xref: /netbsd/distrib/utils/more/linenum.c (revision 6550d01e)
1 /*	$NetBSD: linenum.c,v 1.5 2003/10/13 14:34:25 agc Exp $	*/
2 
3 /*
4  * Copyright (c) 1988 Mark Nudelman
5  * Copyright (c) 1988, 1993
6  *	The Regents of the University of California.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of the University nor the names of its contributors
17  *    may be used to endorse or promote products derived from this software
18  *    without specific prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30  * SUCH DAMAGE.
31  */
32 
33 #include <sys/cdefs.h>
34 #ifndef lint
35 #if 0
36 static char sccsid[] = "@(#)linenum.c	8.1 (Berkeley) 6/6/93";
37 #else
38 __RCSID("$NetBSD: linenum.c,v 1.5 2003/10/13 14:34:25 agc Exp $");
39 #endif
40 #endif /* not lint */
41 
42 /*
43  * Code to handle displaying line numbers.
44  *
45  * Finding the line number of a given file position is rather tricky.
46  * We don't want to just start at the beginning of the file and
47  * count newlines, because that is slow for large files (and also
48  * wouldn't work if we couldn't get to the start of the file; e.g.
49  * if input is a long pipe).
50  *
51  * So we use the function add_lnum to cache line numbers.
52  * We try to be very clever and keep only the more interesting
53  * line numbers when we run out of space in our table.  A line
54  * number is more interesting than another when it is far from
55  * other line numbers.   For example, we'd rather keep lines
56  * 100,200,300 than 100,101,300.  200 is more interesting than
57  * 101 because 101 can be derived very cheaply from 100, while
58  * 200 is more expensive to derive from 100.
59  *
60  * The function currline() returns the line number of a given
61  * position in the file.  As a side effect, it calls add_lnum
62  * to cache the line number.  Therefore currline is occasionally
63  * called to make sure we cache line numbers often enough.
64  */
65 
66 #include <sys/types.h>
67 #include <stdio.h>
68 #include <time.h>
69 
70 #include "less.h"
71 #include "extern.h"
72 
73 /*
74  * Structure to keep track of a line number and the associated file position.
75  * A doubly-linked circular list of line numbers is kept ordered by line number.
76  */
77 struct linenum
78 {
79 	struct linenum *next;		/* Link to next in the list */
80 	struct linenum *prev;		/* Line to previous in the list */
81 	off_t pos;			/* File position */
82 	off_t gap;			/* Gap between prev and next */
83 	int line;			/* Line number */
84 };
85 /*
86  * "gap" needs some explanation: the gap of any particular line number
87  * is the distance between the previous one and the next one in the list.
88  * ("Distance" means difference in file position.)  In other words, the
89  * gap of a line number is the gap which would be introduced if this
90  * line number were deleted.  It is used to decide which one to replace
91  * when we have a new one to insert and the table is full.
92  */
93 
94 #define	NPOOL	50			/* Size of line number pool */
95 
96 #define	LONGTIME	(2)		/* In seconds */
97 
98 int lnloop = 0;				/* Are we in the line num loop? */
99 
100 static struct linenum anchor;		/* Anchor of the list */
101 static struct linenum *freelist;	/* Anchor of the unused entries */
102 static struct linenum pool[NPOOL];	/* The pool itself */
103 static struct linenum *spare;		/* We always keep one spare entry */
104 
105 static void calcgap __P((struct linenum *));
106 static void longloopmessage __P((void));
107 /*
108  * Initialize the line number structures.
109  */
110 void
111 clr_linenum()
112 {
113 	struct linenum *p;
114 
115 	/*
116 	 * Put all the entries on the free list.
117 	 * Leave one for the "spare".
118 	 */
119 	for (p = pool;  p < &pool[NPOOL-2];  p++)
120 		p->next = p+1;
121 	pool[NPOOL-2].next = NULL;
122 	freelist = pool;
123 
124 	spare = &pool[NPOOL-1];
125 
126 	/*
127 	 * Initialize the anchor.
