1 |
|
|
/* $OpenBSD: search.c,v 1.14 2016/08/14 21:47:16 guenther Exp $ */ |
2 |
|
|
|
3 |
|
|
/*- |
4 |
|
|
* Copyright (c) 1992, 1993, 1994 |
5 |
|
|
* The Regents of the University of California. All rights reserved. |
6 |
|
|
* Copyright (c) 1992, 1993, 1994, 1995, 1996 |
7 |
|
|
* Keith Bostic. All rights reserved. |
8 |
|
|
* |
9 |
|
|
* See the LICENSE file for redistribution information. |
10 |
|
|
*/ |
11 |
|
|
|
12 |
|
|
#include "config.h" |
13 |
|
|
|
14 |
|
|
#include <sys/types.h> |
15 |
|
|
#include <sys/queue.h> |
16 |
|
|
|
17 |
|
|
#include <bitstring.h> |
18 |
|
|
#include <ctype.h> |
19 |
|
|
#include <errno.h> |
20 |
|
|
#include <limits.h> |
21 |
|
|
#include <stdio.h> |
22 |
|
|
#include <stdlib.h> |
23 |
|
|
#include <string.h> |
24 |
|
|
#include <unistd.h> |
25 |
|
|
|
26 |
|
|
#include "common.h" |
27 |
|
|
|
28 |
|
|
typedef enum { S_EMPTY, S_EOF, S_NOPREV, S_NOTFOUND, S_SOF, S_WRAP } smsg_t; |
29 |
|
|
|
30 |
|
|
static void search_msg(SCR *, smsg_t); |
31 |
|
|
static int search_init(SCR *, dir_t, char *, size_t, char **, u_int); |
32 |
|
|
|
33 |
|
|
/* |
34 |
|
|
* search_init -- |
35 |
|
|
* Set up a search. |
36 |
|
|
*/ |
37 |
|
|
static int |
38 |
|
|
search_init(SCR *sp, dir_t dir, char *ptrn, size_t plen, char **epp, |
39 |
|
|
u_int flags) |
40 |
|
|
{ |
41 |
|
|
recno_t lno; |
42 |
|
|
int delim; |
43 |
|
|
char *p, *t; |
44 |
|
|
|
45 |
|
|
/* If the file is empty, it's a fast search. */ |
46 |
|
|
if (sp->lno <= 1) { |
47 |
|
|
if (db_last(sp, &lno)) |
48 |
|
|
return (1); |
49 |
|
|
if (lno == 0) { |
50 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
51 |
|
|
search_msg(sp, S_EMPTY); |
52 |
|
|
return (1); |
53 |
|
|
} |
54 |
|
|
} |
55 |
|
|
|
56 |
|
|
if (LF_ISSET(SEARCH_PARSE)) { /* Parse the string. */ |
57 |
|
|
/* |
58 |
|
|
* Use the saved pattern if no pattern specified, or if only |
59 |
|
|
* one or two delimiter characters specified. |
60 |
|
|
* |
61 |
|
|
* !!! |
62 |
|
|
* Historically, only the pattern itself was saved, vi didn't |
63 |
|
|
* preserve addressing or delta information. |
64 |
|
|
*/ |
65 |
|
|
if (ptrn == NULL) |
66 |
|
|
goto prev; |
67 |
|
|
if (plen == 1) { |
68 |
|
|
if (epp != NULL) |
69 |
|
|
*epp = ptrn + 1; |
70 |
|
|
goto prev; |
71 |
|
|
} |
72 |
|
|
if (ptrn[0] == ptrn[1]) { |
73 |
|
|
if (epp != NULL) |
74 |
|
|
*epp = ptrn + 2; |
75 |
|
|
|
76 |
|
|
/* Complain if we don't have a previous pattern. */ |
77 |
|
|
prev: if (sp->re == NULL) { |
78 |
|
|
search_msg(sp, S_NOPREV); |
79 |
|
|
return (1); |
80 |
|
|
} |
81 |
|
|
/* Re-compile the search pattern if necessary. */ |
82 |
|
|
if (!F_ISSET(sp, SC_RE_SEARCH) && re_compile(sp, |
83 |
|
|
sp->re, sp->re_len, NULL, NULL, &sp->re_c, |
84 |
|
|
RE_C_SEARCH | |
85 |
|
|
(LF_ISSET(SEARCH_MSG) ? 0 : RE_C_SILENT))) |
86 |
|
|
return (1); |
87 |
|
|
|
88 |
|
|
/* Set the search direction. */ |
89 |
|
|
if (LF_ISSET(SEARCH_SET)) |
90 |
|
|
sp->searchdir = dir; |
91 |
|
|
return (0); |
92 |
|
|
