* Sebastian Hammer, Adam Dickmeiss
*
* $Log: extract.c,v $
- * Revision 1.32 1995-11-25 10:24:05 adam
+ * Revision 1.50 1996-02-12 18:45:36 adam
+ * New fileVerboseFlag in record group control.
+ *
+ * Revision 1.49 1996/02/05 12:29:57 adam
+ * Logging reduced a bit.
+ * The remaining running time is estimated during register merge.
+ *
+ * Revision 1.48 1996/02/01 20:53:26 adam
+ * The temporary per-record keys are compacted a little, and duplication
+ * of the per-records keys are avoided when they are saved in the record
+ * information buffer.
+ *
+ * Revision 1.47 1996/01/17 14:57:48 adam
+ * Prototype changed for reader functions in extract/retrieve. File
+ * is identified by 'void *' instead of 'int.
+ *
+ * Revision 1.46 1995/12/15 14:57:16 adam
+ * Bug fix.
+ *
+ * Revision 1.45 1995/12/15 12:37:41 adam
+ * In addRecordKeyAny: Writes key only when attrSet != -1.
+ *
+ * Revision 1.44 1995/12/12 16:00:54 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 1.43 1995/12/11 09:12:46 adam
+ * The rec_get function returns NULL if record doesn't exist - will
+ * happen in the server if the result set records have been deleted since
+ * the creation of the set (i.e. the search).
+ * The server saves a result temporarily if it is 'volatile', i.e. the
+ * set is register dependent.
+ *
+ * Revision 1.42 1995/12/07 17:38:46 adam
+ * Work locking mechanisms for concurrent updates/commit.
+ *
+ * Revision 1.41 1995/12/06 16:06:42 adam
+ * Better diagnostics. Work on 'real' dictionary deletion.
+ *
+ * Revision 1.40 1995/12/05 16:57:40 adam
+ * More work on regular patterns.
+ *
+ * Revision 1.39 1995/12/05 13:20:18 adam
+ * Bug fix: file_read sometimes returned early EOF.
+ *
+ * Revision 1.38 1995/12/04 17:59:21 adam
+ * More work on regular expression conversion.
+ *
+ * Revision 1.37 1995/12/04 14:22:27 adam
+ * Extra arg to recType_byName.
+ * Started work on new regular expression parsed input to
+ * structured records.
+ *
+ * Revision 1.36 1995/11/30 08:34:29 adam
+ * Started work on commit facility.
+ * Changed a few malloc/free to xmalloc/xfree.
+ *
+ * Revision 1.35 1995/11/28 14:26:21 adam
+ * Bug fix: recordId with constant wasn't right.
+ * Bug fix: recordId dictionary entry wasn't deleted when needed.
+ *
+ * Revision 1.34 1995/11/28 09:09:38 adam
+ * Zebra config renamed.
+ * Use setting 'recordId' to identify record now.
+ * Bug fix in recindex.c: rec_release_blocks was invokeded even
+ * though the blocks were already released.
+ * File traversal properly deletes records when needed.
+ *
+ * Revision 1.33 1995/11/27 09:56:20 adam
+ * Record info elements better enumerated. Internal store of records.
+ *
+ * Revision 1.32 1995/11/25 10:24:05 adam
* More record fields - they are enumerated now.
* New options: flagStoreData flagStoreKey.
