* See the file LICENSE for details.
* Heikki Levanto
*
- * merge-d.c: merge routines for isamd
+ * $Id: merge-d.c,v 1.22 1999-09-23 18:01:18 heikki Exp $
*
+ * bugs
+ * (none)
+ *
+ * missing
+ *
+ * optimize
+ * - study and optimize block sizes (later)
+ * - find a way to decide the size of an empty diffblock (after merge)
+ * - On allocating more blocks (in append and merge), check the order of
+ * blocks, and if needed, swap them.
+ * - Write a routine to save/load indexes into a block, save only as many
+ * bytes as needed (size, diff, diffindexes)
+ *
+ *
+ * caveat
+ * There is a confusion about the block addresses. cat or type is the category,
+ * pos or block is the block number. pp structures keep these two separate,
+ * and combine when saving the pp. The next pointer in the pp structure is
+ * also a combined address, but needs to be combined every time it is needed,
+ * and separated when the partss are needed... This is done with the isamd_
+ * _block, _type, and _addr macros. The _addr takes block and type as args,
+ * in that order. This conflicts with the order these are often mentioned in
+ * the debug log calls, and other places, leading to small mistakes here
+ * and there.
+ *
+ * Needs cleaning! The way diff blocks are handled in append and reading is
+ * quite different, and likely to give maintenance problems.
+ *
+ * log levels (set isamddebug=x in zebra.cfg (or what ever cfg file you use) )
+ * 0 = no logging. Default
+ * 1 = no logging here. isamd logs overall statistics
+ * 2 = Each call to isamd_append with start address and no more
+ * 3 = Start and type of append, start of merge, and result of append
+ * 4 = Block allocations
+ * 5 = Block-level operations (read/write)
+ * 6 = Details about diff blocks etc.
+ * 7 = Log each record as it passes the system (once)
+ * 8 = Log raw and (de)coded data
+ * 9 = Anything else that may be useful
+ * .. = Anything needed to hunt a specific bug
+ * (note that all tests in the code are like debug>3, which means 4 or above!)
+ *
+ * Design for the new and improved isamd
+ * Key points:
+ * - The first block is only diffs, no straight data
+ * - Additional blocks are straight data
+ * - When a diff block gets filled up, a data block is created by
+ * merging the diffs with the data
+ *
+ * Structure
+ * - Isamd_pp: buffer for diffs and for data
+ * keep both pos, type, and combined address
+ * routine to set the address
+ * - diffbuf: lengths as short ints, or bytes for small blocks
+ * - keys are of key_struct, not just a number of bytes.
+ *
+ * Routines
+ * - isamd_append
+ * - create_new_block if needed
+ * - append_diffs
+ * - load_diffs
+ * - get diffend, start encoding
+ * - while input data
+ * - encode it
+ * - if no room, then realloc block in larger size
+ * - if still no room, merge and exit
+ * - append in the block
+ *
+ * - merge
+ * - just as before, except that merges also input data directly
+ * - writes into new data blocks
+ *
+ *
+ * - isamd.c: load firstpp, load datablock
+ * save firstpp, save datablock
+ * - Readlength, writelength - handling right size of len fields
+ * - isamd_read_main_item: take also a merge input structure, and merge it too
+ * - prefilter: cache two inputs, and check if they cancel.
+ * - single-item optimization
+ *
+ * questions: Should we realloc firstblocks in a different size as the main
+ * blocks. Makes a sideways seek, which is bound to be slowe. But saves some
+ * update time. Compromise: alloc the first one in the size of the datablock,
+ * but increase if necessary. Large blocks get a large diff, ok. Small ones
+ * may get an extra seek in read, but save merges.
*/
+
+#define NEW_ISAM_D 1 /* not yet ready to delete the old one! */
+
#include <stdlib.h>
#include <assert.h>
#include <string.h>
#include <stdio.h>
#include <log.h>
+#include "../index/index.h"
#include "isamd-p.h"
+struct ISAMD_DIFF_s {
+ int diffidx;
+ int maxidx;
+ struct it_key key;
+ void *decodeData;
+ int mode;
+ int difftype;
+};
+
+#define DT_NONE 0 // no diff, marks end of sequence
+#define DT_DIFF 1 // ordinarry diff
+#define DT_MAIN 2 // main data
+#define DT_INPU 3 // input data to be merged
+#define DT_DONE 4 // done with all input here
+
+
+
+/***************************************************************
+ * Input preprocess filter
