2 * line formatting buffer for line adjustment and hyphenation
4 * The line formatting buffer does two main functions: breaking
5 * words into lines (possibly after breaking them at their
6 * hyphenation points), and, if requested, adjusting the space
7 * between words in a line. In this file the first step is
8 * referred to as filling.
10 * Functions like fmt_word() return nonzero on failure, which
11 * means the call should be repeated after fetching previously
12 * formatted lines via fmt_nextline().
19 #define FMT_LLEN(f) MAX(0, (f)->ll - (f)->li)
20 #define FMT_FILL(f) (!n_ce && n_u)
21 #define FMT_ADJ(f) (n_u && !n_na && !n_ce && (n_j & AD_B) == AD_B)
23 static int fmt_fillwords(struct fmt
*f
, int br
);
27 int wid
; /* word's width */
28 int elsn
, elsp
; /* els_neg and els_pos */
29 int gap
; /* the space before this word */
30 int hy
; /* hyphen width if inserted after this word */
31 int str
; /* does the space before it stretch */
32 int cost
; /* the extra cost of line break after this word */
44 int words_n
, words_sz
;
47 int lines_head
, lines_tail
, lines_sz
;
48 /* for paragraph adjustment */
53 int gap
; /* space before the next word */
54 int nls
; /* newlines before the next word */
55 int nls_sup
; /* suppressed newlines */
56 int li
, ll
; /* current line indentation and length */
57 int filled
; /* filled all words in the last fmt_fill() */
58 int eos
; /* last word ends a sentence */
59 int fillreq
; /* fill after the last word (\p) */
62 /* .ll, .in and .ti are delayed until the partial line is output */
63 static void fmt_confupdate(struct fmt
*f
)
66 f
->li
= n_ti
>= 0 ? n_ti
: n_i
;
70 static int fmt_confchanged(struct fmt
*f
)
72 return f
->ll
!= n_l
|| f
->li
!= (n_ti
>= 0 ? n_ti
: n_i
);
75 /* move words inside an fmt struct */
76 static void fmt_movewords(struct fmt
*a
, int dst
, int src
, int len
)
78 memmove(a
->words
+ dst
, a
->words
+ src
, len
* sizeof(a
->words
[0]));
81 /* move words from the buffer to s */
82 static int fmt_wordscopy(struct fmt
*f
, int beg
, int end
,
83 struct sbuf
*s
, int *els_neg
, int *els_pos
)
90 for (i
= beg
; i
< end
; i
++) {
92 sbuf_printf(s
, "%ch'%du'", c_ec
, wcur
->gap
);
93 sbuf_append(s
, wcur
->s
);
94 w
+= wcur
->wid
+ wcur
->gap
;
95 if (wcur
->elsn
< *els_neg
)
96 *els_neg
= wcur
->elsn
;
97 if (wcur
->elsp
> *els_pos
)
98 *els_pos
= wcur
->elsp
;
102 wcur
= &f
->words
[end
- 1];
104 sbuf_append(s
, "\\(hy");
110 static int fmt_nlines(struct fmt
*f
)
112 return f
->lines_head
- f
->lines_tail
;
115 /* the total width of the specified words in f->words[] */
116 static int fmt_wordslen(struct fmt
*f
, int beg
, int end
)
119 for (i
= beg
; i
< end
; i
++)
120 w
+= f
->words
[i
].wid
+ f
->words
[i
].gap
;
121 return beg
< end
? w
+ f
->words
[end
- 1].hy
: 0;
124 /* the number of stretchable spaces in f */
125 static int fmt_spaces(struct fmt
*f
, int beg
, int end
)
128 for (i
= beg
+ 1; i
< end
; i
++)
134 /* the amount of stretchable spaces in f */
135 static int fmt_spacessum(struct fmt
*f
, int beg
, int end
)
138 for (i
= beg
+ 1; i
< end
; i
++)
140 n
+= f
->words
[i
].gap
;
144 /* return the next line in the buffer */
