1 | /* |
2 | * Legal Notice |
3 | * |
4 | * This document and associated source code (the "Work") is a part of a |
5 | * benchmark specification maintained by the TPC. |
6 | * |
7 | * The TPC reserves all right, title, and interest to the Work as provided |
8 | * under U.S. and international laws, including without limitation all patent |
9 | * and trademark rights therein. |
10 | * |
11 | * No Warranty |
12 | * |
13 | * 1.1 TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, THE INFORMATION |
14 | * CONTAINED HEREIN IS PROVIDED "AS IS" AND WITH ALL FAULTS, AND THE |
15 | * AUTHORS AND DEVELOPERS OF THE WORK HEREBY DISCLAIM ALL OTHER |
16 | * WARRANTIES AND CONDITIONS, EITHER EXPRESS, IMPLIED OR STATUTORY, |
17 | * INCLUDING, BUT NOT LIMITED TO, ANY (IF ANY) IMPLIED WARRANTIES, |
18 | * DUTIES OR CONDITIONS OF MERCHANTABILITY, OF FITNESS FOR A PARTICULAR |
19 | * PURPOSE, OF ACCURACY OR COMPLETENESS OF RESPONSES, OF RESULTS, OF |
20 | * WORKMANLIKE EFFORT, OF LACK OF VIRUSES, AND OF LACK OF NEGLIGENCE. |
21 | * ALSO, THERE IS NO WARRANTY OR CONDITION OF TITLE, QUIET ENJOYMENT, |
22 | * QUIET POSSESSION, CORRESPONDENCE TO DESCRIPTION OR NON-INFRINGEMENT |
23 | * WITH REGARD TO THE WORK. |
24 | * 1.2 IN NO EVENT WILL ANY AUTHOR OR DEVELOPER OF THE WORK BE LIABLE TO |
25 | * ANY OTHER PARTY FOR ANY DAMAGES, INCLUDING BUT NOT LIMITED TO THE |
26 | * COST OF PROCURING SUBSTITUTE GOODS OR SERVICES, LOST PROFITS, LOSS |
27 | * OF USE, LOSS OF DATA, OR ANY INCIDENTAL, CONSEQUENTIAL, DIRECT, |
28 | * INDIRECT, OR SPECIAL DAMAGES WHETHER UNDER CONTRACT, TORT, WARRANTY, |
29 | * OR OTHERWISE, ARISING IN ANY WAY OUT OF THIS OR ANY OTHER AGREEMENT |
30 | * RELATING TO THE WORK, WHETHER OR NOT SUCH AUTHOR OR DEVELOPER HAD |
31 | * ADVANCE NOTICE OF THE POSSIBILITY OF SUCH DAMAGES. |
32 | * |
33 | * Contributors: |
34 | * Gradient Systems |
35 | */ |
36 | #include "config.h" |
37 | #include "porting.h" |
38 | #include <stdio.h> |
39 | #include <fcntl.h> |
40 | #include <assert.h> |
41 | #ifdef WIN32 |
42 | #include <io.h> |
43 | #include <stdlib.h> |
44 | #include <search.h> |
45 | #else |
46 | #include <sys/types.h> |
47 | #include <sys/stat.h> |
48 | #include <netinet/in.h> |
49 | #endif |
50 | #ifdef NCR |
51 | #include <sys/types.h> |
52 | #endif |
53 | #ifndef USE_STDLIB_H |
54 | #include <malloc.h> |
55 | #endif |
56 | #include "config.h" |
57 | #include "decimal.h" |
58 | #include "date.h" |
59 | #include "dist.h" |
60 | #include "genrand.h" |
61 | #include "error_msg.h" |
62 | #include "r_params.h" |
63 | #include "dcomp.h" |
64 | #ifdef TEST |
65 | option_t options[] = {{"DISTRIBUTIONS" , OPT_STR, 2, "read distributions from file <s>" , NULL, "tester_dist.idx" }, NULL}; |
66 | |
67 | char params[2]; |
68 | struct { |
69 | char *name; |
70 | } tdefs[] = {NULL}; |
71 | #endif |
72 | |
73 | /* NOTE: these need to be in sync with a_dist.h */ |
74 | #define D_NAME_LEN 20 |
75 | #define FL_LOADED 0x01 |
76 | static int load_dist(d_idx_t *d); |
77 | |
78 | /* |
79 | * Routine: di_compare() |
80 | * Purpose: comparison routine for two d_idx_t entries; used by qsort |
81 | * Algorithm: |
82 | * Data Structures: |
83 | * |
84 | * Params: |
85 | * Returns: |
86 | * Called By: |
87 | * Calls: |
88 | * Assumptions: |
89 | * Side Effects: |
90 | * TODO: None |
91 | */ |
92 | int di_compare(const void *op1, const void *op2) { |
93 | d_idx_t *ie1 = (d_idx_t *)op1, *ie2 = (d_idx_t *)op2; |
94 | |
95 | return (strcasecmp(ie1->name, ie2->name)); |
96 | } |
97 | |
98 | /* |
99 | * Routine: find_dist(char *name) |
100 | * Purpose: translate from dist_t name to d_idx_t * |
101 | * Algorithm: |