} |
93 |
|
|
|
94 |
|
|
/* |
95 |
|
|
* Set the delimiter, and move forward to the terminating |
96 |
|
|
* delimiter, handling escaped delimiters. |
97 |
|
|
* |
98 |
|
|
* QUOTING NOTE: |
99 |
|
|
* Only discard an escape character if it escapes a delimiter. |
100 |
|
|
*/ |
101 |
|
|
for (delim = *ptrn, p = t = ++ptrn;; *t++ = *p++) { |
102 |
|
|
if (--plen == 0 || p[0] == delim) { |
103 |
|
|
if (plen != 0) |
104 |
|
|
++p; |
105 |
|
|
break; |
106 |
|
|
} |
107 |
|
|
if (plen > 1 && p[0] == '\\' && p[1] == delim) { |
108 |
|
|
++p; |
109 |
|
|
--plen; |
110 |
|
|
} |
111 |
|
|
} |
112 |
|
|
if (epp != NULL) |
113 |
|
|
*epp = p; |
114 |
|
|
|
115 |
|
|
plen = t - ptrn; |
116 |
|
|
} |
117 |
|
|
|
118 |
|
|
/* Compile the RE. */ |
119 |
|
|
if (re_compile(sp, ptrn, plen, &sp->re, &sp->re_len, &sp->re_c, |
120 |
|
|
RE_C_SEARCH | |
121 |
|
|
(LF_ISSET(SEARCH_MSG) ? 0 : RE_C_SILENT) | |
122 |
|
|
(LF_ISSET(SEARCH_TAG) ? RE_C_TAG : 0))) |
123 |
|
|
return (1); |
124 |
|
|
|
125 |
|
|
/* Set the search direction. */ |
126 |
|
|
if (LF_ISSET(SEARCH_SET)) |
127 |
|
|
sp->searchdir = dir; |
128 |
|
|
|
129 |
|
|
return (0); |
130 |
|
|
} |
131 |
|
|
|
132 |
|
|
/* |
133 |
|
|
* f_search -- |
134 |
|
|
* Do a forward search. |
135 |
|
|
* |
136 |
|
|
* PUBLIC: int f_search(SCR *, MARK *, MARK *, char *, size_t, char **, u_int); |
137 |
|
|
*/ |
138 |
|
|
int |
139 |
|
|
f_search(SCR *sp, MARK *fm, MARK *rm, char *ptrn, size_t plen, char **eptrn, |
140 |
|
|
u_int flags) |
141 |
|
|
{ |
142 |
|
|
busy_t btype; |
143 |
|
|
recno_t lno; |
144 |
|
|
regmatch_t match[1]; |
145 |
|
|
size_t coff, len; |
146 |
|
|
int cnt, eval, rval, wrapped = 0; |
147 |
|
|
char *l; |
148 |
|
|
|
149 |
|
|
if (search_init(sp, FORWARD, ptrn, plen, eptrn, flags)) |
150 |
|
|
return (1); |
151 |
|
|
|
152 |
|
|
if (LF_ISSET(SEARCH_FILE)) { |
153 |
|
|
lno = 1; |
154 |
|
|
coff = 0; |
155 |
|
|
} else { |
156 |
|
|
if (db_get(sp, fm->lno, DBG_FATAL, &l, &len)) |
157 |
|
|
return (1); |
158 |
|
|
lno = fm->lno; |
159 |
|
|
|
160 |
|
|
/* |
161 |
|
|
* If doing incremental search, start searching at the previous |
162 |
|
|
* column, so that we search a minimal distance and still match |
163 |
|
|
* special patterns, e.g., \< for beginning of a word. |
164 |
|
|
* |
165 |
|
|
* Otherwise, start searching immediately after the cursor. If |
166 |
|
|
* at the end of the line, start searching on the next line. |
167 |
|
|
* This is incompatible (read bug fix) with the historic vi -- |
168 |
|
|
* searches for the '$' pattern never moved forward, and the |
169 |
|
|
* "-t foo" didn't work if the 'f' was the first character in |
170 |
|
|
* the file. |
171 |
|
|
*/ |
172 |
|
|
if (LF_ISSET(SEARCH_INCR)) { |
173 |
|
|
if ((coff = fm->cno) != 0) |
174 |
|
|
--coff; |
175 |
|
|
} else if (fm->cno + 1 >= len) { |
176 |
|
|
coff = 0; |
177 |
|
|
lno = fm->lno + 1; |
178 |
|
|
if (db_get(sp, lno, 0, &l, &len)) { |
179 |
|
|
if (!O_ISSET(sp, O_WRAPSCAN)) { |