+ ***************************************************************/
+
+
+#define FILTER_NOTYET -1 /* no data read in yet, to be done */
+
+struct ISAMD_FILTER_s {
+ ISAMD_I data; /* where the data comes from */
+ ISAMD is; /* for debug flags */
+ struct it_key k1; /* the next item to be returned */
+ int m1; /* mode for k1 */
+ int r1; /* result for read of k1, or NOTYET */
+ struct it_key k2; /* the one after that */
+ int m2;
+ int r2;
+};
+
+typedef struct ISAMD_FILTER_s *FILTER;
+
+
+void filter_fill(FILTER F)
+{
+ while ( (F->r1 == FILTER_NOTYET) || (F->r2 == FILTER_NOTYET) )
+ {
+ if (F->r1==FILTER_NOTYET)
+ { /* move data forward in the filter */
+ F->k1 = F->k2;
+ F->m1 = F->m2;
+ F->r1 = F->r2;
+ if ( 0 != F->r1 ) /* not eof */
+ F->r2 = FILTER_NOTYET; /* say we want more */
+ if (F->is->method->debug > 9)
+ logf(LOG_LOG,"filt_fill: shift %d.%d m=%d r=%d",
+ F->k1.sysno,
+ F->k1.seqno,
+ F->m1, F->r1);
+ }
+ if (F->r2==FILTER_NOTYET)
+ { /* read new bottom value */
+ char *k_ptr = (char*) &F->k2;
+ F->r2 = (F->data->read_item)(F->data->clientData, &k_ptr, &F->m2);
+ if (F->is->method->debug > 9)
+ logf(LOG_LOG,"filt_fill: read %d.%d m=%d r=%d",
+ F->k2.sysno, F->k2.seqno, F->m2, F->r2);
+ }
+ if ( (F->k1.sysno == F->k2.sysno) &&
+ (F->k1.seqno == F->k2.seqno) &&
+ (F->m1 != F->m2) &&
+ (F->r1 >0 ) && (F->r2 >0) )
+ { /* del-ins pair of same key (not eof) , ignore both */
+ if (F->is->method->debug > 9)
+ logf(LOG_LOG,"filt_fill: skipped %d.%d m=%d/%d r=%d/%d",
+ F->k1.sysno, F->k1.seqno,
+ F->m1,F->m2, F->r1,F->r2);
+ F->r1 = FILTER_NOTYET;
+ F->r2 = FILTER_NOTYET;
+ }
+ } /* while */
+} /* filter_fill */
+
+
+FILTER filter_open( ISAMD is, ISAMD_I data )
+{
+ FILTER F = (FILTER) xmalloc(sizeof(struct ISAMD_FILTER_s));
+ F->is = is;
+ F->data = data;
+ F->k1.sysno=0;
+ F->k1.seqno=0;
+ F->k2=F->k1;
+ F->m1 = F->m2 = 0;
+ F->r1 = F->r2 = FILTER_NOTYET;
+ filter_fill(F);
+ return F;
+}
+
+static void filter_close (FILTER F)
+{
+ xfree(F);
+}
+
+static int filter_read( FILTER F,
+ struct it_key *k,
+ int *mode)
+{
+ int res;
+ filter_fill(F);
+ if (F->is->method->debug > 9)
+ logf(LOG_LOG,"filt_read: reading %d.%d m=%d r=%d",
+ F->k1.sysno, F->k1.seqno, F->m1, F->r1);
+ res = F->r1;
+ if(res)
+ {
+ *k = F->k1;
+ *mode= F->m1;
+ }
+ F->r1 = FILTER_NOTYET;
+ return res;
+}
+
+static int filter_isempty(FILTER F)
+{
+ return ( (0 == F->r1) && (0 == F->r2)) ;
+}
+
+static int filter_only_one(FILTER F)
+{
+ return ( (0 != F->r1) && (0 == F->r2));
+}
+
+
+
+
+/***************************************************************
+ * Singleton encoding
+ ***************************************************************/
+/* When there is only a single item, we don't allocate a block
+ * for it, but code it in the directory entry directly, if it
+ * fits.
+ */
+
+#define DEC_SYSBITS 15
+#define DEC_SEQBITS 15
+#define DEC_MASK(n) ((1<<(n))-1)
+
+int is_singleton(ISAMD_P ipos)
+{
+ return ( ipos != 0 ) && ( ipos & 1 );
+}
+
+
+int singleton_encode(struct it_key *k)
+/* encodes the key into one int. If it does not fit, returns 0 */
+{
+ if ( (k->sysno & DEC_MASK(DEC_SYSBITS) ) != k->sysno )
+ return 0; /* no room dor sysno */
+ if ( (k->seqno & DEC_MASK(DEC_SYSBITS) ) != k->seqno )
+ return 0; /* no room dor sysno */
+ return ( (k->sysno | (k->seqno << DEC_SYSBITS) ) << 1) | 1;
+}
+
+void singleton_decode (int code, struct it_key *k)
+{
+ assert ((code & 1) == 1 );
+ code = code >> 1;
+ k->sysno = code & DEC_MASK(DEC_SYSBITS);
+ k->seqno = code >> DEC_SYSBITS ;
+}
+
+
+/***************************************************************
+ * General support routines
+ ***************************************************************/
+
+
+
static char *hexdump(unsigned char *p, int len, char *buff) {
static char localbuff[128];
char bytebuff[8];
}
-/* isamd - heikki's append-only isam
-*/
+static void isamd_reduceblock(ISAMD_PP pp)
+/* takes a large block, and reduces its category if possible */
+/* Presumably the first block in an isam-list */
+{
+ if (pp->pos)
+ return; /* existing block, do not touch */
+ /* TODO: Probably we may touch anyway? */
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_reduce: start p=%d c=%d sz=%d",
+ pp->pos, pp->cat, pp->size);
+ while ( ( pp->cat > 0 ) && (!pp->next) &&
+ (pp->offset < pp->is->method->filecat[pp->cat-1].bsize ) )
+ pp->cat--;
