xref: /openbsd-src/usr.bin/mandoc/out.c (revision d1df930ffab53da22f3324c32bed7ac5709915e6)
1 /*	$OpenBSD: out.c,v 1.45 2018/08/19 23:10:16 schwarze Exp $ */
2 /*
3  * Copyright (c) 2009, 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4  * Copyright (c) 2011,2014,2015,2017,2018 Ingo Schwarze <schwarze@openbsd.org>
5  *
6  * Permission to use, copy, modify, and distribute this software for any
7  * purpose with or without fee is hereby granted, provided that the above
8  * copyright notice and this permission notice appear in all copies.
9  *
10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17  */
18 #include <sys/types.h>
19 
20 #include <assert.h>
21 #include <ctype.h>
22 #include <stdint.h>
23 #include <stdlib.h>
24 #include <string.h>
25 #include <time.h>
26 
27 #include "mandoc_aux.h"
28 #include "mandoc.h"
29 #include "out.h"
30 
31 static	void	tblcalc_data(struct rofftbl *, struct roffcol *,
32 			const struct tbl_opts *, const struct tbl_dat *,
33 			size_t);
34 static	void	tblcalc_literal(struct rofftbl *, struct roffcol *,
35 			const struct tbl_dat *, size_t);
36 static	void	tblcalc_number(struct rofftbl *, struct roffcol *,
37 			const struct tbl_opts *, const struct tbl_dat *);
38 
39 
40 /*
41  * Parse the *src string and store a scaling unit into *dst.
42  * If the string doesn't specify the unit, use the default.
43  * If no default is specified, fail.
44  * Return a pointer to the byte after the last byte used,
45  * or NULL on total failure.
46  */
47 const char *
48 a2roffsu(const char *src, struct roffsu *dst, enum roffscale def)
49 {
50 	char		*endptr;
51 
52 	dst->unit = def == SCALE_MAX ? SCALE_BU : def;
53 	dst->scale = strtod(src, &endptr);
54 	if (endptr == src)
55 		return NULL;
56 
57 	switch (*endptr++) {
58 	case 'c':
59 		dst->unit = SCALE_CM;
60 		break;
61 	case 'i':
62 		dst->unit = SCALE_IN;
63 		break;
64 	case 'f':
65 		dst->unit = SCALE_FS;
66 		break;
67 	case 'M':
68 		dst->unit = SCALE_MM;
69 		break;
70 	case 'm':
71 		dst->unit = SCALE_EM;
72 		break;
73 	case 'n':
74 		dst->unit = SCALE_EN;
75 		break;
76 	case 'P':
77 		dst->unit = SCALE_PC;
78 		break;
79 	case 'p':
80 		dst->unit = SCALE_PT;
81 		break;
82 	case 'u':
83 		dst->unit = SCALE_BU;
84 		break;
85 	case 'v':
86 		dst->unit = SCALE_VS;
87 		break;
88 	default:
89 		endptr--;
90 		if (SCALE_MAX == def)
91 			return NULL;
92 		dst->unit = def;
93 		break;
94 	}
95 	return endptr;
96 }
97 
98 /*
99  * Calculate the abstract widths and decimal positions of columns in a
100  * table.  This routine allocates the columns structures then runs over
101  * all rows and cells in the table.  The function pointers in "tbl" are
102  * used for the actual width calculations.
103  */
104 void
105 tblcalc(struct rofftbl *tbl, const struct tbl_span *sp,
106     size_t offset, size_t rmargin)
107 {
108 	struct roffsu		 su;
109 	const struct tbl_opts	*opts;
110 	const struct tbl_dat	*dp;
111 	struct roffcol		*col;
112 	size_t			 ewidth, xwidth;
113 	int			 spans;
114 	int			 icol, maxcol, necol, nxcol, quirkcol;
115 
116 	/*
117 	 * Allocate the master column specifiers.  These will hold the
118 	 * widths and decimal positions for all cells in the column.  It
119 	 * must be freed and nullified by the caller.
120 	 */
121 
122 	assert(NULL == tbl->cols);
123 	tbl->cols = mandoc_calloc((size_t)sp->opts->cols,
124 	    sizeof(struct roffcol));
125 	opts = sp->opts;
126 
127 	for (maxcol = -1; sp; sp = sp->next) {
128 		if (TBL_SPAN_DATA != sp->pos)
129 			continue;
130 		spans = 1;
131 		/*
132 		 * Account for the data cells in the layout, matching it
133 		 * to data cells in the data section.