102 | * Data Structures: |
103 | * |
104 | * Params: |
105 | * Returns: |
106 | * Called By: |
107 | * Calls: |
108 | * Assumptions: |
109 | * Side Effects: |
110 | * TODO: None |
111 | */ |
112 | d_idx_t *find_dist(char *name) { |
113 | static int index_loaded = 0, entry_count; |
114 | static d_idx_t *idx = NULL; |
115 | d_idx_t key, *id = NULL; |
116 | int i; |
117 | FILE *ifp; |
118 | int32_t temp; |
119 | |
120 | /* load the index if this is the first time through */ |
121 | if (!index_loaded) { |
122 | /* make sure that this is read one thread at a time */ |
123 | if (!index_loaded) /* make sure no one beat us to it */ |
124 | { |
125 | |
126 | /* open the dist file */ |
127 | if ((ifp = fopen(get_str("DISTRIBUTIONS" ), "rb" )) == NULL) { |
128 | fprintf(stderr, "Error: open of distributions failed: " ); |
129 | perror(get_str("DISTRIBUTIONS" )); |
130 | exit(1); |
131 | } |
132 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
133 | fprintf(stderr, "Error: read of index count failed: " ); |
134 | perror(get_str("DISTRIBUTIONS" )); |
135 | exit(2); |
136 | } |
137 | entry_count = ntohl(temp); |
138 | if ((temp = fseek(ifp, -entry_count * IDX_SIZE, SEEK_END)) < 0) { |
139 | fprintf(stderr, "Error: lseek to index failed: " ); |
140 | fprintf(stderr, "attempting to reach %d\nSystem error: " , (int)(-entry_count * IDX_SIZE)); |
141 | perror(get_str("DISTRIBUTIONS" )); |
142 | exit(3); |
143 | } |
144 | idx = (d_idx_t *)malloc(entry_count * sizeof(d_idx_t)); |
145 | MALLOC_CHECK(idx); |
146 | for (i = 0; i < entry_count; i++) { |
147 | memset(idx + i, 0, sizeof(d_idx_t)); |
148 | if (fread(idx[i].name, 1, D_NAME_LEN, ifp) < D_NAME_LEN) { |
149 | fprintf(stderr, "Error: read index failed (1): " ); |
150 | perror(get_str("DISTRIBUTIONS" )); |
151 | exit(2); |
152 | } |
153 | idx[i].name[D_NAME_LEN] = '\0'; |
154 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
155 | fprintf(stderr, "Error: read index failed (2): " ); |
156 | perror(get_str("DISTRIBUTIONS" )); |
157 | exit(2); |
158 | } |
159 | idx[i].index = ntohl(temp); |
160 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
161 | fprintf(stderr, "Error: read index failed (4): " ); |
162 | perror(get_str("DISTRIBUTIONS" )); |
163 | exit(2); |
164 | } |
165 | idx[i].offset = ntohl(temp); |
166 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
167 | fprintf(stderr, "Error: read index failed (5): " ); |
168 | perror(get_str("DISTRIBUTIONS" )); |
169 | exit(2); |
170 | } |
171 | idx[i].str_space = ntohl(temp); |
172 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
173 | fprintf(stderr, "Error: read index failed (6): " ); |
174 | perror(get_str("DISTRIBUTIONS" )); |
175 | exit(2); |
176 | } |
177 | idx[i].length = ntohl(temp); |
178 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
179 | fprintf(stderr, "Error: read index failed (7): " ); |
180 | perror(get_str("DISTRIBUTIONS" )); |
181 | exit(2); |
182 | } |
183 | idx[i].w_width = ntohl(temp); |
184 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
185 | fprintf(stderr, "Error: read index failed (8): " ); |
186 | perror(get_str("DISTRIBUTIONS" )); |
187 | exit(2); |
188 | } |
189 | idx[i].v_width = ntohl(temp); |
190 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
191 | fprintf(stderr, "Error: read index failed (9): " ); |
192 | perror(get_str("DISTRIBUTIONS" )); |
193 | exit(2); |
194 | } |
195 | idx[i].name_space = ntohl(temp); |
196 | idx[i].dist = NULL; |
197 | } |
198 | qsort((void *)idx, entry_count, sizeof(d_idx_t), di_compare); |
199 | index_loaded = 1; |
200 | |
201 | /* make sure that this is read one thread at a time */ |