+ pp->pos = isamd_alloc_block(pp->is, pp->cat);
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_reduce: got p=%d c=%d sz=%d",
+ pp->pos, pp->cat, pp->size);
+} /* reduceblock */
-ISAMD_P isamd_append (ISAMD is, ISAMD_P ipos, ISAMD_I data)
+static int save_first_pp ( ISAMD_PP firstpp)
{
+ isamd_buildfirstblock(firstpp);
+ isamd_write_block(firstpp->is,firstpp->cat,firstpp->pos,firstpp->buf);
+ return isamd_addr(firstpp->pos,firstpp->cat);
+}
+
+
+static void save_last_pp (ISAMD_PP pp)
+{
+ pp->next = 0;/* just to be sure */
+ isamd_buildlaterblock(pp);
+ isamd_write_block(pp->is,pp->cat,pp->pos,pp->buf);
+}
+
+#ifdef UNUSED
+static int save_both_pps (ISAMD_PP firstpp, ISAMD_PP pp)
+{
+ /* order of things: Better to save firstpp first, if there are just two */
+ /* blocks, but last if there are blocks in between, as these have already */
+ /* been saved... optimise later (that's why this is in its own func...*/
+ int retval = save_first_pp(firstpp);
+ if (firstpp!=pp){
+ save_last_pp(pp);
+ isamd_pp_close(pp);
+ }
+ isamd_pp_close(firstpp);
+ return retval;
+} /* save_both_pps */
+#endif
+
+
+
+/***************************************************************
+ * Diffblock handling
+ ***************************************************************/
+
+void isamd_free_diffs(ISAMD_PP pp)
+{
+ int i;
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_free_diffs: pp=%p di=%p", pp, pp->diffinfo);
+ if (!pp->diffinfo)
+ return;
+ for (i=0;pp->diffinfo[i].difftype!=DT_NONE;i++)
+ if(pp->diffinfo[i].decodeData)
+ {
+ if (pp->is->method->debug > 8)
+ logf(LOG_LOG,"isamd_free_diffs [%d]=%p",i,
+ pp->diffinfo[i].decodeData);
+ (*pp->is->method->code_stop)(ISAMD_DECODE,pp->diffinfo[i].decodeData);
+ }
+ xfree(pp->diffinfo);
+ if (pp->diffbuf != pp->buf)
+ xfree (pp->diffbuf);
+ pp->diffbuf=0;
+ pp->diffinfo=0;
+} /* isamd_free_diffs */
+
+
+static void getDiffInfo(ISAMD_PP pp )
+{ /* builds the diff info structures from a diffblock */
+ int maxinfos = pp->is->method->filecat[pp->cat].bsize / 5 +2;
+ /* Each diff takes at least 5 bytes. Probably more, but this is safe */
+ int i=1; /* [0] is used for the main data, [n+1] for merge inputs */
+ int diffsz= maxinfos * sizeof(struct ISAMD_DIFF_s);
+ int maxsz = pp->is->method->filecat[pp->is->max_cat].bsize;
+ int diffidx = ISAMD_BLOCK_OFFSET_1;
+
+ pp->diffinfo = xmalloc( diffsz );
+ pp->offset = pp->size+1; /* used this block up */
+ memset(pp->diffinfo,'\0',diffsz);
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_getDiffInfo: %d=%d:%d->%d, ix=%d mx=%d",
+ isamd_addr(pp->pos, pp->cat), pp->cat, pp->pos, pp->next,
+ diffidx,maxinfos);
+
+ /* duplicate the buffer for diffs */
+ /* (so that we can read the next real buffer(s) */
+ assert(0==pp->diffbuf);
+ pp->diffbuf=xmalloc(maxsz);
+ memcpy(pp->diffbuf, pp->buf, maxsz);
+
+ pp->diffinfo[0].maxidx=-1; /* mark as special */
+ pp->diffinfo[0].difftype=DT_MAIN;
+
+ while (i<maxinfos)
+ {
+ if ( diffidx+sizeof(int) > pp->is->method->filecat[pp->cat].bsize )
+ {
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_getDiffInfo:Near end (no room for len) at ix=%d n=%d",
+ diffidx, i);
+ return; /* whole block done */
+ }
+ memcpy( &pp->diffinfo[i].maxidx, &pp->diffbuf[diffidx], sizeof(int) );
+ pp->diffinfo[i].difftype=DT_DIFF;
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_getDiffInfo: max=%d ix=%d dbuf=%p",
+ pp->diffinfo[i].maxidx, diffidx, pp->diffbuf);
+
+ if ( (pp->is->method->debug > 0) &&
+ (pp->diffinfo[i].maxidx > pp->is->method->filecat[pp->cat].bsize) )
+ {
+ logf(LOG_LOG,"Bad MaxIx!!! %s:%d: diffidx=%d",
+ __FILE__,__LINE__, diffidx);
+ logf(LOG_LOG,"i=%d maxix=%d bsz=%d", i, pp->diffinfo[i].maxidx,
+ pp->is->method->filecat[pp->cat].bsize);
+ logf(LOG_LOG,"pp=%d=%d:%d pp->nx=%d=%d:%d",
+ isamd_addr(pp->pos,pp->cat), pp->pos, pp->cat,
+ pp->next, isamd_type(pp->next), isamd_block(pp->next) );
+ }
+ assert(pp->diffinfo[i].maxidx <= pp->is->method->filecat[pp->cat].bsize+1);
+
+ if (0==pp->diffinfo[i].maxidx)
+ {
+ if (pp->is->method->debug > 5) //!!! 4
+ logf(LOG_LOG,"isamd_getDiffInfo:End mark at ix=%d n=%d",
+ diffidx, i);
+ return; /* end marker */
+ }
+ diffidx += sizeof(int);
+ pp->diffinfo[i].decodeData = (*pp->is->method->code_start)(ISAMD_DECODE);
+ pp->diffinfo[i].diffidx = diffidx;
+ if (pp->is->method->debug > 5)
+ logf(LOG_LOG,"isamd_getDiff[%d]:%d-%d %s",
