xref: /openbsd/usr.bin/vi/ex/ex_txt.c (revision ebcaf48c)
1 /*	$OpenBSD: ex_txt.c,v 1.17 2020/04/30 10:40:21 millert 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 <limits.h>
20 #include <stdio.h>
21 #include <stdlib.h>
22 #include <string.h>
23 
24 #include "../common/common.h"
25 #include "../vi/vi.h"
26 
27 /*
28  * !!!
29  * The backslash characters was special when it preceded a newline as part of
30  * a substitution replacement pattern.  For example, the input ":a\<cr>" would
31  * failed immediately with an error, as the <cr> wasn't part of a substitution
32  * replacement pattern.  This implies a frightening integration of the editor
33  * and the parser and/or the RE engine.  There's no way I'm going to reproduce
34  * those semantics.
35  *
36  * So, if backslashes are special, this code inserts the backslash and the next
37  * character into the string, without regard for the character or the command
38  * being entered.  Since "\<cr>" was illegal historically (except for the one
39  * special case), and the command will fail eventually, no historical scripts
40  * should break (presuming they didn't depend on the failure mode itself or the
41  * characters remaining when failure occurred.
42  */
43 
44 static int	txt_dent(SCR *, TEXT *);
45 static void	txt_prompt(SCR *, TEXT *, CHAR_T, u_int32_t);
46 
47 /*
48  * ex_txt --
49  *	Get lines from the terminal for ex.
50  *
51  * PUBLIC: int ex_txt(SCR *, TEXTH *, CHAR_T, u_int32_t);
52  */
53 int
ex_txt(SCR * sp,TEXTH * tiqh,CHAR_T prompt,u_int32_t flags)54 ex_txt(SCR *sp, TEXTH *tiqh, CHAR_T prompt, u_int32_t flags)
55 {
56 	EVENT ev;
57 	GS *gp;
58 	TEXT ait, *ntp, *tp;
59 	carat_t carat_st;
60 	size_t cnt;
61 	int rval;
62 
63 	rval = 0;
64 
65 	/*
66 	 * Get a TEXT structure with some initial buffer space, reusing the
67 	 * last one if it's big enough.  (All TEXT bookkeeping fields default
68 	 * to 0 -- text_init() handles this.)
69 	 */
70 	if (!TAILQ_EMPTY(tiqh)) {
71 		tp = TAILQ_FIRST(tiqh);
72 		if (TAILQ_NEXT(tp, q) || tp->lb_len < 32) {
73 			text_lfree(tiqh);
74 			goto newtp;
75 		}
76 		tp->len = 0;
77 	} else {
78 newtp:		if ((tp = text_init(sp, NULL, 0, 32)) == NULL)
79 			goto err;
80 		TAILQ_INSERT_HEAD(tiqh, tp, q);
81 	}
82 
83 	/* Set the starting line number. */
84 	tp->lno = sp->lno + 1;
85 
86 	/*
87 	 * If it's a terminal, set up autoindent, put out the prompt, and
88 	 * set it up so we know we were suspended.  Otherwise, turn off
89 	 * the autoindent flag, as that requires less special casing below.
90 	 *
91 	 * XXX
92 	 * Historic practice is that ^Z suspended command mode (but, because
93 	 * it ran in cooked mode, it was unaffected by the autowrite option.)
94 	 * On restart, any "current" input was discarded, whether in insert
95 	 * mode or not, and ex was in command mode.  This code matches historic
96 	 * practice, but not 'cause it's easier.