128 	 */
129 	anchor.next = anchor.prev = &anchor;
130 	anchor.gap = 0;
131 	anchor.pos = (off_t)0;
132 	anchor.line = 1;
133 }
134 
135 /*
136  * Calculate the gap for an entry.
137  */
138 static void
139 calcgap(p)
140 	struct linenum *p;
141 {
142 	/*
143 	 * Don't bother to compute a gap for the anchor.
144 	 * Also don't compute a gap for the last one in the list.
145 	 * The gap for that last one should be considered infinite,
146 	 * but we never look at it anyway.
147 	 */
148 	if (p == &anchor || p->next == &anchor)
149 		return;
150 	p->gap = p->next->pos - p->prev->pos;
151 }
152 
153 /*
154  * Add a new line number to the cache.
155  * The specified position (pos) should be the file position of the
156  * FIRST character in the specified line.
157  */
158 void
159 add_lnum(line, pos)
160 	int line;
161 	off_t pos;
162 {
163 	struct linenum *p;
164 	struct linenum *new;
165 	struct linenum *nextp;
166 	struct linenum *prevp;
167 	off_t mingap;
168 
169 	/*
170 	 * Find the proper place in the list for the new one.
171 	 * The entries are sorted by position.
172 	 */
173 	for (p = anchor.next;  p != &anchor && p->pos < pos;  p = p->next)
174 		if (p->line == line)
175 			/* We already have this one. */
176 			return;
177 	nextp = p;
178 	prevp = p->prev;
179 
180 	if (freelist != NULL)
181 	{
182 		/*
183 		 * We still have free (unused) entries.
184 		 * Use one of them.
185 		 */
186 		new = freelist;
187 		freelist = freelist->next;
188 	} else
189 	{
190 		/*
191 		 * No free entries.
192 		 * Use the "spare" entry.
193 		 */
194 		new = spare;
195 		spare = NULL;
196 	}
197 
198 	/*
199 	 * Fill in the fields of the new entry,
200 	 * and insert it into the proper place in the list.
201 	 */
202 	new->next = nextp;
203 	new->prev = prevp;
204 	new->pos = pos;
205 	new->line = line;
206 
207 	nextp->prev = new;
208 	prevp->next = new;
209 
210 	/*
211 	 * Recalculate gaps for the new entry and the neighboring entries.
212 	 */
213 	calcgap(new);
214 	calcgap(nextp);
215 	calcgap(prevp);
216 
217 	if (spare == NULL)
218 	{
219 		/*
220 		 * We have used the spare entry.
221 		 * Scan the list to find the one with the smallest
222 		 * gap, take it out and make it the spare.
223 		 * We should never remove the last one, so stop when
224 		 * we get to p->next == &anchor.  This also avoids
225 		 * looking at the gap of the last one, which is
226 		 * not computed by calcgap.
227 		 */
228 		mingap = anchor.next->gap;
229 		for (p = anchor.next;  p->next != &anchor;  p = p->next)
230 		{
231 			if (p->gap <= mingap)
232 			{
233 				spare = p;
234 				mingap = p->gap;
235 			}
236 		}
237 		spare->next->prev = spare->prev;
238 		spare->prev->next = spare->next;
239 	}
240 }
241 
242 /*
243  * If we get stuck in a long loop trying to figure out the
244  * line number, print a message to tell the user what we're doing.
245  */
246 static void
247 longloopmessage()
248 {
249 	ierror("Calculating line numbers");
250 	/*
251 	 * Set the lnloop flag here, so if the user interrupts while
252 	 * we are calculating line numbers, the signal handler will
253 	 * turn off line numbers (linenums=0).
254 	 */
255 	lnloop = 1;
256 }
257 
258 /*
259  * Find the line number associated with a given position.
