1 |
/* |
2 |
* integrate Hyper Estraier into PostgreSQL |
3 |
* |
4 |
* Dobrica Pavlinusic <dpavlin@rot13.org> 2005-05-19 |
5 |
* |
6 |
* TODO: |
7 |
* - all |
8 |
* |
9 |
* NOTES: |
10 |
* - clear structures with memset to support hash indexes (who whould like |
11 |
* to create hash index on table returned from function?) |
12 |
* - number of returned rows is set by PostgreSQL evaluator, see: |
13 |
* http://archives.postgresql.org/pgsql-hackers/2005-02/msg00546.php |
14 |
* |
15 |
* Based on: |
16 |
* - C example from PostgreSQL documentation (BSD licence) |
17 |
* - example002.c from Hyper Estraier (GPL) |
18 |
* - _textin/_textout from pgcurl.c (LGPL) |
19 |
* |
20 |
* This code is licenced under GPL |
21 |
*/ |
22 |
|
23 |
#include "postgres.h" |
24 |
#include "fmgr.h" |
25 |
#include "funcapi.h" |
26 |
#include "utils/builtins.h" |
27 |
#include "utils/array.h" |
28 |
#include "miscadmin.h" |
29 |
#include <estraier.h> |
30 |
#include <cabin.h> |
31 |
|
32 |
#define _textin(str) DirectFunctionCall1(textin, CStringGetDatum(str)) |
33 |
#define _textout(str) DatumGetPointer(DirectFunctionCall1(textout, PointerGetDatum(str))) |
34 |
#define GET_STR(textp) DatumGetCString(DirectFunctionCall1(textout, PointerGetDatum(textp))) |
35 |
#define GET_TEXT(cstrp) DatumGetTextP(DirectFunctionCall1(textin, CStringGetDatum(cstrp))) |
36 |
|
37 |
/* prototype */ |
38 |
char *attr2text(ESTDOC *doc, char *attr); |
39 |
|
40 |
ESTDB *db; |
41 |
ESTCOND *cond; |
42 |
ESTDOC *doc; |
43 |
const CBLIST *texts; |
44 |
int ecode, *est_result, resnum, i, j; |
45 |
int limit = 0; |
46 |
int offset = 0; |
47 |
|
48 |
/* define PostgreSQL v1 function */ |
49 |
PG_FUNCTION_INFO_V1(pgest); |
50 |
Datum pgest(PG_FUNCTION_ARGS) { |
51 |
|
52 |
FuncCallContext *funcctx; |
53 |
int call_cntr; |
54 |
int max_calls; |
55 |
TupleDesc tupdesc; |
56 |
TupleTableSlot *slot; |
57 |
AttInMetadata *attinmeta; |
58 |
char *index_path; |
59 |
char *query; |
60 |
char *attr; |
61 |
|
62 |
/* stuff done only on the first call of the function */ |
63 |
if (SRF_IS_FIRSTCALL()) { |
64 |
MemoryContext oldcontext; |
65 |
|
66 |
/* take arguments from function */ |
67 |
//index_path = _textout(PG_GETARG_TEXT_P(0)); |
68 |
index_path = _textout(PG_GETARG_TEXT_P(0)); |
69 |
query = _textout(PG_GETARG_TEXT_P(1)); |
70 |
attr = _textout(PG_GETARG_TEXT_P(2)); |
71 |
limit = PG_GETARG_INT32(3); |
72 |
offset = PG_GETARG_INT32(4); |
73 |
|
74 |
/* create a function context for cross-call persistence */ |
75 |
funcctx = SRF_FIRSTCALL_INIT(); |
76 |
|
77 |
/* switch to memory context appropriate for multiple function calls */ |
78 |
oldcontext = MemoryContextSwitchTo(funcctx->multi_call_memory_ctx); |
79 |
|
80 |
/* open the database */ |
81 |
elog(DEBUG1, "pgest: est_db_open(%s)", index_path); |
82 |
|
83 |
if(!(db = est_db_open(index_path, ESTDBREADER, &ecode))){ |
84 |
elog(ERROR, "est_db_open: can't open %s [%d]: %s", index_path, ecode, est_err_msg(ecode)); |