97 	 */
98 	gp = sp->gp;
99 	if (F_ISSET(gp, G_SCRIPTED))
100 		LF_CLR(TXT_AUTOINDENT);
101 	else {
102 		if (LF_ISSET(TXT_AUTOINDENT)) {
103 			LF_SET(TXT_EOFCHAR);
104 			if (v_txt_auto(sp, sp->lno, NULL, 0, tp))
105 				goto err;
106 		}
107 		txt_prompt(sp, tp, prompt, flags);
108 	}
109 
110 	for (carat_st = C_NOTSET;;) {
111 		if (v_event_get(sp, &ev, 0, 0))
112 			goto err;
113 
114 		/* Deal with all non-character events. */
115 		switch (ev.e_event) {
116 		case E_CHARACTER:
117 			break;
118 		case E_ERR:
119 			goto err;
120 		case E_REPAINT:
121 		case E_WRESIZE:
122 			continue;
123 		case E_EOF:
124 			rval = 1;
125 			/* FALLTHROUGH */
126 		case E_INTERRUPT:
127 			/*
128 			 * Handle EOF/SIGINT events by discarding partially
129 			 * entered text and returning.  EOF returns failure,
130 			 * E_INTERRUPT returns success.
131 			 */
132 			goto notlast;
133 		default:
134 			v_event_err(sp, &ev);
135 			goto notlast;
136 		}
137 
138 		/*
139 		 * Deal with character events.
140 		 *
141 		 * Check to see if the character fits into the input buffer.
142 		 * (Use tp->len, ignore overwrite and non-printable chars.)
143 		 */
144 		BINC_GOTO(sp, tp->lb, tp->lb_len, tp->len + 1);
145 
146 		switch (ev.e_value) {
147 		case K_CR:
148 			/*
149 			 * !!!
150 			 * Historically, <carriage-return>'s in the command
151 			 * weren't special, so the ex parser would return an
152 			 * unknown command error message.  However, if they
153 			 * terminated the command if they were in a map.  I'm
154 			 * pretty sure this still isn't right, but it handles
155 			 * what I've seen so far.
156 			 */
157 			if (!F_ISSET(&ev.e_ch, CH_MAPPED))
158 				goto ins_ch;
159 			/* FALLTHROUGH */
160 		case K_NL:
161 			/*
162 			 * '\' can escape <carriage-return>/<newline>.  We
163 			 * don't discard the backslash because we need it
164 			 * to get the <newline> through the ex parser.
165 			 */
166 			if (LF_ISSET(TXT_BACKSLASH) &&
167 			    tp->len != 0 && tp->lb[tp->len - 1] == '\\')
168 				goto ins_ch;
169 
170 			/*
171 			 * CR returns from the ex command line.
172 			 *
173 			 * XXX
174 			 * Terminate with a nul, needed by filter.
175 			 */
176 			if (LF_ISSET(TXT_CR)) {
177 				tp->lb[tp->len] = '\0';
178 				goto done;
179 			}
180 
181 			/*
182 			 * '.' may terminate text input mode; free the current
183 			 * TEXT.
184 			 */
185 			if (LF_ISSET(TXT_DOTTERM) && tp->len == tp->ai + 1 &&
186 			    tp->lb[tp->len - 1] == '.') {
187 notlast:			TAILQ_REMOVE(tiqh, tp, q);
188 				text_free(tp);
189 				goto done;
190 			}
191 
192 			/* Set up bookkeeping for the new line. */
193 			if ((ntp = text_init(sp, NULL, 0, 32)) == NULL)
194 				goto err;
195 			ntp->lno = tp->lno + 1;
196 
197 			/*
198 			 * Reset the autoindent line value.  0^D keeps the ai
199 			 * line from changing, ^D changes the level, even if
200 			 * there were no characters in the old line.  Note, if
201 			 * using the current tp structure, use the cursor as
202 			 * the length, the autoindent characters may have been
203 			 * erased.
204 			 */
205 			if (LF_ISSET(TXT_AUTOINDENT)) {
206 				if (carat_st == C_NOCHANGE) {
207 					if (v_txt_auto(sp,
208 					    OOBLNO, &ait, ait.ai, ntp))
209 						goto err;
210 					free(ait.lb);
211 				} else
212 					if (v_txt_auto(sp,
213 					    OOBLNO, tp, tp->len, ntp))
214 						goto err;
215 				carat_st = C_NOTSET;
216 			}
217 			txt_prompt(sp, ntp, prompt, flags);
218 
219 			/*
220 			 * Swap old and new TEXT's, and insert the new TEXT
221 			 * into the queue.