202 | fclose(ifp); |
203 | } |
204 | } |
205 | |
206 | /* find the distribution, if it exists and move to it */ |
207 | strcpy(key.name, name); |
208 | id = (d_idx_t *)bsearch((void *)&key, (void *)idx, entry_count, sizeof(d_idx_t), di_compare); |
209 | if (id != NULL) /* found a valid distribution */ |
210 | if (id->flags != FL_LOADED) /* but it needs to be loaded */ |
211 | load_dist(id); |
212 | |
213 | return (id); |
214 | } |
215 | |
216 | /* |
217 | * Routine: load_dist(int fd, dist_t *d) |
218 | * Purpose: load a particular distribution |
219 | * Algorithm: |
220 | * Data Structures: |
221 | * |
222 | * Params: |
223 | * Returns: |
224 | * Called By: |
225 | * Calls: |
226 | * Assumptions: |
227 | * Side Effects: |
228 | * TODO: None |
229 | */ |
230 | static int load_dist(d_idx_t *di) { |
231 | int res = 0, i, j; |
232 | dist_t *d; |
233 | int32_t temp; |
234 | FILE *ifp; |
235 | |
236 | if (di->flags != FL_LOADED) /* make sure no one beat us to it */ |
237 | { |
238 | if ((ifp = fopen(get_str("DISTRIBUTIONS" ), "rb" )) == NULL) { |
239 | fprintf(stderr, "Error: open of distributions failed: " ); |
240 | perror(get_str("DISTRIBUTIONS" )); |
241 | exit(1); |
242 | } |
243 | |
244 | if ((temp = fseek(ifp, di->offset, SEEK_SET)) < 0) { |
245 | fprintf(stderr, "Error: lseek to distribution failed: " ); |
246 | perror("load_dist()" ); |
247 | exit(2); |
248 | } |
249 | |
250 | di->dist = (dist_t *)malloc(sizeof(struct DIST_T)); |
251 | MALLOC_CHECK(di->dist); |
252 | d = di->dist; |
253 | |
254 | // fprintf(stderr, "\ndist %s ", di->name); |
255 | |
256 | /* load the type information */ |
257 | d->type_vector = (int *)malloc(sizeof(int32_t) * di->v_width); |
258 | MALLOC_CHECK(d->type_vector); |
259 | for (i = 0; i < di->v_width; i++) { |
260 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
261 | fprintf(stderr, "Error: read of type vector failed for '%s': " , di->name); |
262 | perror("load_dist()" ); |
263 | exit(3); |
264 | } |
265 | d->type_vector[i] = ntohl(temp); |
266 | // fprintf(stderr, "type %d, ", d->type_vector[i]); |
267 | } |
268 | |
269 | /* load the weights */ |
270 | d->weight_sets = (int **)malloc(sizeof(int *) * di->w_width); |
271 | d->maximums = (int *)malloc(sizeof(int32_t) * di->w_width); |
272 | MALLOC_CHECK(d->weight_sets); |
273 | MALLOC_CHECK(d->maximums); |
274 | for (i = 0; i < di->w_width; i++) { |
275 | *(d->weight_sets + i) = (int *)malloc(di->length * sizeof(int32_t)); |
276 | MALLOC_CHECK(*(d->weight_sets + i)); |
277 | d->maximums[i] = 0; |
278 | for (j = 0; j < di->length; j++) { |
279 | if (fread(&temp, 1, sizeof(int32_t), ifp) < 0) { |
280 | fprintf(stderr, "Error: read of weights failed: " ); |
281 | perror("load_dist()" ); |
282 | exit(3); |
283 | } |
284 | *(*(d->weight_sets + i) + j) = ntohl(temp); |
285 | /* calculate the maximum weight and convert sets to cummulative |
286 | */ |
287 | d->maximums[i] += d->weight_sets[i][j]; |
288 | d->weight_sets[i][j] = d->maximums[i]; |
289 | } |
290 | } |
291 | |
292 | /* load the value offsets */ |
293 | d->value_sets = (int **)malloc(sizeof(int *) * di->v_width); |
294 | MALLOC_CHECK(d->value_sets); |
295 | for (i = 0; i < di->v_width; i++) { |
296 | *(d->value_sets + i) = (int *)malloc(di->length * sizeof(int32_t)); |
297 | MALLOC_CHECK(*(d->value_sets + i)); |
298 | for (j = 0; j < di->length; j++) { |
299 | if (fread(&temp, 1, sizeof(int32_t), ifp) != sizeof(int32_t)) { |
300 | fprintf(stderr, "Error: read of values failed: " ); |
301 | perror("load_dist()" ); |
302 | exit(4); |
303 | } |
304 | *(*(d->value_sets + i) + j) = ntohl(temp); |
305 | } |
306 | } |
307 | |