260  * Return 0 if we can't figure it out.
261  */
262 int
263 find_linenum(pos)
264 	off_t pos;
265 {
266 	struct linenum *p;
267 	int lno;
268 	int loopcount;
269 	off_t cpos;
270 	time_t startime;
271 
272 	if (!linenums)
273 		/*
274 		 * We're not using line numbers.
275 		 */
276 		return (0);
277 	if (pos == NULL_POSITION)
278 		/*
279 		 * Caller doesn't know what he's talking about.
280 		 */
281 		return (0);
282 	if (pos == (off_t)0)
283 		/*
284 		 * Beginning of file is always line number 1.
285 		 */
286 		return (1);
287 
288 	/*
289 	 * Find the entry nearest to the position we want.
290 	 */
291 	for (p = anchor.next;  p != &anchor && p->pos < pos;  p = p->next)
292 		continue;
293 	if (p->pos == pos)
294 		/* Found it exactly. */
295 		return (p->line);
296 
297 	/*
298 	 * This is the (possibly) time-consuming part.
299 	 * We start at the line we just found and start
300 	 * reading the file forward or backward till we
301 	 * get to the place we want.
302 	 *
303 	 * First decide whether we should go forward from the
304 	 * previous one or backwards from the next one.
305 	 * The decision is based on which way involves
306 	 * traversing fewer bytes in the file.
307 	 */
308 	flush();
309 	(void)time(&startime);
310 	if (p == &anchor || pos - p->prev->pos < p->pos - pos)
311 	{
312 		/*
313 		 * Go forward.
314 		 */
315 		p = p->prev;
316 		if (ch_seek(p->pos))
317 			return (0);
318 		loopcount = 0;
319 		for (lno = p->line, cpos = p->pos;  cpos < pos;  lno++)
320 		{
321 			/*
322 			 * Allow a signal to abort this loop.
323 			 */
324 			cpos = forw_raw_line(cpos);
325 			if (sigs || cpos == NULL_POSITION)
326 				return (0);
327 			if (loopcount >= 0 && ++loopcount > 100) {
328 				loopcount = 0;
329 				if (time((time_t *)NULL)
330 				    >= startime + LONGTIME) {
331 					longloopmessage();
332 					loopcount = -1;
333 				}
334 			}
335 		}
336 		lnloop = 0;
337 		/*
338 		 * If the given position is not at the start of a line,
339 		 * make sure we return the correct line number.
340 		 */
341 		if (cpos > pos)
342 			lno--;
343 	} else
344 	{
345 		/*
346 		 * Go backward.
347 		 */
348 		if (ch_seek(p->pos))
349 			return (0);
350 		loopcount = 0;
351 		for (lno = p->line, cpos = p->pos;  cpos > pos;  lno--)
352 		{
353 			/*
354 			 * Allow a signal to abort this loop.
355 			 */
356 			cpos = back_raw_line(cpos);
357 			if (sigs || cpos == NULL_POSITION)
358 				return (0);
359 			if (loopcount >= 0 && ++loopcount > 100) {
360 				loopcount = 0;
361 				if (time((time_t *)NULL)
362 				    >= startime + LONGTIME) {
363 					longloopmessage();
364 					loopcount = -1;
365 				}
366 			}
367 		}
368 		lnloop = 0;
369 	}
370 
371 	/*
372 	 * We might as well cache it.
373 	 */
374 	add_lnum(lno, cpos);
375 	return (lno);
376 }
377 
378 /*
379  * Return the line number of the "current" line.
380  * The argument "where" tells which line is to be considered
381  * the "current" line (e.g. TOP, BOTTOM, MIDDLE, etc).
382  */
383 int
384 currline(where)
385 	int where;
386 {
387 	off_t pos;
388 
389 	if ((pos = position(where)) == NULL_POSITION)
390 		pos = ch_length();
391 	return(find_linenum(pos));
392 }
393