222 			 */
223 			tp = ntp;
224 			TAILQ_INSERT_TAIL(tiqh, tp, q);
225 			break;
226 		case K_CARAT:			/* Delete autoindent chars. */
227 			if (tp->len <= tp->ai && LF_ISSET(TXT_AUTOINDENT))
228 				carat_st = C_CARATSET;
229 			goto ins_ch;
230 		case K_ZERO:			/* Delete autoindent chars. */
231 			if (tp->len <= tp->ai && LF_ISSET(TXT_AUTOINDENT))
232 				carat_st = C_ZEROSET;
233 			goto ins_ch;
234 		case K_CNTRLD:			/* Delete autoindent char. */
235 			/*
236 			 * !!!
237 			 * Historically, the ^D command took (but then ignored)
238 			 * a count.  For simplicity, we don't return it unless
239 			 * it's the first character entered.  The check for len
240 			 * equal to 0 is okay, TXT_AUTOINDENT won't be set.
241 			 */
242 			if (LF_ISSET(TXT_CNTRLD)) {
243 				for (cnt = 0; cnt < tp->len; ++cnt)
244 					if (!isblank(tp->lb[cnt]))
245 						break;
246 				if (cnt == tp->len) {
247 					tp->len = 1;
248 					tp->lb[0] = ev.e_c;
249 					tp->lb[1] = '\0';
250 
251 					/*
252 					 * Put out a line separator, in case
253 					 * the command fails.
254 					 */
255 					(void)putchar('\n');
256 					goto done;
257 				}
258 			}
259 
260 			/*
261 			 * POSIX 1003.1b-1993, paragraph 7.1.1.9, states that
262 			 * the EOF characters are discarded if there are other
263 			 * characters to process in the line, i.e. if the EOF
264 			 * is not the first character in the line.  For this
265 			 * reason, historic ex discarded the EOF characters,
266 			 * even if occurring in the middle of the input line.
267 			 * We match that historic practice.
268 			 *
269 			 * !!!
270 			 * The test for discarding in the middle of the line is
271 			 * done in the switch, because the CARAT forms are N+1,
272 			 * not N.
273 			 *
274 			 * !!!
275 			 * There's considerable magic to make the terminal code
276 			 * return the EOF character at all.  See that code for
277 			 * details.
278 			 */
279 			if (!LF_ISSET(TXT_AUTOINDENT) || tp->len == 0)
280 				continue;
281 			switch (carat_st) {
282 			case C_CARATSET:		/* ^^D */
283 				if (tp->len > tp->ai + 1)
284 					continue;
285 
286 				/* Save the ai string for later. */
287 				ait.lb = NULL;
288 				ait.lb_len = 0;
289 				BINC_GOTO(sp, ait.lb, ait.lb_len, tp->ai);
290 				memcpy(ait.lb, tp->lb, tp->ai);
291 				ait.ai = ait.len = tp->ai;
292 
293 				carat_st = C_NOCHANGE;
294 				goto leftmargin;
295 			case C_ZEROSET:			/* 0^D */
296 				if (tp->len > tp->ai + 1)
297 					continue;
298 
299 				carat_st = C_NOTSET;
300 leftmargin:			(void)gp->scr_ex_adjust(sp, EX_TERM_CE);
301 				tp->ai = tp->len = 0;
302 				break;
303 			case C_NOTSET:			/* ^D */
304 				if (tp->len > tp->ai)
305 					continue;
306 
307 				if (txt_dent(sp, tp))
308 					goto err;
309 				break;
310 			default:
311 				abort();
312 			}
313 
314 			/* Clear and redisplay the line. */
315 			(void)gp->scr_ex_adjust(sp, EX_TERM_CE);
316 			txt_prompt(sp, tp, prompt, flags);
317 			break;
318 		default:
319 			/*
320 			 * See the TXT_BEAUTIFY comment in vi/v_txt_ev.c.
