+        " commit        Commit changes\n"
+        " clean         Clean shadow files\n"
         "Options:\n"
        " -t <type>     Index files as <type> (grs or text).\n"
        " -c <config>   Read configuration file <config>.\n"
        " -g <group>    Index files according to group settings.\n"
        " -d <database> Records belong to Z39.50 database <database>.\n"
        " -m <mbytes>   Use <mbytes> before flushing keys to disk.\n"
-       " -v <level>    Set logging to <level>.\n");
+        " -n            Don't use shadow system\n"
+       " -v <level>    Set logging to <level>\n"
+        " -V            Show version\n"
+                 );
         exit (1);
     }
-    while ((ret = options ("t:c:g:d:m:v:", argv, argc, &arg)) != -2)
+    log_event_end (abort_func, NULL);
+    while ((ret = options ("Vt:c:g:d:m:v:n", argv, argc, &arg)) != -2)
     {
         if (ret == 0)
         {
+            const char *rval;
             if(cmd == 0) /* command */
             {
                 if (!common_resource)
                 {
-                    const char *rval;
                     common_resource = res_open (configName ?
                                                 configName : FNAME_CONFIG);
                     if (!common_resource)
@@ -157,16 +234,72 @@ int main (int argc, char **argv)
                         exit (1);
                     }
                     data1_tabpath = res_get (common_resource, "profilePath");
-                    rval = res_get (common_resource, "commitEnable");
-                    if (rval && atoi(rval))
-                        bf_cache ();
+                    bf_lockDir (res_get (common_resource, "lockPath"));
+                   init_charmap();
                 }
                 if (!strcmp (arg, "update"))
                     cmd = 'u';
+                else if (!strcmp (arg, "update1"))
+                    cmd = 'U';
+                else if (!strcmp (arg, "update2"))
+                    cmd = 'm';
+                else if (!strcmp (arg, "dump"))
+                    cmd = 's';
                 else if (!strcmp (arg, "del") || !strcmp(arg, "delete"))
                     cmd = 'd';
                 else if (!strcmp (arg, "commit"))
-                    commit_at_end = 1;
+                {
+                    zebraIndexLock (1);
+                    rval = res_get (common_resource, "shadow");
+                    if (rval && *rval)
+                        bf_cache (1);
+                    else
+                    {
+                        logf (LOG_FATAL, "Cannot perform commit");
+                        logf (LOG_FATAL, "No shadow area defined");
+                        exit (1);
+                    }
+                    if (bf_commitExists ())
+                    {
+                        logf (LOG_LOG, "Commit start");
+                        zebraIndexLockMsg ("c");
+                        zebraIndexWait (1);
+                        logf (LOG_LOG, "Commit execute");
+                        bf_commitExec ();
+                        sync ();
+                        zebraIndexLockMsg ("d");
+                        zebraIndexWait (0);
+                        logf (LOG_LOG, "Commit clean");
+                        bf_commitClean ();
+                    }
+                    else
+                        logf (LOG_LOG, "Nothing to commit");
+                }
+                else if (!strcmp (arg, "clean"))
+                {
+                    zebraIndexLock (1);
+                    if (bf_commitExists ())
+                    {
+                        zebraIndexLockMsg ("d");
+                        zebraIndexWait (0);
+                        logf (LOG_LOG, "Commit clean");
+                        bf_commitClean ();
+                    }
+                    else
+                        logf (LOG_LOG, "Nothing to clean");
+                }
+                else if (!strcmp (arg, "stat") || !strcmp (arg, "status"))
+                {
+                    zebraIndexLock (0);
+                    rval = res_get (common_resource, "shadow");
+                    if (rval && *rval)
+                    {
+                        bf_cache (1);
+                        zebraIndexLockMsg ("r");
+                    }
+                    rec_prstat ();
+                    inv_prstat  (FNAME_WORD_DICT, FNAME_WORD_ISAM);
+                }
                 else
                 {
                     logf (LOG_FATAL, "Unknown command: %s", arg);
@@ -177,20 +310,69 @@ int main (int argc, char **argv)
             {
                 struct recordGroup rGroup;
 
-                memcpy (&rGroup, &rGroupDef, sizeof(rGroup));
-                if (!key_open_flag)
+                zebraIndexLock (0);
+                rval = res_get (common_resource, "shadow");
+                if (rval && *rval && !disableCommit)
                 {
-                    key_open (mem_max);
-                    key_open_flag = 1;
+                    bf_cache (1);
+                    zebraIndexLockMsg ("r");
+                }
+                else
+                {
+                    bf_cache (0);
+                    zebraIndexLockMsg ("w");
                 }
+                zebraIndexWait (0);
+                
+                memcpy (&rGroup, &rGroupDef, sizeof(rGroup));
                 rGroup.path = arg;
-                if (cmd == 'u')
+                switch (cmd)
+                {
+                case 'u':
+                    key_open (mem_max);
+                    logf (LOG_LOG, "Updating %s", rGroup.path);
                     repositoryUpdate (&rGroup);
-                else if (cmd == 'd')
+                    nsections = key_close ();
+                    break;
+                case 'U':
+                    key_open (mem_max);
+                    logf (LOG_LOG, "Updating (pass 1) %s", rGroup.path);
+                    repositoryUpdate (&rGroup);
+                    key_close ();
+                    nsections = 0;
+                    break;
+                case 'd':
+                    key_open (mem_max);
+                    logf (LOG_LOG, "Deleting %s", rGroup.path);
                     repositoryDelete (&rGroup);
+                    nsections = key_close ();
+                    break;
+                case 's':
+                    logf (LOG_LOG, "Dumping %s", rGroup.path);
+                    repositoryShow (&rGroup);
+                    nsections = 0;
+                    break;
+                case 'm':
+                    nsections = -1;
+                    break;
+                default:
+                    nsections = 0;
+                }
                 cmd = 0;
+                if (nsections)
+                {
+                    logf (LOG_LOG, "Merging with index");
+                    key_input (FNAME_WORD_DICT, FNAME_WORD_ISAM, nsections,
+                               60);
+                    sync ();
+                }
             }
         }
+        else if (ret == 'V')
+        {
+            fprintf (stderr, "Zebra %s %s\n",
+                     ZEBRAVER, ZEBRADATE);
+        }
         else if (ret == 'v')
         {
             log_init (log_mask_str(arg), prog, NULL);
@@ -211,26 +393,16 @@ int main (int argc, char **argv)
             configName = arg;
         else if (ret == 't')
             rGroupDef.recordType = arg;
+        else if (ret == 'n')
+            disableCommit = 1;
         else
         {
             logf (LOG_FATAL, "Unknown option '-%s'", arg);
             exit (1);
         }
     }
-    if (key_open_flag)
-    {
-        nsections = key_close ();
-        if (nsections)
-        {
-            logf (LOG_LOG, "Merging with index");
-            key_input (FNAME_WORD_DICT, FNAME_WORD_ISAM, nsections, 60);
-        }
-    }
-    if (commit_at_end)
-    {
-        logf (LOG_LOG, "commiting");
-        bf_commit ();
-    }
+    if (common_resource)
+        zebraIndexUnlock ();
     exit (0);
 }