85 |
SRF_RETURN_DONE(funcctx); |
86 |
} |
87 |
|
88 |
elog(INFO, "pgest: query[%s] attr[%s] limit %d offset %d", query, attr, limit, offset); |
89 |
|
90 |
/* create a search condition object */ |
91 |
if (!(cond = est_cond_new())) { |
92 |
elog(INFO, "pgest: est_cond_new failed"); |
93 |
SRF_RETURN_DONE(funcctx); |
94 |
} |
95 |
|
96 |
/* set the search phrase to the search condition object */ |
97 |
if (strlen(query) > 0) |
98 |
est_cond_set_phrase(cond, query); |
99 |
|
100 |
/* minimum valid attribute length is 10: @a STREQ a */ |
101 |
if (attr != NULL && strlen(attr) >= 10) { |
102 |
elog(INFO,"est_cond_add_attr(%s)", attr); |
103 |
est_cond_add_attr(cond, attr); |
104 |
} |
105 |
|
106 |
/* get the result of search */ |
107 |
est_result = est_db_search(db, cond, &resnum, NULL); |
108 |
|
109 |
/* total number of tuples to be returned */ |
110 |
if (limit && limit < resnum) { |
111 |
funcctx->max_calls = limit - offset; |
112 |
} else { |
113 |
funcctx->max_calls = resnum - offset; |
114 |
} |
115 |
|
116 |
/* check if results exists */ |
117 |
if ( 0 == funcctx->max_calls ) |
118 |
elog(INFO, "pgest: no results for: %s", query ); |
119 |
|
120 |
elog(DEBUG1, "pgest: found %d hits for %s", resnum, query); |
121 |
|
122 |
/* Build a tuple description for a __pgest tuple */ |
123 |
tupdesc = RelationNameGetTupleDesc("__pgest"); |
124 |
|
125 |
/* allocate a slot for a tuple with this tupdesc */ |
126 |
slot = TupleDescGetSlot(tupdesc); |
127 |
|
128 |
/* assign slot to function context */ |
129 |
funcctx->slot = slot; |
130 |
|
131 |
/* |
132 |
* generate attribute metadata needed later to produce tuples from raw |
133 |
* C strings |
134 |
*/ |
135 |
attinmeta = TupleDescGetAttInMetadata(tupdesc); |
136 |
funcctx->attinmeta = attinmeta; |
137 |
|
138 |
MemoryContextSwitchTo(oldcontext); |
139 |
|
140 |
elog(DEBUG1, "SRF_IS_FIRSTCALL done"); |
141 |
} |
142 |
|
143 |
/* stuff done on every call of the function */ |
144 |
funcctx = SRF_PERCALL_SETUP(); |
145 |
|
146 |
call_cntr = funcctx->call_cntr; |
147 |
max_calls = funcctx->max_calls; |
148 |
slot = funcctx->slot; |
149 |
attinmeta = funcctx->attinmeta; |
150 |
|
151 |
if (limit && call_cntr > limit - 1) { |
152 |
elog(INFO, "call_cntr: %d limit: %d", call_cntr, limit); |
153 |
SRF_RETURN_DONE(funcctx); |
154 |
} |
155 |
|
156 |
if (call_cntr < max_calls) { |
157 |
char **values; |
158 |
HeapTuple tuple; |
159 |
Datum result; |
160 |
|
161 |
elog(DEBUG1, "pgest: loop count %d", call_cntr); |
162 |
|
163 |
if (! est_result) { |
164 |
elog(ERROR, "pgest: no estraier results"); |
165 |
SRF_RETURN_DONE(funcctx); |
166 |
} |
167 |
|
168 |
/* |
169 |
* Prepare a values array for storage in our slot. |
170 |
* This should be an array of C strings which will |
171 |
* be processed later by the type input functions. |
172 |
*/ |
173 |
|
174 |
if (doc = est_db_get_doc(db, est_result[call_cntr + offset], 0)) { |