321 			 *
322 			 * Silently eliminate any iscntrl() character that was
323 			 * not already handled specially, except for <tab> and
324 			 * <ff>.
325 			 */
326 ins_ch:			if (LF_ISSET(TXT_BEAUTIFY) && iscntrl(ev.e_c) &&
327 			    ev.e_value != K_FORMFEED && ev.e_value != K_TAB)
328 				break;
329 
330 			tp->lb[tp->len++] = ev.e_c;
331 			break;
332 		}
333 	}
334 	/* NOTREACHED */
335 
336 done:	return (rval);
337 
338 err:
339 alloc_err:
340 	return (1);
341 }
342 
343 /*
344  * txt_prompt --
345  *	Display the ex prompt, line number, ai characters.  Characters had
346  *	better be printable by the terminal driver, but that's its problem,
347  *	not ours.
348  */
349 static void
txt_prompt(SCR * sp,TEXT * tp,CHAR_T prompt,u_int32_t flags)350 txt_prompt(SCR *sp, TEXT *tp, CHAR_T prompt, u_int32_t flags)
351 {
352 	/* Display the prompt. */
353 	if (LF_ISSET(TXT_PROMPT))
354 		(void)printf("%c", prompt);
355 
356 	/* Display the line number. */
357 	if (LF_ISSET(TXT_NUMBER) && O_ISSET(sp, O_NUMBER))
358 		(void)printf("%6lu  ", (u_long)tp->lno);
359 
360 	/* Print out autoindent string. */
361 	if (LF_ISSET(TXT_AUTOINDENT))
362 		(void)printf("%.*s", (int)tp->ai, tp->lb);
363 	(void)fflush(stdout);
364 }
365 
366 /*
367  * txt_dent --
368  *	Handle ^D outdents.
369  *
370  * Ex version of vi/v_ntext.c:txt_dent().  See that code for the (usual)
371  * ranting and raving.  This is a fair bit simpler as ^T isn't special.
372  */
373 static int
txt_dent(SCR * sp,TEXT * tp)374 txt_dent(SCR *sp, TEXT *tp)
375 {
376 	u_long sw, ts;
377 	size_t cno, off, scno, spaces, tabs;
378 
379 	ts = O_VAL(sp, O_TABSTOP);
380 	sw = O_VAL(sp, O_SHIFTWIDTH);
381 
382 	/* Get the current screen column. */
383 	for (off = scno = 0; off < tp->len; ++off)
384 		if (tp->lb[off] == '\t')
385 			scno += COL_OFF(scno, ts);
386 		else
387 			++scno;
388 
389 	/* Get the previous shiftwidth column. */
390 	cno = scno--;
391 	scno -= scno % sw;
392 
393 	/*
394 	 * Since we don't know what comes before the character(s) being
395 	 * deleted, we have to resolve the autoindent characters .  The
396 	 * example is a <tab>, which doesn't take up a full shiftwidth
397 	 * number of columns because it's preceded by <space>s.  This is
398 	 * easy to get if the user sets shiftwidth to a value less than
399 	 * tabstop, and then uses ^T to indent, and ^D to outdent.
400 	 *
401 	 * Count up spaces/tabs needed to get to the target.
402 	 */
403 	cno = 0;
404 	tabs = 0;
405 	if (!O_ISSET(sp, O_EXPANDTAB)) {
406 		for (; cno + COL_OFF(cno, ts) <= scno; ++tabs)
407 			cno += COL_OFF(cno, ts);
408 	}
409 	spaces = scno - cno;
410 
411 	/* Make sure there's enough room. */
412 	BINC_RET(sp, tp->lb, tp->lb_len, tabs + spaces + 1);
413 
414 	/* Adjust the final ai character count. */
415 	tp->ai = tabs + spaces;
416 
417 	/* Enter the replacement characters. */
418 	for (tp->len = 0; tabs > 0; --tabs)
419 		tp->lb[tp->len++] = '\t';
420 	for (; spaces > 0; --spaces)
421 		tp->lb[tp->len++] = ' ';
422 	return (0);
423 }
424