308 | /* load the column aliases, if they were defined */ |
309 | if (di->name_space) { |
310 | d->names = (char *)malloc(di->name_space); |
311 | MALLOC_CHECK(d->names); |
312 | if (fread(d->names, 1, di->name_space * sizeof(char), ifp) < 0) { |
313 | fprintf(stderr, "Error: read of names failed: " ); |
314 | perror("load_dist()" ); |
315 | exit(599); |
316 | } |
317 | } |
318 | |
319 | /* and finally the values themselves */ |
320 | d->strings = (char *)malloc(sizeof(char) * di->str_space); |
321 | MALLOC_CHECK(d->strings); |
322 | if (fread(d->strings, 1, di->str_space * sizeof(char), ifp) < 0) { |
323 | fprintf(stderr, "Error: read of strings failed: " ); |
324 | perror("load_dist()" ); |
325 | exit(5); |
326 | } |
327 | |
328 | fclose(ifp); |
329 | |
330 | // |
331 | // fprintf(stderr, "%s {\n", di->name); |
332 | // |
333 | // // type_vector |
334 | // fprintf(stderr, "{"); |
335 | // |
336 | // for (int i = 0 ; i < di->v_width; i++) { |
337 | // fprintf(stderr, "%d", d->type_vector[i]); |
338 | // |
339 | // if (i < di->v_width-1) { |
340 | // fprintf(stderr, ", "); |
341 | // |
342 | // } |
343 | // } |
344 | // fprintf(stderr, "},"); |
345 | // |
346 | //// weight_sets |
347 | // fprintf(stderr, "{"); |
348 | // for (int i = 0 ; i < di->w_width; i++) { |
349 | // fprintf(stderr, "{"); |
350 | // |
351 | // for (int j = 0 ; j < di->length; j++) { |
352 | // |
353 | // fprintf(stderr, "%d", d->weight_sets[i][j]); |
354 | // |
355 | // |
356 | // |
357 | // if (j < di->length-1) { |
358 | // fprintf(stderr, ", "); |
359 | // |
360 | // } |
361 | // } |
362 | // fprintf(stderr, "},"); |
363 | // |
364 | // |
365 | // |
366 | // if (i < di->w_width-1) { |
367 | // fprintf(stderr, ", "); |
368 | // |
369 | // } |
370 | // } |
371 | // fprintf(stderr, "},"); |
372 | // |
373 | // |
374 | // // maximums |
375 | // fprintf(stderr, "{"); |
376 | // |
377 | // for (int i = 0 ; i < di->w_width; i++) { |
378 | // fprintf(stderr, "%d", d->maximums[i]); |
379 | // |
380 | // if (i < di->w_width-1) { |
381 | // fprintf(stderr, ", "); |
382 | // |
383 | // } |
384 | // } |
385 | // fprintf(stderr, "},"); |
386 | // |
387 | // |
388 | // |
389 | // // value sets |
390 | // fprintf(stderr, "{"); |
391 | // for (int i = 0 ; i < di->v_width; i++) { |
392 | // fprintf(stderr, "{"); |
393 | // |
394 | // for (int j = 0 ; j < di->length; j++) { |
395 | // |
396 | // fprintf(stderr, "%d", d->value_sets[i][j]); |
397 | // |
398 | // |
399 | // |
400 | // if (j < di->length-1) { |
401 | // fprintf(stderr, ", "); |
402 | // |
403 | // } |
404 | // } |
405 | // fprintf(stderr, "},"); |
406 | // |
407 | // |
408 | // |
409 | // if (i < di->v_width-1) { |
410 | // fprintf(stderr, ", "); |
411 | // |
412 | // } |
413 | // } |
414 | // fprintf(stderr, "},"); |
415 | // |
416 | // |
417 | //// strings |
418 | // |
419 | // |
420 | // fprintf(stderr, "{"); |
421 | // |
422 | // for (int i = 0 ; i < di->str_space; i++) { |
423 | // fprintf(stderr, "%d", (int) d->strings[i]); |
424 | // |
425 | // if (i < di->str_space-1) { |
426 | // fprintf(stderr, ", "); |
427 | // |
428 | // } |
429 | // } |
430 | // fprintf(stderr, "},"); |
431 | // |
432 | // |
433 | // |
434 | // // names |
435 | // |
436 | // fprintf(stderr, "{"); |
437 | // |
438 | // for (int i = 0 ; i < di->name_space; i++) { |
439 | // fprintf(stderr, "%d", (int) d->names[i]); |
440 | // |
441 | // if (i < di->name_space-1) { |
442 | // fprintf(stderr, ", "); |
443 | // |
444 | // } |
445 | // } |
446 | // fprintf(stderr, "},"); |
447 | // |
448 | // |
449 | // |
450 | // // size |
451 | // fprintf(stderr, "%d}\n", d->size); |
452 | |