+ i,diffidx-sizeof(int),pp->diffinfo[i].maxidx,
+ hexdump((char *)&pp->diffbuf[diffidx-4],8,0) );
+ diffidx=pp->diffinfo[i].maxidx;
+ if ( diffidx > pp->is->method->filecat[pp->cat].bsize )
+ return; /* whole block done */
+ ++i;
+ }
+ assert (!"too many diff sequences in the block");
+}
+
+/***************************************************************
+ * Main block operations
+ ***************************************************************/
+
+
+static ISAMD_PP get_new_main_block( ISAMD_PP firstpp, ISAMD_PP pp)
+{ /* allocates a new block for the main data, and links it in */
+ int newblock;
+ if (0 == firstpp->next)
+ { /* special case, pp not yet allocated. */
+ /*Started as largest size, that's fine */
+ pp->pos = isamd_alloc_block(pp->is,pp->cat);
+ firstpp->next = isamd_addr(pp->pos,pp->cat);
+ if (pp->is->method->debug >3)
+ logf(LOG_LOG,"isamd_build: Alloc 1. dblock p=%d=%d:%d",
+ isamd_addr(pp->pos,pp->cat), pp->cat, pp->pos);
+ }
+ newblock=isamd_alloc_block(pp->is,pp->cat);
+ pp->next=isamd_addr(newblock,pp->cat);
+ isamd_buildlaterblock(pp);
+ isamd_write_block(pp->is,pp->cat,pp->pos,pp->buf);
+ if (pp->is->method->debug >3)
+ logf(LOG_LOG,"isamd_build: Alloc nxt %d=%d:%d -> %d=%d:%d",
+ isamd_addr(pp->pos,pp->cat), pp->cat, pp->pos,
+ isamd_addr(newblock,pp->cat), pp->cat, newblock);
+ pp->next=0;
+ pp->pos=newblock;
+ pp->size=pp->offset=ISAMD_BLOCK_OFFSET_N;
+ return pp;
+} /* get_new_main_block */
+
+
+static ISAMD_PP append_main_item(ISAMD_PP firstpp,
+ ISAMD_PP pp,
+ struct it_key *i_key)
+{ /* appends one item in the main data block, allocates new if needed */
+ char *i_item= (char *) i_key; /* same as char */
+ char *i_ptr=i_item;
+ char codebuff[128];
+ char *c_ptr = codebuff;
+ int codelen;
+ char hexbuff[64];
+
+ int maxsize = pp->is->method->filecat[pp->is->max_cat].bsize;
+
+ c_ptr=codebuff;
+ i_ptr=i_item;
+ (*pp->is->method->code_item)(ISAMD_ENCODE, pp->decodeClientData,
+ &c_ptr, &i_ptr);
+ codelen = c_ptr - codebuff;
+ assert ( (codelen<128) && (codelen>0));
+ if (pp->is->method->debug >7)
+ logf(LOG_LOG,"isamd:build: coded %s nk=%d,ofs=%d-%d",
+ hexdump(codebuff, c_ptr-codebuff,hexbuff), firstpp->numKeys+1,
+ pp->offset, pp->offset+codelen);
+
+ if (pp->offset + codelen > maxsize )
+ { /* oops, block full - get a new one */
+ pp = get_new_main_block( firstpp, pp );
+ /* reset encoging and code again */
+ (*pp->is->method->code_reset)(pp->decodeClientData);
+ c_ptr=codebuff;
+ i_ptr=i_item;
+ (*pp->is->method->code_item)(ISAMD_ENCODE, pp->decodeClientData,
+ &c_ptr, &i_ptr);
+ codelen = c_ptr - codebuff;
+ assert ( (codelen<128) && (codelen>0));
+ if (pp->is->method->debug >7)
+ logf(LOG_LOG,"isamd:build: recoded into %s (nk=%d)",
+ hexdump(codebuff, c_ptr-codebuff,hexbuff), firstpp->numKeys+1);
+ } /* block full */
+
+ assert (pp->offset + codelen <= maxsize );
+
+ /* write the data into pp, now we must have room */
+ memcpy(&(pp->buf[pp->offset]),codebuff,codelen);
+ pp->offset += codelen;
+ pp->size += codelen;
+ firstpp->numKeys++;
+ /* clear the next 4 bytes in block, to avoid confusions with diff lens */
+ /* dirty, it should not be done here, but something slips somewhere, and */
+ /* I hope this fixes it... - Heikki */
+ codelen = pp->offset;
+ while ( (codelen < maxsize ) && (codelen <= pp->offset+4) )
+ pp->buf[codelen++] = '\0';
+ return pp;
+} /* append_main_item */
+
+
+/***************************************************************
+ * Read with merge
+ ***************************************************************/
+
+/* Reads one item and corrects for the diffs, if any */
+/* return 1 for ok, 0 for eof */
+int isamd_read_item_merge (
+ ISAMD_PP pp,
+ char **dst,
+ struct it_key *p_key, /* the data item that didn't fit*/
+ /* ISAMD_I data) */ /* more input data comes here */
+ FILTER filt) /* more input data comes here */
+{ /* The last two args can be null for ordinary reads */
+ char *keyptr;
+ char *codeptr;
+ char *codestart;
+ int winner=0; /* which diff holds the day */
+ int i; /* looping diffs */
+ int cmp;
+ int retry=1;
+ int oldoffs;
+ int rc;
+
+ if (!pp->diffinfo)
+ { /* first time */
+ getDiffInfo(pp);
+
+ for(i=1; pp->diffinfo[i].difftype!=DT_NONE; i++)
+ ; /* find last diff */
+ if (p_key)
+ { /* we have an extra item to inject into the merge */
+ if (pp->is->method->debug >9) //!!!!!