*
static int records_updated = 0;
static int records_deleted = 0;
-#define MATCH_DICT "match"
-
void key_open (int mem)
{
if (mem < 50000)
key_buf_used = 0;
key_file_no = 0;
- if (!(matchDict = dict_open (MATCH_DICT, 20, 1)))
+ if (!(matchDict = dict_open (GMATCH_DICT, 50, 1)))
{
- logf (LOG_FATAL, "dict_open fail of %s", MATCH_DICT);
+ logf (LOG_FATAL, "dict_open fail of %s", GMATCH_DICT);
exit (1);
}
assert (!records);
int buf_used;
int buf_max;
char *buf;
+ char prevAttrSet;
+ short prevAttrUse;
} reckeys;
static void addRecordKey (const RecWord *p)
char attrSet;
short attrUse;
size_t i;
+ int lead = 0;
if (reckeys.buf_used+1024 > reckeys.buf_max)
{
char *b;
- b = malloc (reckeys.buf_max += 65000);
+ b = xmalloc (reckeys.buf_max += 128000);
if (reckeys.buf_used > 0)
memcpy (b, reckeys.buf, reckeys.buf_used);
- free (reckeys.buf);
+ xfree (reckeys.buf);
reckeys.buf = b;
}
dst = reckeys.buf + reckeys.buf_used;
+
+ attrSet = p->attrSet;
+ if (reckeys.buf_used > 0 && reckeys.prevAttrSet == attrSet)
+ lead |= 1;
+ else
+ reckeys.prevAttrSet = attrSet;
+ attrUse = p->attrUse;
+ if (reckeys.buf_used > 0 && reckeys.prevAttrUse == attrUse)
+ lead |= 2;
+ else
+ reckeys.prevAttrUse = attrUse;
+
switch (p->which)
{
case Word_String:
- attrSet = p->attrSet;
- memcpy (dst, &attrSet, sizeof(attrSet));
- dst += sizeof(attrSet);
+ *dst++ = lead;
- attrUse = p->attrUse;
- memcpy (dst, &attrUse, sizeof(attrUse));
- dst += sizeof(attrUse);
-
+ if (!(lead & 1))
+ {
+ memcpy (dst, &attrSet, sizeof(attrSet));
+ dst += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (dst, &attrUse, sizeof(attrUse));
+ dst += sizeof(attrUse);
+ }
for (i = 0; p->u.string[i]; i++)
*dst++ = p->u.string[i];
*dst++ = '\0';
static void flushRecordKeys (SYSNO sysno, int cmd, struct recKeys *reckeys,
const char *databaseName)
{
+ char attrSet = -1;
+ short attrUse = -1;
int off = 0;
while (off < reckeys->buf_used)
{
const char *src = reckeys->buf + off;
- char attrSet;
- short attrUse;
struct it_key key;
-
- memcpy (&attrSet, src, sizeof(attrSet));
- src += sizeof(attrSet);
-
- memcpy (&attrUse, src, sizeof(attrUse));
- src += sizeof(attrUse);
+ int lead;
+
+ lead = *src++;
+ if (!(lead & 1))
+ {
+ memcpy (&attrSet, src, sizeof(attrSet));
+ src += sizeof(attrSet);
+ }
+ if (!(lead & 2))
+ {
+ memcpy (&attrUse, src, sizeof(attrUse));
+ src += sizeof(attrUse);
+ }
if (key_buf_used + 1024 > (ptr_top-ptr_i)*sizeof(char*))
key_flush ();
++ptr_i;
w.attrUse = 1016;
addRecordKey (&w);
}
- addRecordKey (p);
+ if (p->attrSet != -1)
+ addRecordKey (p);
}
-static char *file_buf;
-static int file_offset;
-static int file_bufsize;
+#define FILE_READ_BUFSIZE 4096
+struct file_read_info {
+ int file_noread;
+ int fd;
+#if FILE_READ_BUFSIZE
+ char *file_buf;
+ int file_offset;
+ int file_bufsize;
+#endif
+};
-static void file_read_start (int fd)
+static struct file_read_info *file_read_start (int fd)
{
- file_offset = 0;