180 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
181 |
|
|
search_msg(sp, S_EOF); |
182 |
|
|
return (1); |
183 |
|
|
} |
184 |
|
|
lno = 1; |
185 |
|
|
wrapped = 1; |
186 |
|
|
} |
187 |
|
|
} else |
188 |
|
|
coff = fm->cno + 1; |
189 |
|
|
} |
190 |
|
|
|
191 |
|
|
btype = BUSY_ON; |
192 |
|
|
for (cnt = INTERRUPT_CHECK, rval = 1;; ++lno, coff = 0) { |
193 |
|
|
if (cnt-- == 0) { |
194 |
|
|
if (INTERRUPTED(sp)) |
195 |
|
|
break; |
196 |
|
|
if (LF_ISSET(SEARCH_MSG)) { |
197 |
|
|
search_busy(sp, btype); |
198 |
|
|
btype = BUSY_UPDATE; |
199 |
|
|
} |
200 |
|
|
cnt = INTERRUPT_CHECK; |
201 |
|
|
} |
202 |
|
|
if ((wrapped && lno > fm->lno) || db_get(sp, lno, 0, &l, &len)) { |
203 |
|
|
if (wrapped) { |
204 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
205 |
|
|
search_msg(sp, S_NOTFOUND); |
206 |
|
|
break; |
207 |
|
|
} |
208 |
|
|
if (!O_ISSET(sp, O_WRAPSCAN)) { |
209 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
210 |
|
|
search_msg(sp, S_EOF); |
211 |
|
|
break; |
212 |
|
|
} |
213 |
|
|
lno = 0; |
214 |
|
|
wrapped = 1; |
215 |
|
|
continue; |
216 |
|
|
} |
217 |
|
|
|
218 |
|
|
/* If already at EOL, just keep going. */ |
219 |
|
|
if (len != 0 && coff == len) |
220 |
|
|
continue; |
221 |
|
|
|
222 |
|
|
/* Set the termination. */ |
223 |
|
|
match[0].rm_so = coff; |
224 |
|
|
match[0].rm_eo = len; |
225 |
|
|
|
226 |
|
|
/* Search the line. */ |
227 |
|
|
eval = regexec(&sp->re_c, l, 1, match, |
228 |
|
|
(match[0].rm_so == 0 ? 0 : REG_NOTBOL) | REG_STARTEND); |
229 |
|
|
if (eval == REG_NOMATCH) |
230 |
|
|
continue; |
231 |
|
|
if (eval != 0) { |
232 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
233 |
|
|
re_error(sp, eval, &sp->re_c); |
234 |
|
|
else |
235 |
|
|
(void)sp->gp->scr_bell(sp); |
236 |
|
|
break; |
237 |
|
|
} |
238 |
|
|
|
239 |
|
|
/* Warn if the search wrapped. */ |
240 |
|
|
if (wrapped && LF_ISSET(SEARCH_WMSG)) |
241 |
|
|
search_msg(sp, S_WRAP); |
242 |
|
|
|
243 |
|
|
rm->lno = lno; |
244 |
|
|
rm->cno = match[0].rm_so; |
245 |
|
|
|
246 |
|
|
/* |
247 |
|
|
* If a change command, it's possible to move beyond the end |
248 |
|
|
* of a line. Historic vi generally got this wrong (e.g. try |
249 |
|
|
* "c?$<cr>"). Not all that sure this gets it right, there |
250 |
|
|
* are lots of strange cases. |
251 |
|
|
*/ |
252 |
|
|
if (!LF_ISSET(SEARCH_EOL) && rm->cno >= len) |
253 |
|
|
rm->cno = len != 0 ? len - 1 : 0; |
254 |
|
|
|
255 |
|
|
rval = 0; |
256 |
|
|
break; |
257 |
|
|
} |
258 |
|
|
|
259 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
260 |
|
|
search_busy(sp, BUSY_OFF); |
261 |
|
|
return (rval); |
262 |
|
|
} |
263 |
|
|
|
264 |
|
|
/* |
265 |
|
|
* b_search -- |
266 |
|
|
* Do a backward search. |
267 |
|
|
* |
268 |
|
|
* PUBLIC: int b_search(SCR *, MARK *, MARK *, char *, size_t, char **, u_int); |
269 |
|
|
*/ |
270 |
|
|
int |
271 |
|
|
b_search(SCR *sp, MARK *fm, MARK *rm, char *ptrn, size_t plen, char **eptrn, |
272 |
|
|
u_int flags) |
273 |
|
|
{ |