453 | di->flags = FL_LOADED; |
454 | } |
455 | |
456 | return (res); |
457 | } |
458 | |
459 | /* |
460 | * Routine: void *dist_op() |
461 | * Purpose: select a value/weight from a distribution |
462 | * Algorithm: |
463 | * Data Structures: |
464 | * |
465 | * Params: char *d_name |
466 | * int vset: which set of values |
467 | * int wset: which set of weights |
468 | * Returns: appropriate data type cast as a void * |
469 | * Called By: |
470 | * Calls: |
471 | * Assumptions: |
472 | * Side Effects: |
473 | * TODO: 20000317 Need to be sure this is portable to NT and others |
474 | */ |
475 | int dist_op(void *dest, int op, char *d_name, int vset, int wset, int stream) { |
476 | d_idx_t *d; |
477 | dist_t *dist; |
478 | int level, index = 0, dt; |
479 | char *char_val; |
480 | int i_res = 1; |
481 | |
482 | if ((d = find_dist(d_name)) == NULL) { |
483 | char msg[80]; |
484 | sprintf(msg, "Invalid distribution name '%s'" , d_name); |
485 | INTERNAL(msg); |
486 | assert(d != NULL); |
487 | } |
488 | |
489 | dist = d->dist; |
490 | |
491 | if (op == 0) { |
492 | genrand_integer(&level, DIST_UNIFORM, 1, dist->maximums[wset - 1], 0, stream); |
493 | while (level > dist->weight_sets[wset - 1][index] && index < d->length) |
494 | index += 1; |
495 | dt = vset - 1; |
496 | if ((index >= d->length) || (dt > d->v_width)) |
497 | INTERNAL("Distribution overrun" ); |
498 | char_val = dist->strings + dist->value_sets[dt][index]; |
499 | } else { |
500 | index = vset - 1; |
501 | dt = wset - 1; |
502 | if (index >= d->length || index < 0) { |
503 | fprintf(stderr, "Runtime ERROR: Distribution over-run/under-run\n" ); |
504 | fprintf(stderr, "Check distribution definitions and usage for %s.\n" , d->name); |
505 | fprintf(stderr, "index = %d, length=%d.\n" , index, d->length); |
506 | exit(1); |
507 | } |
508 | char_val = dist->strings + dist->value_sets[dt][index]; |
509 | } |
510 | |
511 | switch (dist->type_vector[dt]) { |
512 | case TKN_VARCHAR: |
513 | if (dest) |
514 | *(char **)dest = (char *)char_val; |
515 | break; |
516 | case TKN_INT: |
517 | i_res = atoi(char_val); |
518 | if (dest) |
519 | *(int *)dest = i_res; |
520 | break; |
521 | case TKN_DATE: |
522 | if (dest == NULL) { |
523 | dest = (date_t *)malloc(sizeof(date_t)); |
524 | MALLOC_CHECK(dest); |
525 | } |
526 | strtodt(*(date_t **)dest, char_val); |
527 | break; |
528 | case TKN_DECIMAL: |
529 | if (dest == NULL) { |
530 | dest = (decimal_t *)malloc(sizeof(decimal_t)); |
531 | MALLOC_CHECK(dest); |
532 | } |
533 | strtodec(*(decimal_t **)dest, char_val); |
534 | break; |
535 | } |
536 | |
537 | return ((dest == NULL) ? i_res : index + 1); /* shift back to the 1-based indexing scheme */ |
538 | } |
539 | |
540 | /* |
541 | * Routine: int dist_weight |
542 | * Purpose: return the weight of a particular member of a distribution |
543 | * Algorithm: |
544 | * Data Structures: |
545 | * |
546 | * Params: distribution *d |
547 | * int index: which "row" |
548 | * int wset: which set of weights |
549 | * Returns: |
550 | * Called By: |
551 | * Calls: |
552 | * Assumptions: |
553 | * Side Effects: |
554 | * TODO: |
555 | * 20000405 need to add error checking |
556 | */ |
557 | int dist_weight(int *dest, char *d, int index, int wset) { |
558 | d_idx_t *d_idx; |
559 | dist_t *dist; |
560 | int res; |
561 | |
562 | if ((d_idx = find_dist(d)) == NULL) { |
563 | char msg[80]; |
564 | sprintf(msg, "Invalid distribution name '%s'" , d); |
565 | INTERNAL(msg); |
566 | } |
567 | |
568 | dist = d_idx->dist; |
569 | |
570 | res = dist->weight_sets[wset - 1][index - 1]; |
571 | /* reverse the accumulation of weights */ |
572 | if (index > 1) |