175 |
|
176 |
elog(DEBUG1, "URI: %s\n Title: %s\n", |
177 |
est_doc_attr(doc, "@uri"), |
178 |
est_doc_attr(doc, "@title") |
179 |
); |
180 |
|
181 |
values = (char **) palloc(4 * sizeof(char *)); |
182 |
|
183 |
// values[0] = (char *) palloc(strlen(_estval) * sizeof(char)); |
184 |
|
185 |
values[0] = (char *) attr2text(doc,"@id"); |
186 |
values[1] = (char *) attr2text(doc,"@uri"); |
187 |
values[2] = (char *) attr2text(doc,"@title"); |
188 |
values[3] = (char *) attr2text(doc,"@type"); |
189 |
|
190 |
/* destloy the document object */ |
191 |
elog(DEBUG2, "est_doc_delete"); |
192 |
est_doc_delete(doc); |
193 |
} else { |
194 |
elog(INFO, "no result from estraier"); |
195 |
values[0] = DatumGetCString( "" ); |
196 |
values[1] = DatumGetCString( "" ); |
197 |
values[2] = DatumGetCString( "" ); |
198 |
values[3] = DatumGetCString( "" ); |
199 |
} |
200 |
|
201 |
|
202 |
elog(DEBUG2, "build tuple"); |
203 |
/* build a tuple */ |
204 |
tuple = BuildTupleFromCStrings(attinmeta, values); |
205 |
|
206 |
elog(DEBUG2, "make tuple into datum"); |
207 |
/* make the tuple into a datum */ |
208 |
result = TupleGetDatum(slot, tuple); |
209 |
|
210 |
elog(DEBUG2, "cleanup"); |
211 |
/* clean up ? */ |
212 |
/* |
213 |
pfree(values[0]); |
214 |
pfree(values[1]); |
215 |
pfree(values[2]); |
216 |
pfree(values[3]); |
217 |
pfree(values); |
218 |
*/ |
219 |
|
220 |
elog(DEBUG2, "cleanup over"); |
221 |
|
222 |
SRF_RETURN_NEXT(funcctx, result); |
223 |
} else { |
224 |
elog(DEBUG1, "loop over"); |
225 |
|
226 |
if(!est_db_close(db, &ecode)){ |
227 |
elog(INFO, "est_db_close error: %s", est_err_msg(ecode)); |
228 |
} |
229 |
|
230 |
/* do when there is no more left */ |
231 |
SRF_RETURN_DONE(funcctx); |
232 |
} |
233 |
} |
234 |
|
235 |
/* work in progress */ |
236 |
PG_FUNCTION_INFO_V1(pgest2); |
237 |
Datum pgest2(PG_FUNCTION_ARGS) |
238 |
{ |
239 |
int nrows = 3; |
240 |
int16 typlen; |
241 |
bool typbyval; |
242 |
char typalign; |
243 |
ReturnSetInfo *rsinfo = (ReturnSetInfo *) fcinfo->resultinfo; |
244 |
AttInMetadata *attinmeta; |
245 |
TupleDesc tupdesc; |
246 |
Tuplestorestate *tupstore = NULL; |
247 |
HeapTuple tuple; |
248 |
MemoryContext per_query_ctx; |
249 |
MemoryContext oldcontext; |
250 |
Datum dvalue; |
251 |
char **values; |
252 |
int ncols; |
253 |
int i, j; |
254 |
|
255 |
/* check to see if caller supports us returning a tuplestore */ |
256 |
if (!rsinfo || !(rsinfo->allowedModes & SFRM_Materialize)) |
257 |
ereport(ERROR, |
258 |
(errcode(ERRCODE_SYNTAX_ERROR), |
259 |
errmsg("materialize mode required, but it is not " \ |
260 |
"allowed in this context"))); |
261 |
|
262 |
/* get the requested return tuple description */ |
263 |
tupdesc = rsinfo->expectedDesc; |
264 |
ncols = tupdesc->natts; |
265 |
|
266 |
/* |
267 |
* The requested tuple description better match up with the array |
268 |
* we were given. |
269 |
*/ |
270 |
/* OK, use it */ |
271 |