274 |
|
|
busy_t btype; |
275 |
|
|
recno_t lno; |
276 |
|
|
regmatch_t match[1]; |
277 |
|
|
size_t coff, last, len; |
278 |
|
|
int cnt, eval, rval, wrapped; |
279 |
|
|
char *l; |
280 |
|
|
|
281 |
|
|
if (search_init(sp, BACKWARD, ptrn, plen, eptrn, flags)) |
282 |
|
|
return (1); |
283 |
|
|
|
284 |
|
|
/* |
285 |
|
|
* If doing incremental search, set the "starting" position past the |
286 |
|
|
* current column, so that we search a minimal distance and still |
287 |
|
|
* match special patterns, e.g., \> for the end of a word. This is |
288 |
|
|
* safe when the cursor is at the end of a line because we only use |
289 |
|
|
* it for comparison with the location of the match. |
290 |
|
|
* |
291 |
|
|
* Otherwise, start searching immediately before the cursor. If in |
292 |
|
|
* the first column, start search on the previous line. |
293 |
|
|
*/ |
294 |
|
|
if (LF_ISSET(SEARCH_INCR)) { |
295 |
|
|
lno = fm->lno; |
296 |
|
|
coff = fm->cno + 1; |
297 |
|
|
} else { |
298 |
|
|
if (fm->cno == 0) { |
299 |
|
|
if (fm->lno == 1 && !O_ISSET(sp, O_WRAPSCAN)) { |
300 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
301 |
|
|
search_msg(sp, S_SOF); |
302 |
|
|
return (1); |
303 |
|
|
} |
304 |
|
|
lno = fm->lno - 1; |
305 |
|
|
} else |
306 |
|
|
lno = fm->lno; |
307 |
|
|
coff = fm->cno; |
308 |
|
|
} |
309 |
|
|
|
310 |
|
|
btype = BUSY_ON; |
311 |
|
|
for (cnt = INTERRUPT_CHECK, rval = 1, wrapped = 0;; --lno, coff = 0) { |
312 |
|
|
if (cnt-- == 0) { |
313 |
|
|
if (INTERRUPTED(sp)) |
314 |
|
|
break; |
315 |
|
|
if (LF_ISSET(SEARCH_MSG)) { |
316 |
|
|
search_busy(sp, btype); |
317 |
|
|
btype = BUSY_UPDATE; |
318 |
|
|
} |
319 |
|
|
cnt = INTERRUPT_CHECK; |
320 |
|
|
} |
321 |
|
|
if ((wrapped && lno < fm->lno) || lno == 0) { |
322 |
|
|
if (wrapped) { |
323 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
324 |
|
|
search_msg(sp, S_NOTFOUND); |
325 |
|
|
break; |
326 |
|
|
} |
327 |
|
|
if (!O_ISSET(sp, O_WRAPSCAN)) { |
328 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
329 |
|
|
search_msg(sp, S_SOF); |
330 |
|
|
break; |
331 |
|
|
} |
332 |
|
|
if (db_last(sp, &lno)) |
333 |
|
|
break; |
334 |
|
|
if (lno == 0) { |
335 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
336 |
|
|
search_msg(sp, S_EMPTY); |
337 |
|
|
break; |
338 |
|
|
} |
339 |
|
|
++lno; |
340 |
|
|
wrapped = 1; |
341 |
|
|
continue; |
342 |
|
|
} |
343 |
|
|
|
344 |
|
|
if (db_get(sp, lno, 0, &l, &len)) |
345 |
|
|
break; |
346 |
|
|
|
347 |
|
|
/* Set the termination. */ |
348 |
|
|
match[0].rm_so = 0; |
349 |
|
|
match[0].rm_eo = len; |
350 |
|
|
|
351 |
|
|
/* Search the line. */ |
352 |
|
|
eval = regexec(&sp->re_c, l, 1, match, |
353 |
|
|
(match[0].rm_eo == len ? 0 : REG_NOTEOL) | REG_STARTEND); |
354 |
|
|
if (eval == REG_NOMATCH) |
355 |
|
|
continue; |
356 |
|
|
if (eval != 0) { |
357 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
358 |
|
|
re_error(sp, eval, &sp->re_c); |
359 |
|
|
else |
360 |
|
|
(void)sp->gp->scr_bell(sp); |