573 | res -= dist->weight_sets[wset - 1][index - 2]; |
574 | |
575 | if (dest == NULL) |
576 | return (res); |
577 | |
578 | *dest = res; |
579 | |
580 | return (0); |
581 | } |
582 | |
583 | /* |
584 | * Routine: int DistNameIndex() |
585 | * Purpose: return the index of a column alias |
586 | * Algorithm: |
587 | * Data Structures: |
588 | * |
589 | * Params: |
590 | * Returns: |
591 | * Called By: |
592 | * Calls: |
593 | * Assumptions: |
594 | * Side Effects: |
595 | * TODO: |
596 | */ |
597 | int DistNameIndex(char *szDist, int nNameType, char *szName) { |
598 | d_idx_t *d_idx; |
599 | dist_t *dist; |
600 | int res; |
601 | char *cp = NULL; |
602 | |
603 | if ((d_idx = find_dist(szDist)) == NULL) |
604 | return (-1); |
605 | dist = d_idx->dist; |
606 | |
607 | if (dist->names == NULL) |
608 | return (-1); |
609 | |
610 | res = 0; |
611 | cp = dist->names; |
612 | do { |
613 | if (strcasecmp(szName, cp) == 0) |
614 | break; |
615 | cp += strlen(cp) + 1; |
616 | res += 1; |
617 | } while (res < (d_idx->v_width + d_idx->w_width)); |
618 | |
619 | if (res >= 0) { |
620 | if ((nNameType == VALUE_NAME) && (res < d_idx->v_width)) |
621 | return (res + 1); |
622 | if ((nNameType == WEIGHT_NAME) && (res > d_idx->v_width)) |
623 | return (res - d_idx->v_width + 1); |
624 | } |
625 | |
626 | return (-1); |
627 | } |
628 | |
629 | /* |
630 | * Routine: int distsize(char *name) |
631 | * Purpose: return the size of a distribution |
632 | * Algorithm: |
633 | * Data Structures: |
634 | * |
635 | * Params: |
636 | * Returns: |
637 | * Called By: |
638 | * Calls: |
639 | * Assumptions: |
640 | * Side Effects: |
641 | * TODO: |
642 | * 20000405 need to add error checking |
643 | */ |
644 | int distsize(char *name) { |
645 | d_idx_t *dist; |
646 | |
647 | dist = find_dist(name); |
648 | |
649 | if (dist == NULL) |
650 | return (-1); |
651 | |
652 | return (dist->length); |
653 | } |
654 | |
655 | /* |
656 | * Routine: int IntegrateDist(char *szDistName, int nPct, int nStartIndex, int |
657 | *nWeightSet) Purpose: return the index of the entry which, starting from |
658 | *nStartIndex, would create a range comprising nPct of the total contained in |
659 | *nWeightSet NOTE: the value can "wrap" -- that is, the returned value can be |
660 | *less than nStartIndex Algorithm: Data Structures: |
661 | * |
662 | * Params: |
663 | * Returns: |
664 | * Called By: |
665 | * Calls: |
666 | * Assumptions: |
667 | * Side Effects: |
668 | * TODO: |
669 | */ |
670 | |
671 | int IntegrateDist(char *szDistName, int nPct, int nStartIndex, int nWeightSet) { |
672 | d_idx_t *pDistIndex; |
673 | int nGoal, nSize; |
674 | |
675 | if ((nPct <= 0) || (nPct >= 100)) |
676 | return (QERR_RANGE_ERROR); |
677 | |
678 | pDistIndex = find_dist(szDistName); |
679 | if (pDistIndex == NULL) |
680 | return (QERR_BAD_NAME); |
681 | |
682 | if (nStartIndex > pDistIndex->length) |
683 | return (QERR_RANGE_ERROR); |
684 | |
685 | nGoal = pDistIndex->dist->maximums[nWeightSet]; |
686 | nGoal = nGoal * nPct / 100; |
687 | nSize = distsize(szDistName); |
688 | |
689 | while (nGoal >= 0) { |
690 | nStartIndex++; |
691 | nGoal -= dist_weight(NULL, szDistName, nStartIndex % nSize, nWeightSet); |
692 | } |
693 | |
694 | return (nStartIndex); |
695 | } |
696 | |
697 | /* |
698 | * Routine: int dist_type(char *name, int nValueSet) |
699 | * Purpose: return the type of the n-th value set in a distribution |
700 | * Algorithm: |
701 | * Data Structures: |
702 | * |
703 | * Params: |
704 | * Returns: |
705 | * Called By: |
706 | * Calls: |
707 | * Assumptions: |
708 | * Side Effects: |
709 | * TODO: |
710 | */ |