- file_buf = xmalloc (4096);
- file_bufsize = read (fd, file_buf, 4096);
+ struct file_read_info *fi = xmalloc (sizeof(*fi));
+
+ fi->fd = fd;
+ fi->file_noread = 0;
+#if FILE_READ_BUFSIZE
+ fi->file_offset = 0;
+ fi->file_buf = xmalloc (FILE_READ_BUFSIZE);
+ fi->file_bufsize = read (fd, fi->file_buf, FILE_READ_BUFSIZE);
+#endif
+ return fi;
}
-static void file_read_stop (int fd)
+static void file_read_stop (struct file_read_info *fi)
{
- xfree (file_buf);
+ assert (fi);
+#if FILE_READ_BUFSIZE
+ xfree (fi->file_buf);
+ fi->file_buf = NULL;
+#endif
+ xfree (fi);
}
-static int file_read (int fd, char *buf, size_t count)
+static int file_read (void *handle, char *buf, size_t count)
{
- int l = file_bufsize - file_offset;
+ struct file_read_info *p = handle;
+ int fd = p->fd;
+#if FILE_READ_BUFSIZE
+ int l = p->file_bufsize - p->file_offset;
if (count > l)
{
int r;
if (l > 0)
- memcpy (buf, file_buf + file_offset, l);
+ memcpy (buf, p->file_buf + p->file_offset, l);
count = count-l;
- if (count > file_bufsize)
+ if (count > FILE_READ_BUFSIZE)
{
if ((r = read (fd, buf + l, count)) == -1)
{
logf (LOG_FATAL|LOG_ERRNO, "read");
exit (1);
}
- file_bufsize = 0;
- file_offset = 0;
- return r;
+ p->file_bufsize = 0;
+ p->file_offset = 0;
+ p->file_noread += l+r;
+ return l+r;
}
- file_bufsize = r = read (fd, file_buf, 4096);
+ p->file_bufsize = r = read (fd, p->file_buf, FILE_READ_BUFSIZE);
if (r == -1)
{
logf (LOG_FATAL|LOG_ERRNO, "read");
}
else if (r <= count)
{
- file_offset = r;
- memcpy (buf + l, file_buf, r);
- return l + r;
+ p->file_offset = r;
+ memcpy (buf + l, p->file_buf, r);
+ p->file_noread += l+r;
+ return l+r;
}
else
{
- file_offset = count;
- memcpy (buf + l, file_buf, count - l);
+ p->file_offset = count;
+ memcpy (buf + l, p->file_buf, count - l);
+ p->file_noread += count;
return count;
}
}
- memcpy (buf, file_buf + file_offset, count);
- file_offset += count;
+ memcpy (buf, p->file_buf + p->file_offset, count);
+ p->file_offset += count;
+ p->file_noread += count;
return count;
+#else
+ int r;
+ r = read (fd, buf, count);
+ if (r > 0)
+ p->file_noread += r;
+ return r;
+#endif
}
static int atois (const char **s)
static char *fileMatchStr (struct recKeys *reckeys, struct recordGroup *rGroup,
const char *fname,
- const char *recordType,
const char *spec)
{
static char dstBuf[2048];
}
if (first)
{
- logf (LOG_WARN, "Record in file %s didn't contain match"
- " fields in (%d,%d)", fname, attrSet, attrUse);
+ logf (LOG_WARN, "Record didn't contain match"
+ " fields in (%d,%d)", attrSet, attrUse);
return NULL;
}
}
special[spec_len] = '\0';
s = s1;
- if (strcmp (special, "group"))
+ if (!strcmp (special, "group"))
spec_src = rGroup->groupName;
- else if (strcmp (special, "database"))
+ else if (!strcmp (special, "database"))
spec_src = rGroup->databaseName;
- else if (strcmp (special, "filename"))
+ else if (!strcmp (special, "filename"))
spec_src = fname;
- else if (strcmp (special, "type"))