134 		 */
135 		for (dp = sp->first; dp; dp = dp->next) {
136 			/* Do not used spanned cells in the calculation. */
137 			if (0 < --spans)
138 				continue;
139 			spans = dp->spans;
140 			if (1 < spans)
141 				continue;
142 			icol = dp->layout->col;
143 			while (maxcol < icol)
144 				tbl->cols[++maxcol].spacing = SIZE_MAX;
145 			col = tbl->cols + icol;
146 			col->flags |= dp->layout->flags;
147 			if (dp->layout->flags & TBL_CELL_WIGN)
148 				continue;
149 			if (dp->layout->wstr != NULL &&
150 			    dp->layout->width == 0 &&
151 			    a2roffsu(dp->layout->wstr, &su, SCALE_EN)
152 			    != NULL)
153 				dp->layout->width =
154 				    (*tbl->sulen)(&su, tbl->arg);
155 			if (col->width < dp->layout->width)
156 				col->width = dp->layout->width;
157 			if (dp->layout->spacing != SIZE_MAX &&
158 			    (col->spacing == SIZE_MAX ||
159 			     col->spacing < dp->layout->spacing))
160 				col->spacing = dp->layout->spacing;
161 			tblcalc_data(tbl, col, opts, dp,
162 			    dp->block == 0 ? 0 :
163 			    dp->layout->width ? dp->layout->width :
164 			    rmargin ? (rmargin + sp->opts->cols / 2)
165 			    / (sp->opts->cols + 1) : 0);
166 		}
167 	}
168 
169 	/*
170 	 * Align numbers with text.
171 	 * Count columns to equalize and columns to maximize.
172 	 * Find maximum width of the columns to equalize.
173 	 * Find total width of the columns *not* to maximize.
174 	 */
175 
176 	necol = nxcol = 0;
177 	ewidth = xwidth = 0;
178 	for (icol = 0; icol <= maxcol; icol++) {
179 		col = tbl->cols + icol;
180 		if (col->width > col->nwidth)
181 			col->decimal += (col->width - col->nwidth) / 2;
182 		else
183 			col->width = col->nwidth;
184 		if (col->spacing == SIZE_MAX || icol == maxcol)
185 			col->spacing = 3;
186 		if (col->flags & TBL_CELL_EQUAL) {
187 			necol++;
188 			if (ewidth < col->width)
189 				ewidth = col->width;
190 		}
191 		if (col->flags & TBL_CELL_WMAX)
192 			nxcol++;
193 		else
194 			xwidth += col->width;
195 	}
196 
197 	/*
198 	 * Equalize columns, if requested for any of them.
199 	 * Update total width of the columns not to maximize.
200 	 */
201 
202 	if (necol) {
203 		for (icol = 0; icol <= maxcol; icol++) {
204 			col = tbl->cols + icol;
205 			if ( ! (col->flags & TBL_CELL_EQUAL))
206 				continue;
207 			if (col->width == ewidth)
208 				continue;
209 			if (nxcol && rmargin)
210 				xwidth += ewidth - col->width;
211 			col->width = ewidth;
212 		}
213 	}
214 
215 	/*
216 	 * If there are any columns to maximize, find the total
217 	 * available width, deducting 3n margins between columns.
218 	 * Distribute the available width evenly.
219 	 */
220 
221 	if (nxcol && rmargin) {
222 		xwidth += 3*maxcol +
223 		    (opts->opts & (TBL_OPT_BOX | TBL_OPT_DBOX) ?
224 		     2 : !!opts->lvert + !!opts->rvert);
225 		if (rmargin <= offset + xwidth)
226 			return;
227 		xwidth = rmargin - offset - xwidth;
228 
229 		/*
230 		 * Emulate a bug in GNU tbl width calculation that
231 		 * manifests itself for large numbers of x-columns.
232 		 * Emulating it for 5 x-columns gives identical
233 		 * behaviour for up to 6 x-columns.