145 char *fmt_nextline(struct fmt
*f
, int *w
,
146 int *li
, int *ll
, int *els_neg
, int *els_pos
)
149 if (f
->lines_head
== f
->lines_tail
)
151 l
= &f
->lines
[f
->lines_tail
++];
157 return sbuf_out(&l
->sbuf
);
160 static struct line
*fmt_mkline(struct fmt
*f
)
163 if (f
->lines_head
== f
->lines_tail
) {
167 if (f
->lines_head
== f
->lines_sz
) {
169 f
->lines
= mextend(f
->lines
, f
->lines_head
,
170 f
->lines_sz
, sizeof(f
->lines
[0]));
172 l
= &f
->lines
[f
->lines_head
++];
179 /* extract words from beg to end; shrink or stretch spaces if needed */
180 static int fmt_extractline(struct fmt
*f
, int beg
, int end
, int str
)
182 int fmt_div
, fmt_rem
;
183 int w
, i
, nspc
, llen
;
185 if (!(l
= fmt_mkline(f
)))
188 w
= fmt_wordslen(f
, beg
, end
);
189 nspc
= fmt_spaces(f
, beg
, end
);
190 if (nspc
&& FMT_ADJ(f
) && (llen
< w
|| str
)) {
191 fmt_div
= (llen
- w
) / nspc
;
192 fmt_rem
= (llen
- w
) % nspc
;
197 for (i
= beg
+ 1; i
< end
; i
++)
199 f
->words
[i
].gap
+= fmt_div
+ (fmt_rem
-- > 0);
201 l
->wid
= fmt_wordscopy(f
, beg
, end
, &l
->sbuf
, &l
->elsn
, &l
->elsp
);
205 static int fmt_sp(struct fmt
*f
)
207 if (fmt_fillwords(f
, 1))
209 if (fmt_extractline(f
, 0, f
->words_n
, 0))
219 /* fill as many lines as possible; if br, put the remaining words in a line */
220 int fmt_fill(struct fmt
*f
, int br
)
222 if (fmt_fillwords(f
, br
))
233 void fmt_space(struct fmt
*fmt
)
235 fmt
->gap
+= font_swid(dev_font(n_f
), n_s
, n_ss
);
238 int fmt_newline(struct fmt
*f
)
249 if (f
->nls
== 0 && !f
->filled
&& !f
->words_n
)
255 /* format the paragraph after the next word (\p) */
256 int fmt_fillreq(struct fmt
*f
)
259 if (fmt_fillwords(f
, 0))
261 f
->fillreq
= f
->words_n
+ 1;
265 static void fmt_wb2word(struct fmt
*f
, struct word
*word
, struct wb
*wb
,
266 int hy
, int str
, int gap
, int cost
)
268 int len
= strlen(wb_buf(wb
));
269 word
->s
= xmalloc(len
+ 1);
270 memcpy(word
->s
, wb_buf(wb
), len
+ 1);
271 word
->wid
= wb_wid(wb
);
272 word
->elsn
= wb
->els_neg
;
273 word
->elsp
= wb
->els_pos
;
274 word
->hy
= hy
? wb_hywid(wb
) : 0;
280 /* find explicit break positions: dashes, \:, \%, and \~ */
281 static int fmt_hyphmarks(char *word
, int *hyidx
, int *hyins
, int *hygap
)
286 while ((c
= escread(&s
, &d
)) > 0)
288 if (c
< 0 || !strcmp(c_hc
, d
))
290 while ((c
= escread(&s
, &d
)) >= 0 && n
< NHYPHSWORD
) {
291 if (!c
&& !strcmp(c_hc
, d
)) {
293 hyidx
[n
++] = s
- word
;
295 if (!c
&& c_hydash(d
)) {
297 hyidx
[n
++] = s
- word
;
299 if (!c
&& !strcmp(c_nb
, d
)) {
301 hyidx
[n
++] = s
- word
;
307 static struct word
*fmt_mkword(struct fmt
*f
)
309 if (f
->words_n
== f
->words_sz
) {
311 f
->words
= mextend(f
->words
, f
->words_n
,
312 f
->words_sz
, sizeof(f
->words
[0]));
314 return &f
->words
[f
->words_n
++];
317 static void fmt_insertword(struct fmt
*f
, struct wb
*wb
, int gap
)
319 int hyidx
[NHYPHSWORD
]; /* sub-word boundaries */
320 int hyins
[NHYPHSWORD
] = {0}; /* insert dash */
321 int hygap
[NHYPHSWORD
] = {0}; /* stretchable no-break space */
322 char *src
= wb_buf(wb
);
328 n
= fmt_hyphmarks(src
, hyidx
, hyins
, hygap
);
330 fmt_wb2word(f