- spec_src = recordType;
+ else if (!strcmp (special, "type"))
+ spec_src = rGroup->recordType;
else
spec_src = NULL;
if (spec_src)
while (*s && *s != stopMarker)
{
if (i < 63)
- tmpString[i++] = *s;
+ tmpString[i++] = *s++;
}
if (*s)
s++;
return dstBuf;
}
+struct recordLogInfo {
+ const char *fname;
+ char *op;
+ struct recordGroup *rGroup;
+};
+
+static void recordLogPreamble (int level, const char *msg, void *info)
+{
+ struct recordLogInfo *p = info;
+ FILE *outf = log_file ();
+
+ if (level & LOG_LOG)
+ return ;
+ if (p->op)
+ fprintf (outf, "%s of ", p->op);
+ fprintf (outf, "%s type %s\n", p->rGroup->recordType, p->fname);
+ log_event_start (NULL, NULL);
+}
+
static int recordExtract (SYSNO *sysno, const char *fname,
struct recordGroup *rGroup, int deleteFlag,
- int fd,
- const char *file_type,
- RecType recType)
+ struct file_read_info *fi, RecType recType,
+ char *subType)
{
struct recExtractCtrl extractCtrl;
int r;
char *matchStr;
SYSNO sysnotmp;
Record rec;
+ struct recordLogInfo logInfo;
- extractCtrl.fd = fd;
+ logInfo.fname = fname;
+ logInfo.op = NULL;
+ logInfo.rGroup = rGroup;
+ log_event_start (recordLogPreamble, &logInfo);
- /* extract keys */
- extractCtrl.subType = "";
- extractCtrl.init = wordInit;
- extractCtrl.add = addRecordKeyAny;
-
- reckeys.buf_used = 0;
- file_read_start (extractCtrl.fd);
- extractCtrl.readf = file_read;
- r = (*recType->extract)(&extractCtrl);
- file_read_stop (extractCtrl.fd);
- close (extractCtrl.fd);
-
- if (r)
+ if (fi->fd != -1)
{
- logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
- return 0;
+ extractCtrl.fh = fi;
+ /* extract keys */
+ extractCtrl.subType = subType;
+ extractCtrl.init = wordInit;
+ extractCtrl.add = addRecordKeyAny;
+
+ reckeys.buf_used = 0;
+ reckeys.prevAttrUse = -1;
+ reckeys.prevAttrSet = -1;
+ extractCtrl.readf = file_read;
+ r = (*recType->extract)(&extractCtrl);
+
+ if (r)
+ {
+ logf (LOG_WARN, "Couldn't extract file %s, code %d", fname, r);
+ return 0;
+ }
}
/* perform match if sysno not known and if match criteria is specified */
{
sysnotmp = 0;
sysno = &sysnotmp;
- if (rGroup->fileMatch)
+ if (rGroup->recordId && *rGroup->recordId)
{
char *rinfo;
- matchStr = fileMatchStr(&reckeys, rGroup, fname, file_type,
- rGroup->fileMatch);
+ matchStr = fileMatchStr (&reckeys, rGroup, fname,
+ rGroup->recordId);
if (matchStr)
{
rinfo = dict_lookup (matchDict, matchStr);
}
else
{
- logf (LOG_WARN, "Record not inserted");
+ logf (LOG_WARN, "Bad match criteria");
return 0;
}
}
{
if (deleteFlag)
{
- logf (LOG_LOG, "? record %s", fname);
+ logf (LOG_LOG, "Cannot delete new record");
return 1;
}
- logf (LOG_LOG, "add record %s", fname);
+ logInfo.op = "add";
+ if (rGroup->fileVerboseFlag)
+ logf (LOG_LOG, "add %s %s", rGroup->recordType, fname);
rec = rec_new (records);
*sysno = rec->sysno;
if (matchStr)
+ {
dict_insert (matchDict, matchStr, sizeof(*sysno), sysno);
+ }
flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName);
records_inserted++;
struct recKeys delkeys;
rec = rec_get (records, *sysno);
-
+ assert (rec);
delkeys.buf_used = rec->size[recInfo_delKeys];
delkeys.buf = rec->info[recInfo_delKeys];
flushRecordKeys (*sysno, 0, &delkeys, rec->info[recInfo_databaseName]);
if (deleteFlag)
{
+ logInfo.op = "delete";
if (!delkeys.buf_used)
{
- logf (LOG_WARN, "cannot delete %s - no delete keys",
- fname);
+ logf (LOG_WARN, "cannot delete; storeKeys false");
}
else
- logf (LOG_LOG, "delete record %s", fname);
- records_deleted++;
- rec_del (records, &rec);
+ {
+ if (rGroup->fileVerboseFlag)
+ logf (LOG_LOG, "delete %s %s", rGroup->recordType, fname);
+ records_deleted++;
+ if (matchStr)
+ dict_delete (matchDict, matchStr);
+ rec_del (records, &rec);
+ }
return 1;
}
else
{
+ logInfo.op = "update";
if (!delkeys.buf_used)
{
- logf (LOG_WARN, "cannot update %s - no delete keys",
- fname);
+ logf (LOG_WARN, "cannot update; storeKeys false");
}
else
{
- logf (LOG_LOG, "update record %s", fname);
+ if (rGroup->fileVerboseFlag)
+ logf (LOG_LOG, "update %s %s", rGroup->recordType, fname);
flushRecordKeys (*sysno, 1, &reckeys, rGroup->databaseName);
records_updated++;
}
}
}
- free (rec->info[recInfo_fileType]);
+ xfree (rec->info[recInfo_fileType]);
rec->info[recInfo_fileType] =
- rec_strdup (file_type, &rec->size[recInfo_fileType]);
+ rec_strdup (rGroup->recordType, &rec->size[recInfo_fileType]);
- free (rec->info[recInfo_filename]);
+ xfree (rec->info[recInfo_filename]);
rec->info[recInfo_filename] =
rec_strdup (fname, &rec->size[recInfo_filename]);
- free (rec->info[recInfo_delKeys]);
+ xfree (rec->info[recInfo_delKeys]);
if (reckeys.buf_used > 0 && rGroup->flagStoreKeys == 1)
{
- logf (LOG_LOG, "Storing keys...");
- rec->info[recInfo_delKeys] = malloc (reckeys.buf_used);
+#if 1
+ rec->size[recInfo_delKeys] = reckeys.buf_used;
+ rec->info[recInfo_delKeys] = reckeys.buf;
+ reckeys.buf = NULL;
+ reckeys.buf_max = 0;
+#else
+ rec->info[recInfo_delKeys] = xmalloc (reckeys.buf_used);
rec->size[recInfo_delKeys] = reckeys.buf_used;
memcpy (rec->info[recInfo_delKeys], reckeys.buf,
rec->size[recInfo_delKeys]);
+#endif
}
else
{
rec->info[recInfo_delKeys] = NULL;
rec->size[recInfo_delKeys] = 0;
}
- free (rec->info[recInfo_databaseName]);
+
+ xfree (rec->info[recInfo_storeData]);
+ if (rGroup->flagStoreData == 1)
+ {
+ rec->size[recInfo_storeData] = fi->file_noread;
+ rec->info[recInfo_storeData] = xmalloc (fi->file_noread);
+#if FILE_READ_BUFSIZE
+ if (fi->file_noread < FILE_READ_BUFSIZE)
+ memcpy (rec->info[recInfo_storeData], fi->file_buf,
+ fi->file_noread);
+ else
+#endif
+ {
+ if (lseek (fi->fd, 0L, SEEK_SET) < 0)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "seek to 0 in %s", fname);
+ exit (1);
+ }
+ if (read (fi->fd, rec->info[recInfo_storeData], fi->file_noread)
+ < fi->file_noread)
+ {
+ logf (LOG_ERRNO|LOG_FATAL, "read %d bytes of %s",
+ fi->file_noread, fname);