+ logf(LOG_LOG,"isamd_read_item: going to merge with %d.%d",
+ p_key->sysno, p_key->seqno);
+ pp->diffinfo[i].key = *p_key; /* the key merge could not handle */
+ pp->diffinfo[i].mode = pp->diffinfo[i].key.seqno & 1;
+ pp->diffinfo[i].key.seqno >>= 1;
+ pp->diffinfo[i].difftype=DT_INPU;
+ if (pp->is->method->debug > 7)
+ logf(LOG_LOG,"isamd_read_item: inpu key %d sys=%d seq=%d=2*%d+%d",
+ i, p_key->sysno,
+ pp->diffinfo[i].key.seqno*2 + pp->diffinfo[1].mode,
+ pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].mode);
+ p_key->sysno=p_key->seqno=0; /* used it up */
+ }
+
+ if (filt)
+ { /* we have a whole input stream to inject */
+ pp->diffinfo[i].difftype=DT_INPU;
+ }
+ } /* first time */
+
+ while (retry)
+
+ {
+ retry=0;
+ winner = 0;
+ for (i=0; (!retry) && (pp->diffinfo[i].difftype); i++)
+ {
+ if (0==pp->diffinfo[i].key.sysno)
+ {/* read a new one, if possible */
+ if ((pp->diffinfo[i].difftype==DT_DIFF) &&
+ (pp->diffinfo[i].diffidx < pp->diffinfo[i].maxidx))
+ { /* a normal kind of diff */
+ oldoffs=pp->diffinfo[i].diffidx;
+ codeptr= codestart = &(pp->diffbuf[pp->diffinfo[i].diffidx]);
+ keyptr=(char *)&(pp->diffinfo[i].key);
+ (*pp->is->method->code_item)(ISAMD_DECODE,
+ pp->diffinfo[i].decodeData, &keyptr, &codeptr);
+ pp->diffinfo[i].diffidx += codeptr-codestart;
+ pp->diffinfo[i].mode = pp->diffinfo[i].key.seqno & 1;
+ pp->diffinfo[i].key.seqno = pp->diffinfo[i].key.seqno >>1 ;
+ if (pp->is->method->debug > 9)
+ logf(LOG_LOG,"isamd_read_item: dif[%d] at %d-%d: %s",
+ i,oldoffs, pp->diffinfo[i].diffidx,
+ hexdump(pp->buf+oldoffs, pp->diffinfo[i].diffidx-oldoffs,0));
+ if (pp->is->method->debug > 7)
+ logf(LOG_LOG,"isamd_read_item: rd dif[%d] %d.%d (%x.%x)",
+ i,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno);
+ }
+ else if ( pp->diffinfo[i].difftype==DT_MAIN)
+ { /* read a main item */
+ assert(i==0); /* main data goes before any diffs */
+ oldoffs=pp->offset;
+ keyptr=(char*) &(pp->diffinfo[0].key);
+ rc= isamd_read_main_item(pp,&keyptr);
+ if (0==rc)
+ { /* eof */
+ if (pp->is->method->debug > 7)
+ logf(LOG_LOG,"isamd_read_item: eof (rc=%d) main ",
+ rc);
+ pp->diffinfo[i].maxidx=-1;
+ pp->diffinfo[i].key.sysno=0;
+ pp->diffinfo[i].key.seqno=0;
+ pp->diffinfo[i].difftype= DT_DONE;
+ }
+ else
+ { /* not eof */
+ pp->diffinfo[i].mode = 1;
+ if (pp->is->method->debug > 7)
+ logf(LOG_LOG,"isamd_read_item: rd main %d-%d %d.%d (%x.%x) m=%d",
+ oldoffs,pp->offset,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].mode);
+ } /* not eof */
+ }
+ else if (pp->diffinfo[i].difftype==DT_INPU)
+ {
+ keyptr = (char *) &pp->diffinfo[i].key;
+ /* rc = (*data->read_item)(data->clientData, &keyptr, &pp->diffinfo[i].mode); */
+ rc = filter_read(filt, &pp->diffinfo[i].key,
+ &pp->diffinfo[i].mode);
+ if (!rc)
+ { /* did not get it */
+ pp->diffinfo[i].key.sysno=0;
+ pp->diffinfo[i].maxidx=0; /* signal the end */
+ pp->diffinfo[i].difftype=DT_DONE;
+ }
+ if (pp->is->method->debug >7)
+ logf(LOG_LOG,"merge: read inpu m=%d %d.%d (%x.%x)",
+ pp->diffinfo[i].mode,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno );
+ } /* read an input item */
+ } /* read a new one */
+
+ if (pp->is->method->debug > 8)
+ logf(LOG_LOG,"isamd_read_item: considering d%d %d.%d ix=%d mx=%d",
+ i, pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].diffidx, pp->diffinfo[i].maxidx);
+
+ if ( 0!= pp->diffinfo[i].key.sysno)
+ { /* got a key, compare */
+ if (i!=winner)
+ cmp=key_compare(&pp->diffinfo[i].key, &pp->diffinfo[winner].key);
+ else
+ cmp=-1;
+ if (0==pp->diffinfo[winner].key.sysno)
+ cmp=-1; /* end of main sequence, take all diffs */
+ if (cmp<0)
+ {
+ if (pp->is->method->debug > 8)
+ logf(LOG_LOG,"isamd_read_item: ins [%d]%d.%d < [%d]%d.%d",
+ i,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ winner,
+ pp->diffinfo[winner].key.sysno, pp->diffinfo[winner].key.seqno);
+ if (pp->diffinfo[i].mode) /* insert diff, should always be */
+ winner = i;
+ else
+ {
+ if (pp->is->method->debug > 1)
+ logf(LOG_LOG,"delete diff for nonexisting item");
+ assert(!"delete diff for nonexisting item");
+ /* is an assert too steep here? Not really.*/
+ }
+ } /* earlier key */
+ else if (cmp==0)
+ {
+ if (!pp->diffinfo[i].mode) /* delete diff. should always be */
+ {
+ if (pp->is->method->debug > 8)
+ logf(LOG_LOG,"isamd_read_item: del %d at%d %d.%d (%x.%x)",
+ i, winner,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno);
+ pp->diffinfo[winner].key.sysno=0; /* delete it */
+ }
+ else
+ if (pp->is->method->debug > 2)