attinmeta = TupleDescGetAttInMetadata(tupdesc); |
272 |
|
273 |
/* Now go to work */ |
274 |
rsinfo->returnMode = SFRM_Materialize; |
275 |
|
276 |
per_query_ctx = fcinfo->flinfo->fn_mcxt; |
277 |
oldcontext = MemoryContextSwitchTo(per_query_ctx); |
278 |
|
279 |
/* initialize our tuplestore */ |
280 |
tupstore = tuplestore_begin_heap(true, false, SortMem); |
281 |
|
282 |
values = (char **) palloc(ncols * sizeof(char *)); |
283 |
|
284 |
for (i = 0; i < nrows; i++) |
285 |
{ |
286 |
for (j = 0; j < ncols; j++) |
287 |
{ |
288 |
values[j] = DatumGetCString( "foo" ); |
289 |
} |
290 |
/* construct the tuple */ |
291 |
tuple = BuildTupleFromCStrings(attinmeta, values); |
292 |
|
293 |
/* now store it */ |
294 |
tuplestore_puttuple(tupstore, tuple); |
295 |
} |
296 |
|
297 |
tuplestore_donestoring(tupstore); |
298 |
rsinfo->setResult = tupstore; |
299 |
|
300 |
/* |
301 |
* SFRM_Materialize mode expects us to return a NULL Datum. The actual |
302 |
* tuples are in our tuplestore and passed back through |
303 |
* rsinfo->setResult. rsinfo->setDesc is set to the tuple description |
304 |
* that we actually used to build our tuples with, so the caller can |
305 |
* verify we did what it was expecting. |
306 |
*/ |
307 |
rsinfo->setDesc = tupdesc; |
308 |
MemoryContextSwitchTo(oldcontext); |
309 |
|
310 |
return (Datum) 0; |
311 |
} |
312 |
|
313 |
|
314 |
/* make text var from attr */ |
315 |
char *attr2text(ESTDOC *doc, char *attr) { |
316 |
char *val; |
317 |
const char *attrval; |
318 |
int len; |
319 |
int attrlen; |
320 |
|
321 |
elog(DEBUG1, "doc: %08x, attr: %s", doc, attr); |
322 |
|
323 |
if ( (attrval = est_doc_attr(doc, attr)) && (attrlen = strlen(attrval)) ) { |
324 |
val = (char *) palloc(attrlen * sizeof(char)); |
325 |
} else { |
326 |
return (Datum) NULL; |
327 |
} |
328 |
|
329 |
len = strlen(attrval); |
330 |
elog(DEBUG1, "attr2text(%s) = '%s' %d bytes", attr, attrval, len); |
331 |
|
332 |
len++; |
333 |
len *= sizeof(char); |
334 |
|
335 |
elog(DEBUG2, "palloc(%d)", len); |
336 |
|
337 |
val = palloc(len); |
338 |
|
339 |
memset(val, 0, len); |
340 |
strncpy(val, attrval, len); |
341 |
|
342 |
elog(DEBUG2, "val=%s", val); |
343 |
|
344 |
return val; |
345 |
} |
346 |
|
347 |
/* make integer variable from property */ |
348 |
/* |
349 |
char *prop2int(SW_RESULT sw_res, char *propname) { |
350 |
char *val; |
351 |
unsigned long prop; |
352 |
int len; |
353 |
|
354 |
elog(DEBUG2, "prop2int(%s)", propname); |
355 |
|
356 |
prop = estResultPropertyULong( sw_res, propname ); |
357 |
if (error_or_abort( est_handle )) return NULL; |
358 |
|
359 |
elog(DEBUG1, "prop2int(%s) = %lu", propname, prop); |
360 |
|
361 |
len = 128 * sizeof(char); |
362 |
elog(DEBUG2, "palloc(%d)", len); |
363 |
|
364 |
val = palloc(len); |
365 |
memset(val, 0, len); |
366 |
|
367 |
snprintf(val, len, "%lu", prop); |
368 |
|
369 |
elog(DEBUG2, "val=%s", val); |
370 |
|
371 |
return val; |
372 |
} |
373 |
*/ |