711 | int dist_type(char *name, int nValueSet) { |
712 | d_idx_t *dist; |
713 | |
714 | dist = find_dist(name); |
715 | |
716 | if (dist == NULL) |
717 | return (-1); |
718 | |
719 | if (nValueSet < 1 || nValueSet > dist->v_width) |
720 | return (-1); |
721 | |
722 | return (dist->dist->type_vector[nValueSet - 1]); |
723 | } |
724 | |
725 | /* |
726 | * Routine: |
727 | * Purpose: |
728 | * Algorithm: |
729 | * Data Structures: |
730 | * |
731 | * Params: |
732 | * Returns: |
733 | * Called By: |
734 | * Calls: |
735 | * Assumptions: |
736 | * Side Effects: |
737 | * TODO: None |
738 | */ |
739 | void dump_dist(char *name) { |
740 | d_idx_t *pIndex; |
741 | int i, j; |
742 | char *pCharVal = NULL; |
743 | int nVal; |
744 | |
745 | pIndex = find_dist(name); |
746 | if (pIndex == NULL) |
747 | ReportErrorNoLine(QERR_BAD_NAME, name, 1); |
748 | printf("create %s;\n" , pIndex->name); |
749 | printf("set types = (" ); |
750 | for (i = 0; i < pIndex->v_width; i++) { |
751 | if (i > 0) |
752 | printf(", " ); |
753 | printf("%s" , dist_type(name, i + 1) == 7 ? "int" : "varchar" ); |
754 | } |
755 | printf(");\n" ); |
756 | printf("set weights = %d;\n" , pIndex->w_width); |
757 | for (i = 0; i < pIndex->length; i++) { |
758 | printf("add(" ); |
759 | for (j = 0; j < pIndex->v_width; j++) { |
760 | if (j) |
761 | printf(", " ); |
762 | if (dist_type(name, j + 1) != 7) { |
763 | dist_member(&pCharVal, name, i + 1, j + 1); |
764 | printf("\"%s\"" , pCharVal); |
765 | } else { |
766 | dist_member(&nVal, name, i + 1, j + 1); |
767 | printf("%d" , nVal); |
768 | } |
769 | } |
770 | printf("; " ); |
771 | for (j = 0; j < pIndex->w_width; j++) { |
772 | if (j) |
773 | printf(", " ); |
774 | printf("%d" , dist_weight(NULL, name, i + 1, j + 1)); |
775 | } |
776 | printf(");\n" ); |
777 | } |
778 | |
779 | return; |
780 | } |
781 | |
782 | /* |
783 | * Routine: dist_active(char *szName, int nWeightSet) |
784 | * Purpose: return number of entries with non-zero weght values |
785 | * Algorithm: |
786 | * Data Structures: |
787 | * |
788 | * Params: |
789 | * Returns: |
790 | * Called By: |
791 | * Calls: |
792 | * Assumptions: |
793 | * Side Effects: |
794 | * TODO: None |
795 | */ |
796 | int dist_active(char *szName, int nWeightSet) { |
797 | int nSize, nResult = 0, i; |
798 | |
799 | nSize = distsize(szName); |
800 | for (i = 1; i <= nSize; i++) { |
801 | if (dist_weight(NULL, szName, i, nWeightSet) != 0) |
802 | nResult += 1; |
803 | } |
804 | |
805 | return (nResult); |
806 | } |
807 | |
808 | /* |
809 | * Routine: DistSizeToShiftWidth(char *szDist) |
810 | * Purpose: Determine the number of bits required to select a member of the |
811 | * distribution Algorithm: Data Structures: |
812 | * |
813 | * Params: |
814 | * Returns: |
815 | * Called By: |
816 | * Calls: |
817 | * Assumptions: |
818 | * Side Effects: |
819 | * TODO: None |
820 | */ |
821 | int DistSizeToShiftWidth(char *szDist, int nWeightSet) { |
822 | int nBits = 1, nTotal = 2, nMax; |
823 | d_idx_t *d; |
824 | |
825 | d = find_dist(szDist); |
826 | nMax = dist_max(d->dist, nWeightSet); |
827 | |
828 | while (nTotal < nMax) { |
829 | nBits += 1; |
830 | nTotal <<= 1; |
831 | } |
832 | |
833 | return (nBits); |
834 | } |
835 | |
836 | /* |
837 | * Routine: |
838 | * Purpose: |
839 | * Algorithm: |
840 | * Data Structures: |
841 | * |
842 | * Params: |
843 | * Returns: |
844 | * Called By: |
845 | * Calls: |
846 | * Assumptions: |
847 | * Side Effects: |
848 | * TODO: None |
849 | */ |
850 | int MatchDistWeight(void *dest, char *szDist, int nWeight, int nWeightSet, int ValueSet) { |
851 | d_idx_t *d; |
852 | dist_t *dist; |