+ logf(LOG_LOG,"isamd_read_item: duplicate ins %d at%d %d.%d (%x.%x)",
+ i, winner,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno,
+ pp->diffinfo[i].key.sysno, pp->diffinfo[i].key.seqno);
+ /* skip the insert, since we already have it in the base */
+ /* Should we fail an assertion here??? */
+ pp->diffinfo[i].key.sysno=0; /* done with the delete */
+ retry=1; /* start all over again */
+ } /* matching key */
+ /* else it is a later key, its turn will come */
+ } /* got a key */
+ } /* for each diff */
+ } /* not retry */
+
+ if ( pp->diffinfo[winner].key.sysno)
+ {
+ if (pp->is->method->debug > 7)
+ logf(LOG_LOG,"isamd_read_item: got %d %d.%d (%x.%x)",
+ winner,
+ pp->diffinfo[winner].key.sysno, pp->diffinfo[winner].key.seqno,
+ pp->diffinfo[winner].key.sysno, pp->diffinfo[winner].key.seqno);
+ memcpy(*dst, &pp->diffinfo[winner].key, sizeof(struct it_key) );
+ *dst += sizeof(struct it_key);
+ pp->diffinfo[winner].key.sysno=0; /* used that one up */
+ cmp= 1;
+ }
+ else
+ {
+ if (pp->is->method->debug > 7)
+ logf(LOG_LOG,"isamd_read_item: eof w=%d %d.%d (%x.%x)",
+ winner,
+ pp->diffinfo[winner].key.sysno, pp->diffinfo[winner].key.seqno,
+ pp->diffinfo[winner].key.sysno, pp->diffinfo[winner].key.seqno);
+ assert(winner==0); /* if nothing found, nothing comes from a diff */
+ cmp= 0; /* eof */
+ }
+ if (pp->is->method->debug >9)
+ logf(LOG_LOG,"mergeDB4: sysno[1]=%d", pp->diffinfo[1].key.sysno); /*!*/
+ return cmp;
+
+} /* isamd_read_item */
+
+
+int isamd_read_item (ISAMD_PP pp, char **dst)
+{
+ return isamd_read_item_merge(pp,dst,0,0);
+}
+
+
+/***************************************************************
+ * Merge
+ ***************************************************************/
+
+static int merge ( ISAMD_PP firstpp, /* first pp (with diffs) */
+ struct it_key *p_key, /* the data item that didn't fit*/
+ /* ISAMD_I data) */ /* more input data comes here */
+ FILTER filt) /* more input data arriving here */
+{
+ int diffidx;
+ int killblk=0;
+ struct it_key r_key;
+ char * r_ptr;
+ int r_more = 1;
+ ISAMD_PP pp;
+ ISAMD_PP readpp=firstpp;
+ int retval=0;
+ int diffcat = firstpp->cat; /* keep the category of the diffblock even */
+ /* if it is going to be empty now. */
+ /* Alternative: Make it the minimal, and */
+ /* resize later. Saves disk, but will lead */
+ /* into bad seeks. */
+
+ ++(readpp->is->files[0].no_merges);
+
+ /* set up diffs as they should be for reading */
+ diffidx = ISAMD_BLOCK_OFFSET_1;
+ //readpp->diffbuf=readpp->buf; // diffinfo has to duplicate it!
+ //getDiffInfo(readpp); // first read will make the diffinfo, at init
+
+ if (readpp->is->method->debug >4)
+ logf(LOG_LOG,"isamd_merge: f=%d=%d:%d n=%d=%d:%d",
+ isamd_addr(firstpp->pos,firstpp->cat), firstpp->cat, firstpp->pos,
+ firstpp->next, isamd_type(firstpp->next), isamd_block(firstpp->next));
+
+ /* release our data block. Do before reading, when pos is stable ! */
+ killblk=firstpp->pos;
+ if (killblk)
+ {
+ isamd_release_block(firstpp->is, firstpp->cat, killblk);
+ if (readpp->is->method->debug >3)
+ logf(LOG_LOG,"isamd_merge: released old firstblock %d (%d:%d)",
+ isamd_addr(killblk,firstpp->cat), firstpp->cat, killblk );
+ }
+
+
+ r_ptr= (char *) &r_key;
+/* r_more = isamd_read_item_merge( readpp, &r_ptr, p_key, data); */
+ r_more = isamd_read_item_merge( readpp, &r_ptr, p_key, filt);
+ if (!r_more)
+ { /* oops, all data has been deleted! what to do??? */
+ /* never mind, we have at least one more delta to add to the block */
+ /* pray that is not a delete as well... */
+ r_key.sysno = 0;
+ r_key.seqno = 0;
+ if (readpp->is->method->debug >5)
+ logf(LOG_LOG,"isamd_merge:all data has been deleted (nk=%d) ",
+ readpp->numKeys);
+ //assert (readpp->numKeys == 0); /* no longer true! */
+ }
+
+
+ /* set up the new blocks for simple writing */
+ firstpp=isamd_pp_open(readpp->is,isamd_addr(0, diffcat));
+ firstpp->pos=isamd_alloc_block(firstpp->is,diffcat);
+ if (readpp->is->method->debug >3)
+ logf(LOG_LOG,"isamd_merge: allocated new firstpp %d=%d:%d",
+ isamd_addr(firstpp->pos,firstpp->cat), firstpp->cat, firstpp->pos );
+
+ pp=isamd_pp_open(readpp->is,isamd_addr(0,readpp->is->max_cat) );
+ pp->offset=pp->size=ISAMD_BLOCK_OFFSET_N;
+
+ while (r_more)
+ {
+ if (readpp->is->method->debug >6)
+ logf(LOG_LOG,"isamd_merge: got key %d.%d",
+ r_key.sysno, r_key.seqno );
+ pp= append_main_item(firstpp, pp, &r_key);
+
+ if ( (readpp->pos != killblk ) && (0!=readpp->pos) )