234 		 */
235 
236 		if (nxcol == 5) {
237 			quirkcol = xwidth % nxcol + 2;
238 			if (quirkcol != 3 && quirkcol != 4)
239 				quirkcol = -1;
240 		} else
241 			quirkcol = -1;
242 
243 		necol = 0;
244 		ewidth = 0;
245 		for (icol = 0; icol <= maxcol; icol++) {
246 			col = tbl->cols + icol;
247 			if ( ! (col->flags & TBL_CELL_WMAX))
248 				continue;
249 			col->width = (double)xwidth * ++necol / nxcol
250 			    - ewidth + 0.4995;
251 			if (necol == quirkcol)
252 				col->width--;
253 			ewidth += col->width;
254 		}
255 	}
256 }
257 
258 static void
259 tblcalc_data(struct rofftbl *tbl, struct roffcol *col,
260     const struct tbl_opts *opts, const struct tbl_dat *dp, size_t mw)
261 {
262 	size_t		 sz;
263 
264 	/* Branch down into data sub-types. */
265 
266 	switch (dp->layout->pos) {
267 	case TBL_CELL_HORIZ:
268 	case TBL_CELL_DHORIZ:
269 		sz = (*tbl->len)(1, tbl->arg);
270 		if (col->width < sz)
271 			col->width = sz;
272 		break;
273 	case TBL_CELL_LONG:
274 	case TBL_CELL_CENTRE:
275 	case TBL_CELL_LEFT:
276 	case TBL_CELL_RIGHT:
277 		tblcalc_literal(tbl, col, dp, mw);
278 		break;
279 	case TBL_CELL_NUMBER:
280 		tblcalc_number(tbl, col, opts, dp);
281 		break;
282 	case TBL_CELL_DOWN:
283 		break;
284 	default:
285 		abort();
286 	}
287 }
288 
289 static void
290 tblcalc_literal(struct rofftbl *tbl, struct roffcol *col,
291     const struct tbl_dat *dp, size_t mw)
292 {
293 	const char	*str;	/* Beginning of the first line. */
294 	const char	*beg;	/* Beginning of the current line. */
295 	char		*end;	/* End of the current line. */
296 	size_t		 lsz;	/* Length of the current line. */
297 	size_t		 wsz;	/* Length of the current word. */
298 
299 	if (dp->string == NULL || *dp->string == '\0')
300 		return;
301 	str = mw ? mandoc_strdup(dp->string) : dp->string;
302 	lsz = 0;
303 	for (beg = str; beg != NULL && *beg != '\0'; beg = end) {
304 		end = mw ? strchr(beg, ' ') : NULL;
305 		if (end != NULL) {
306 			*end++ = '\0';
307 			while (*end == ' ')
308 				end++;
309 		}
310 		wsz = (*tbl->slen)(beg, tbl->arg);
311 		if (mw && lsz && lsz + 1 + wsz <= mw)
312 			lsz += 1 + wsz;
313 		else
314 			lsz = wsz;
315 		if (col->width < lsz)
316 			col->width = lsz;
317 	}
318 	if (mw)
319 		free((void *)str);
320 }
321 
322 static void
323 tblcalc_number(struct rofftbl *tbl, struct roffcol *col,
324 		const struct tbl_opts *opts, const struct tbl_dat *dp)
325 {
326 	const char	*cp, *lastdigit, *lastpoint;
327 	size_t		 intsz, totsz;
328 	char		 buf[2];
329 
330 	if (dp->string == NULL || *dp->string == '\0')
331 		return;
332 
333 	/*
334 	 * Find the last digit and
335 	 * the last decimal point that is adjacent to a digit.
336 	 * The alignment indicator "\&" overrides everything.
337 	 */
338 
339 	lastdigit = lastpoint = NULL;
340 	for (cp = dp->string; cp[0] != '\0'; cp++) {
341 		if (cp[0] == '\\' && cp[1] == '&') {
342 			lastdigit = lastpoint = cp;
343 			break;
344 		} else if (cp[0] == opts->decimal &&
345 		    (isdigit((unsigned char)cp[1]) ||
346 		     (cp > dp->string && isdigit((unsigned char)cp[-1]))))
347 			lastpoint = cp;
348 		else if (isdigit((unsigned char)cp[0]))
349 			lastdigit = cp;
350 	}
351 
352 	/* Not a number, treat as a literal string. */
353 
354 	totsz = (*tbl->slen)(dp->string, tbl->arg);
355 	if (lastdigit == NULL) {
356 		if (col->width < totsz)
357 			col->width = totsz;
358 		return;
359 	}
360 
361 	/* Measure the width of the integer part. */
362 
363 	if (lastpoint == NULL)
364 		lastpoint = lastdigit + 1;
365 	intsz = 0;
366 	buf[1] = '\0';
367 	for (cp = dp->string; cp < lastpoint; cp++) {
368 		buf[0] = cp[0];
369 		intsz += (*tbl->slen)(buf, tbl->arg);
370 	}
371 
372 	/*
373          * If this number has more integer digits than all numbers
374          * seen on earlier lines, shift them all to the right.
375 	 * If it has fewer, shift this number to the right.
376 	 */
377 
378 	if (intsz > col->decimal) {
379 		col->nwidth += intsz - col->decimal;
380 		col->decimal = intsz;
381 	} else
382 		totsz += col->decimal - intsz;
383 
384 	/* Update the maximum total width seen so far. */
385 
386 	if (totsz > col->nwidth)
387 		col->nwidth = totsz;
388 }
389