853 | int index = 0, dt, i_res, nRetcode; |
854 | char *char_val; |
855 | |
856 | if ((d = find_dist(szDist)) == NULL) { |
857 | char msg[80]; |
858 | sprintf(msg, "Invalid distribution name '%s'" , szDist); |
859 | INTERNAL(msg); |
860 | } |
861 | |
862 | dist = d->dist; |
863 | nWeight %= dist->maximums[nWeightSet - 1]; |
864 | |
865 | while (nWeight > dist->weight_sets[nWeightSet - 1][index] && index < d->length) |
866 | index += 1; |
867 | dt = ValueSet - 1; |
868 | if (index >= d->length) |
869 | index = d->length - 1; |
870 | char_val = dist->strings + dist->value_sets[dt][index]; |
871 | |
872 | switch (dist->type_vector[dt]) { |
873 | case TKN_VARCHAR: |
874 | if (dest) |
875 | *(char **)dest = (char *)char_val; |
876 | break; |
877 | case TKN_INT: |
878 | i_res = atoi(char_val); |
879 | if (dest) |
880 | *(int *)dest = i_res; |
881 | break; |
882 | case TKN_DATE: |
883 | if (dest == NULL) { |
884 | dest = (date_t *)malloc(sizeof(date_t)); |
885 | MALLOC_CHECK(dest); |
886 | } |
887 | strtodt(*(date_t **)dest, char_val); |
888 | break; |
889 | case TKN_DECIMAL: |
890 | if (dest == NULL) { |
891 | dest = (decimal_t *)malloc(sizeof(decimal_t)); |
892 | MALLOC_CHECK(dest); |
893 | } |
894 | strtodec(*(decimal_t **)dest, char_val); |
895 | break; |
896 | } |
897 | |
898 | nRetcode = 1; |
899 | index = 1; |
900 | while (index < dist->maximums[nWeightSet - 1]) { |
901 | nRetcode += 1; |
902 | index *= 2; |
903 | } |
904 | |
905 | return (nRetcode); |
906 | } |
907 | |
908 | /* |
909 | * Routine: findDistValue(char *szValue, char *szDistName, int nValueSet) |
910 | * Purpose: Return the row number where the entry is found |
911 | * Algorithm: |
912 | * Data Structures: |
913 | * |
914 | * Params: |
915 | * Returns: |
916 | * Called By: |
917 | * Calls: |
918 | * Assumptions: |
919 | * Side Effects: |
920 | * TODO: |
921 | * 20031024 jms this routine needs to handle all data types, not just varchar |
922 | */ |
923 | int findDistValue(char *szValue, char *szDistName, int ValueSet) { |
924 | int nRetValue = 1, nDistMax; |
925 | char szDistValue[128]; |
926 | |
927 | nDistMax = distsize(szDistName); |
928 | |
929 | for (nRetValue = 1; nRetValue < nDistMax; nRetValue++) { |
930 | dist_member(&szDistValue, szDistName, nRetValue, ValueSet); |
931 | if (strcmp(szValue, szDistValue) == 0) |
932 | break; |
933 | } |
934 | |
935 | if (nRetValue <= nDistMax) |
936 | return (nRetValue); |
937 | return (-1); |
938 | } |
939 | |
940 | #ifdef TEST |
941 | main() { |
942 | int i_res; |
943 | char *c_res; |
944 | decimal_t dec_res; |
945 | |
946 | init_params(); |
947 | |
948 | dist_member(&i_res, "test_dist" , 1, 1); |
949 | if (i_res != 10) { |
950 | printf("dist_member(\"test_dist\", 1, 1): %d != 10\n" , i_res); |
951 | exit(1); |
952 | } |
953 | dist_member(&i_res, "test_dist" , 1, 2); |
954 | if (i_res != 60) { |
955 | printf("dist_member(\"test_dist\", 1, 2): %d != 60\n" , i_res); |
956 | exit(1); |
957 | } |
958 | dist_member((void *)&c_res, "test_dist" , 1, 3); |
959 | if (strcmp(c_res, "El Camino" )) { |
960 | printf("dist_member(\"test_dist\", 1, 3): %s != El Camino\n" , c_res); |
961 | exit(1); |
962 | } |
963 | dist_member((void *)&dec_res, "test_dist" , 1, 4); |
964 | if (strcmp(dec_res.number, "1" ) || strcmp(dec_res.fraction, "23" )) { |
965 | printf("dist_member(\"test_dist\", 1, 4): %s.%s != 1.23\n" , dec_res.number, dec_res.fraction); |
966 | exit(1); |
967 | } |
968 | dist_weight(&i_res, "test_dist" , 2, 2); |
969 | if (3 != i_res) { |
970 | printf("dist_weight(\"test_dist\", 2, 2): %d != 3\n" , i_res); |
971 | exit(1); |
972 | } |
973 | } |
974 | #endif /* TEST */ |
975 | |