+ { /* pos can get to 0 at end of main seq, if still diffs left...*/
+ if (readpp->is->method->debug >3)
+ logf(LOG_LOG,"isamd_merge: released block %d (%d:%d) now %d=%d:%d",
+ isamd_addr(killblk,readpp->cat), readpp->cat, killblk,
+ isamd_addr(readpp->pos,readpp->cat),readpp->cat, readpp->pos );
+ isamd_release_block(readpp->is, readpp->cat, readpp->pos);
+ killblk=readpp->pos;
+ }
+
+ /* (try to) read next item */
+ r_ptr= (char *) &r_key;
+ r_more = isamd_read_item_merge( readpp, &r_ptr,0,filt);
+
+ } /* while read */
+
+
+// firstpp->diffs=0;
+
+
+ isamd_reduceblock(pp); /* reduce size if possible */
+ if (0==firstpp->next)
+ firstpp->next = isamd_addr(pp->pos,pp->cat);
+ save_last_pp(pp);
+ if (readpp->is->method->debug >4)
+ logf(LOG_LOG,"isamd_merge: saved last block %d=%d:%d",
+ isamd_addr(pp->pos,pp->cat), pp->cat, pp->pos);
+ isamd_pp_close(pp);
+
+ if (readpp->is->method->debug >5)
+ logf(LOG_LOG,"isamd_merge: closing readpp %d=%d:%d di=%p",
+ isamd_addr(readpp->pos,readpp->cat), readpp->cat, readpp->pos,
+ readpp->diffinfo);
+ isamd_pp_close(readpp); /* pos is 0 by now, at eof. close works anyway */
+
+ if (readpp->is->method->debug >2)
+ logf(LOG_LOG,"isamd_merge: merge ret f=%d=%d:%d pp=%d=%d:%d",
+ isamd_addr(firstpp->pos,pp->cat), firstpp->cat, firstpp->pos,
+ isamd_addr(pp->pos,pp->cat), pp->cat, pp->pos);
+
+ firstpp->size = firstpp->offset = ISAMD_BLOCK_OFFSET_1; /* nothing there */
+ memset(firstpp->buf,'\0',firstpp->is->method->filecat[firstpp->cat].bsize);
+ save_first_pp(firstpp);
+ retval = isamd_addr(firstpp->pos, firstpp->cat);
+ isamd_pp_close(firstpp);
+
+ return retval;
+
+} /* merge */
+
+
+
+
+/***************************************************************
+ * Appending diffs
+ ***************************************************************/
+
+
+
+static int append_diffs(
+ ISAMD is,
+ ISAMD_P ipos,
+ /*ISAMD_I data)*/
+ FILTER filt)
+{
+ struct it_key i_key; /* one input item */
+ char *i_item = (char *) &i_key; /* same as chars */
+ char *i_ptr=i_item;
+ int i_more =1;
+ int i_mode; /* 0 for delete, 1 for insert */
+
+ ISAMD_PP firstpp;
+ char hexbuff[64];
+ int diffidx=0;
+ int maxsize=0;
+ int difflenidx;
+ char codebuff[128];
+ char *c_ptr = codebuff;
+ int codelen;
+ int merge_rc;
+ int retval=0;
+
+ if (0==ipos)
+ {
+ firstpp=isamd_pp_open(is, isamd_addr(0,0) );
+ firstpp->size=firstpp->offset=ISAMD_BLOCK_OFFSET_1;
+ /* create in smallest category, will expand later */
+ ++(is->files[0].no_fbuilds);
+ }
+ else
+ {
+ firstpp=isamd_pp_open(is, ipos);
+ ++(is->files[0].no_appds);
+ }
+
+ if (is->method->debug >2)
+ logf(LOG_LOG,"isamd_appd: Start ipos=%d=%d:%d n=%d=%d:%d nk=%d",
+ ipos, isamd_type(ipos), isamd_block(ipos),
+ firstpp->next, isamd_type(firstpp->next), isamd_block(firstpp->next),
+ firstpp->numKeys);
+ maxsize = is->method->filecat[firstpp->cat].bsize;
+
+ difflenidx = diffidx = firstpp->size;
+
+ diffidx+=sizeof(int); /* difflen will be stored here */
+
+ /* read first input */
+ //i_ptr = i_item; //!!!
+ i_more = filter_read(filt, &i_key, &i_mode);
+ /* i_more = (*data->read_item)(data->clientData, &i_ptr, &i_mode); */
+
+ if (is->method->debug >6)
+ logf(LOG_LOG,"isamd_appd: start m=%d %d.%d=%x.%x: %d",
+ i_mode,
+ i_key.sysno, i_key.seqno,
+ i_key.sysno, i_key.seqno,
+ i_key.sysno*2+i_mode);
+
+ while (i_more)
+ {
+ /* store the mode bit inside key */
+ assert( ((i_key.seqno<<1)>>1) == i_key.seqno); /* can spare the bit */
+ i_key.seqno = i_key.seqno * 2 + i_mode;
+
+ c_ptr=codebuff;
+ i_ptr=i_item;
+ (*is->method->code_item)(ISAMD_ENCODE, firstpp->decodeClientData,
+ &c_ptr, &i_ptr);
+ codelen = c_ptr - codebuff;
+ assert ( (codelen<128) && (codelen>0));
+ if (is->method->debug >7)
+ logf(LOG_LOG,"isamd_appd: coded %d: %s (nk=%d) (ix=%d)",
+ codelen, hexdump(codebuff, codelen,hexbuff),
+ firstpp->numKeys,diffidx);
+
+ if (diffidx + codelen > maxsize )
+ { /* block full */
+ if (firstpp->cat < firstpp->is->max_cat)
+ { /* just increase the block size */
+ if (firstpp->pos > 0) /* free the old block if allocated */
+ isamd_release_block(is, firstpp->cat, firstpp->pos);
+ ++firstpp->cat;
+ maxsize = is->method->filecat[firstpp->cat].bsize;
+ firstpp->pos=0; /* need to allocate it when saving */
+ if (is->method->debug >3)
+ logf(LOG_LOG,"isamd_appd: increased diff block to %d (%d)",
+ firstpp->cat, maxsize);
+ }
+ else
+ { /* max size already - can't help, need to merge it */
+ if (is->method->debug >7)
+ logf(LOG_LOG,"isamd_appd: block full");
+ if (is->method->debug >9) //!!!!!