+ exit (1);
+ }
+ }
+ }
+ else
+ {
+ rec->info[recInfo_storeData] = NULL;
+ rec->size[recInfo_storeData] = 0;
+ }
+ xfree (rec->info[recInfo_databaseName]);
rec->info[recInfo_databaseName] =
rec_strdup (rGroup->databaseName, &rec->size[recInfo_databaseName]);
rec_put (records, &rec);
+ log_event_start (NULL, NULL);
return 1;
}
-int fileExtract (SYSNO *sysno, const char *fname, struct recordGroup *rGroup,
- int deleteFlag)
+int fileExtract (SYSNO *sysno, const char *fname,
+ const struct recordGroup *rGroupP, int deleteFlag)
{
int i, fd;
char gprefix[128];
char ext[128];
char ext_res[128];
- const char *file_type;
+ char subType[128];
RecType recType;
+ struct recordGroup rGroupM;
+ struct recordGroup *rGroup = &rGroupM;
+ struct file_read_info *fi;
+ memcpy (rGroup, rGroupP, sizeof(*rGroupP));
+
if (!rGroup->groupName || !*rGroup->groupName)
*gprefix = '\0';
else
break;
}
/* determine file type - depending on extension */
- sprintf (ext_res, "%sfileExtension.%s", gprefix, ext);
- if (!(file_type = res_get (common_resource, ext_res)))
+ if (!rGroup->recordType)
+ {
+ sprintf (ext_res, "%srecordType.%s", gprefix, ext);
+ if (!(rGroup->recordType = res_get (common_resource, ext_res)))
+ {
+ sprintf (ext_res, "%srecordType", gprefix);
+ if (!(rGroup->recordType = res_get (common_resource, ext_res)))
+ {
+ if (rGroup->fileVerboseFlag)
+ logf (LOG_LOG, "? %s", fname);
+ return 0;
+ }
+ }
+ }
+ if (!rGroup->recordType)
+ {
+ if (rGroup->fileVerboseFlag)
+ logf (LOG_LOG, "? record %s", fname);
return 0;
- if (!(recType = recType_byName (file_type)))
+ }
+ if (!(recType = recType_byName (rGroup->recordType, subType)))
+ {
+ logf (LOG_WARN, "No such record type: %s", rGroup->recordType);
return 0;
+ }
/* determine match criteria */
- if (rGroup->fileMatch)
+ if (!rGroup->recordId)
{
- sprintf (ext_res, "%sfileMatch.%s", gprefix, ext);
- rGroup->fileMatch = res_get (common_resource, ext_res);
- if (!rGroup->fileMatch)
- {
- sprintf (ext_res, "%sfileMatch", gprefix);
- rGroup->fileMatch = res_get (common_resource, ext_res);
- }
+ sprintf (ext_res, "%srecordId.%s", gprefix, ext);
+ rGroup->recordId = res_get (common_resource, ext_res);
}
/* determine database name */
if (rGroup->flagStoreData == -1)
rGroup->flagStoreData = 0;
-
if (rGroup->flagStoreKeys == -1)
{
const char *sval;
if (rGroup->flagStoreKeys == -1)
rGroup->flagStoreKeys = 0;
-
- /* open input file */
- if ((fd = open (fname, O_RDONLY)) == -1)
+ if (sysno && deleteFlag)
+ fd = -1;
+ else
{
- logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
- return 0;
+ if ((fd = open (fname, O_RDONLY)) == -1)
+ {
+ logf (LOG_WARN|LOG_ERRNO, "open %s", fname);
+ return 0;
+ }
}
- recordExtract (sysno, fname, rGroup, deleteFlag, fd,
- file_type, recType);
- close (fd);
+ fi = file_read_start (fd);
+ recordExtract (sysno, fname, rGroup, deleteFlag, fi, recType, subType);
+ file_read_stop (fi);
+ if (fd != -1)
+ close (fd);
return 1;
}