361 |
|
|
break; |
362 |
|
|
} |
363 |
|
|
|
364 |
|
|
/* Check for a match starting past the cursor. */ |
365 |
|
|
if (coff != 0 && match[0].rm_so >= coff) |
366 |
|
|
continue; |
367 |
|
|
|
368 |
|
|
/* Warn if the search wrapped. */ |
369 |
|
|
if (wrapped && LF_ISSET(SEARCH_WMSG)) |
370 |
|
|
search_msg(sp, S_WRAP); |
371 |
|
|
|
372 |
|
|
/* |
373 |
|
|
* We now have the first match on the line. Step through the |
374 |
|
|
* line character by character until find the last acceptable |
375 |
|
|
* match. This is painful, we need a better interface to regex |
376 |
|
|
* to make this work. |
377 |
|
|
*/ |
378 |
|
|
for (;;) { |
379 |
|
|
last = match[0].rm_so++; |
380 |
|
|
if (match[0].rm_so >= len) |
381 |
|
|
break; |
382 |
|
|
match[0].rm_eo = len; |
383 |
|
|
eval = regexec(&sp->re_c, l, 1, match, |
384 |
|
|
(match[0].rm_so == 0 ? 0 : REG_NOTBOL) | |
385 |
|
|
REG_STARTEND); |
386 |
|
|
if (eval == REG_NOMATCH) |
387 |
|
|
break; |
388 |
|
|
if (eval != 0) { |
389 |
|
|
if (LF_ISSET(SEARCH_MSG)) |
390 |
|
|
re_error(sp, eval, &sp->re_c); |
391 |
|
|
else |
392 |
|
|
(void)sp->gp->scr_bell(sp); |
393 |
|
|
goto err; |
394 |
|
|
} |
395 |
|
|
if (coff && match[0].rm_so >= coff) |
396 |
|
|
break; |
397 |
|
|
} |
398 |
|
|
rm->lno = lno; |
399 |
|
|
|
400 |
|
|
/* See comment in f_search(). */ |
401 |
|
|
if (!LF_ISSET(SEARCH_EOL) && last >= len) |
402 |
|
|
rm->cno = len != 0 ? len - 1 : 0; |
403 |
|
|
else |
404 |
|
|
rm->cno = last; |
405 |
|
|
rval = 0; |
406 |
|
|
break; |
407 |
|
|
} |
408 |
|
|
|
409 |
|
|
err: if (LF_ISSET(SEARCH_MSG)) |
410 |
|
|
search_busy(sp, BUSY_OFF); |
411 |
|
|
return (rval); |
412 |
|
|
} |
413 |
|
|
|
414 |
|
|
/* |
415 |
|
|
* search_msg -- |
416 |
|
|
* Display one of the search messages. |
417 |
|
|
*/ |
418 |
|
|
static void |
419 |
|
|
search_msg(SCR *sp, smsg_t msg) |
420 |
|
|
{ |
421 |
|
|
switch (msg) { |
422 |
|
|
case S_EMPTY: |
423 |
|
|
msgq(sp, M_ERR, "File empty; nothing to search"); |
424 |
|
|
break; |
425 |
|
|
case S_EOF: |
426 |
|
|
msgq(sp, M_ERR, |
427 |
|
|
"Reached end-of-file without finding the pattern"); |
428 |
|
|
break; |
429 |
|
|
case S_NOPREV: |
430 |
|
|
msgq(sp, M_ERR, "No previous search pattern"); |
431 |
|
|
break; |
432 |
|
|
case S_NOTFOUND: |
433 |
|
|
msgq(sp, M_ERR, "Pattern not found"); |
434 |
|
|
break; |
435 |
|
|
case S_SOF: |
436 |
|
|
msgq(sp, M_ERR, |
437 |
|
|
"Reached top-of-file without finding the pattern"); |
438 |
|
|
break; |
439 |
|
|
case S_WRAP: |
440 |
|
|
msgq(sp, M_ERR, "Search wrapped"); |
441 |
|
|
break; |
442 |
|
|
default: |
443 |
|
|
abort(); |
444 |
|
|
} |
445 |
|
|
} |
446 |
|
|
|
447 |
|
|
/* |
448 |
|
|
* search_busy -- |
449 |
|
|
* Put up the busy searching message. |
450 |
|
|
* |
451 |
|
|
* PUBLIC: void search_busy(SCR *, busy_t); |
452 |
|
|
*/ |
453 |
|
|
void |
454 |
|
|
search_busy(SCR *sp, busy_t btype) |
455 |
|
|
{ |
456 |
|
|
sp->gp->scr_busy(sp, "Searching...", btype); |
457 |
|
|
} |