+ logf(LOG_LOG,"isamd_appd: going to merge with m=%d %d.%d",
+ i_mode, i_key.sysno, i_key.seqno);
+ merge_rc = merge (firstpp, &i_key, filt);
+ if (0!=merge_rc)
+ return merge_rc; /* merge handled them all ! */
+ assert(!"merge returned zero ??");
+ } /* need to merge */
+ } /* block full */
+
+ assert ( diffidx+codelen <= maxsize );
+
+ /* save the diff */
+ memcpy(&(firstpp->buf[diffidx]),codebuff,codelen);
+ diffidx += codelen;
+ firstpp->size = firstpp->offset = diffidx;
+
+ if (i_mode)
+ firstpp->numKeys++; /* insert diff */
+ else
+ firstpp->numKeys--; /* delete diff */
+
+ /* update length of this diff run */
+ memcpy(&(firstpp->buf[difflenidx]),&diffidx,sizeof(diffidx));
+
+ /* (try to) read the next input */
+ i_ptr = i_item;
+ i_more = filter_read(filt, &i_key, &i_mode);
+ /* i_more = (*data->read_item)(data->clientData, &i_ptr, &i_mode); */
+ if ( (i_more) && (is->method->debug >6) )
+ logf(LOG_LOG,"isamd_appd: got m=%d %d.%d=%x.%x: %d",
+ i_mode,
+ i_key.sysno, i_key.seqno,
+ i_key.sysno, i_key.seqno,
+ i_key.sysno*2+i_mode);
+ } /* more loop */
+
+ /* clear the next difflen, if room for such */
+ difflenidx = diffidx;
+ while ( (difflenidx-diffidx<=sizeof(int)+1) && (difflenidx<maxsize))
+ firstpp->buf[difflenidx++]='\0';
+
+ if (0==firstpp->pos) /* need to (re)alloc the block */
+ firstpp->pos = isamd_alloc_block(is, firstpp->cat);
+
+ retval = save_first_pp( firstpp );
+ isamd_pp_close(firstpp);
-} /* isamh_append */
+ return retval;
+} /* append_diffs */
+
+
+
+
+/*************************************************************
+ * isamd_append itself
+ *************************************************************/
+
+ISAMD_P isamd_append (ISAMD is, ISAMD_P ipos, ISAMD_I data)
+{
+ FILTER F = filter_open(is,data);
+ ISAMD_P rc=0;
+
+ if ( filter_isempty(F) ) /* can be, if del-ins of the same */
+ {
+ if (is->method->debug >9)
+ logf(LOG_LOG,"isamd_appd: nothing to do");
+ filter_close(F);
+ return ipos; /* without doing anything at all */
+ }
+
+ if ( ( 0==ipos) && filter_only_one(F) )
+ {
+ struct it_key k;
+ int mode;
+ filter_read(F,&k,&mode);
+ assert(mode);
+ rc = singleton_encode(&k);
+ if (is->method->debug >9)
+ logf(LOG_LOG,"isamd_appd: singleton %d (%x)",
+ rc,rc);
+ assert ( (rc==0) || is_singleton(rc) );
+ }
+ if ( 0==rc) /* either not single, or it did not fit */
+ {
+ rc = append_diffs(is,ipos,F)<<1; /* not singleton */
+ assert ( ! is_singleton(rc) ); /*!*/
+ }
+ filter_close(F);
+
+ if (is->method->debug >9)
+ logf(LOG_LOG,"isamd_appd: ret %d=%x (%d=%x)",
+ rc,rc,ipos,ipos);
+
+ return rc;
+} /* isamd_append */
+
+
+
+
+
/*
* $Log: merge-d.c,v $
- * Revision 1.1 1999-07-14 13:14:47 heikki
+ * Revision 1.22 1999-09-23 18:01:18 heikki
+ * singleton optimising
+ *
+ * Revision 1.21 1999/09/21 17:36:43 heikki
+ * Added filter function. Not much of effect on the small test set...
+ *
+ * Revision 1.20 1999/09/20 15:48:06 heikki
+ * Small changes
+ *
+ * Revision 1.19 1999/09/13 13:28:28 heikki
+ * isam-d optimizing: merging input data in the same go
+ *
+ * Revision 1.18 1999/08/25 18:09:24 heikki
+ * Starting to optimize
+ *
+ * Revision 1.17 1999/08/24 13:17:42 heikki
+ * Block sizes, comments
+ *
+ * Revision 1.16 1999/08/24 10:12:02 heikki
+ * Comments about optimising
+ *
+ * Revision 1.15 1999/08/22 08:26:34 heikki
+ * COmments
+ *
+ * Revision 1.14 1999/08/20 12:25:58 heikki
+ * Statistics in isamd
+ *
+ * Revision 1.13 1999/08/18 13:59:19 heikki
+ * Fixed another unlikely difflen bug
+ *
+ * Revision 1.12 1999/08/18 13:28:17 heikki
+ * Set log levels to decent values
+ *
+ * Revision 1.11 1999/08/18 10:37:11 heikki
+ * Fixed (another) difflen bug
+ *
+ * Revision 1.10 1999/08/18 09:13:31 heikki
+ * Fixed a detail
+ *
+ * Revision 1.9 1999/08/17 19:46:53 heikki
+ * Fixed a memory leak
+ *
+ * Revision 1.8 1999/08/07 11:30:59 heikki
+ * Bug fixing (still a mem leak somewhere)
+ *
+ * Revision 1.7 1999/08/04 14:21:18 heikki
+ * isam-d seems to be working.
+ *
+ * Revision 1.6 1999/07/23 15:43:05 heikki
+ * Hunted a few bugs in isam-d. Still crashes on the long test run
+ *
+ * Revision 1.5 1999/07/23 13:58:52 heikki
+ * merged closer to working, still fails on filling a separate, large block
+ *
+ * Revision 1.4 1999/07/21 14:53:55 heikki
+ * isamd read and write functions work, except when block full
+ * Merge missing still. Need to split some functions
+ *
+ * Revision 1.1 1999/07/14 13:14:47 heikki
* Created empty
*
*