, fmt_mkword(f
), wb
, 0, 1, gap
, wb_cost(wb
));
333 /* update f->fillreq considering the new sub-words */
334 if (f
->fillreq
== f
->words_n
+ 1)
338 for (i
= 0; i
<= n
; i
++) {
339 int ihy
= i
< n
&& hyins
[i
]; /* dash width */
340 int istr
= i
== 0 || hygap
[i
- 1]; /* stretchable */
341 int igap
; /* gap width */
342 int icost
; /* hyphenation cost */
343 beg
= src
+ (i
> 0 ? hyidx
[i
- 1] : 0);
344 end
= src
+ (i
< n
? hyidx
[i
] : strlen(src
));
345 if (i
< n
&& hygap
[i
]) /* remove \~ */
347 wb_catstr(&wbc
, beg
, end
);
348 wb_fnszget(&wbc
, &cf
, &cs
, &cm
);
349 icost
= i
== n
? wb_cost(&wbc
) : hygap
[i
] * 10000000;
350 igap
= i
== 0 ? gap
: hygap
[i
- 1] * font_swid(dev_font(cf
), cs
, n_ss
);
351 fmt_wb2word(f
, fmt_mkword(f
), &wbc
, ihy
, istr
, igap
, icost
);
353 wb_fnszset(&wbc
, cf
, cs
, cm
); /* restoring wbc */
358 /* the amount of space necessary before the next word */
359 static int fmt_wordgap(struct fmt
*f
)
361 int nls
= f
->nls
|| f
->nls_sup
;
362 int swid
= font_swid(dev_font(n_f
), n_s
, n_ss
);
363 if (f
->eos
&& f
->words_n
)
364 if ((nls
&& !f
->gap
) || (!nls
&& f
->gap
== 2 * swid
))
365 return swid
+ font_swid(dev_font(n_f
), n_s
, n_sss
);
366 return (nls
&& !f
->gap
&& f
->words_n
) ? swid
: f
->gap
;
369 /* insert wb into fmt */
370 int fmt_word(struct fmt
*f
, struct wb
*wb
)
374 if (fmt_confchanged(f
))
375 if (fmt_fillwords(f
, 0))
377 if (FMT_FILL(f
) && f
->nls
&& f
->gap
)
380 if (!f
->words_n
) /* apply the new .l and .i */
382 f
->gap
= fmt_wordgap(f
);
384 fmt_insertword(f
, wb
, f
->filled
? 0 : f
->gap
);
392 /* approximate 8 * sqrt(cost) */
393 static long scaledown(long cost
)
397 for (i
= 0; i
< 14; i
++)
398 ret
+= ((cost
>> (i
* 2)) & 3) << (i
+ 3);
399 return ret
< (1 << 13) ? ret
: (1 << 13);
402 /* the cost of putting lwid words in a line of length llen */
403 static long FMT_COST(int llen
, int lwid
, int swid
, int nspc
)
405 /* the ratio that the stretchable spaces of the line should be spread */
406 long ratio
= abs((llen
- lwid
) * 100l / (swid
? swid
: 1));
407 /* ratio too large; scaling it down */
409 ratio
= 4000 + scaledown(ratio
- 4000);
410 /* assigning a cost of 100 to each space stretching 100 percent */
411 return ratio
* ratio
/ 100l * (nspc
? nspc
: 1);
414 /* the number of hyphenations in consecutive lines ending at pos */
415 static int fmt_hydepth(struct fmt
*f
, int pos
)
418 while (pos
> 0 && f
->words
[pos
- 1].hy
&& ++n
< 5)
419 pos
= f
->best_pos
[pos
];
423 static long hycost(int depth
)
425 if (n_hlm
> 0 && depth
> n_hlm
)
428 return n_hycost
+ n_hycost2
+ n_hycost3
;
430 return n_hycost
+ n_hycost2
;
431 return depth
? n_hycost
: 0;
434 /* the cost of putting a line break before word pos */
435 static long fmt_findcost(struct fmt
*f
, int pos
)
439 int llen
= MAX(1, FMT_LLEN(f
));
440 int lwid
= 0; /* current line length */
441 int swid
= 0; /* amount of stretchable spaces */
442 int nspc
= 0; /* number of stretchable spaces */
445 if (f
->best_pos
[pos
] >= 0)
446 return f
->best
[pos
] + f
->words
[pos
- 1].cost
;
447 lwid
= f
->words
[pos
- 1].hy
; /* non-zero if the last word is hyphenated */
448 hyphenated
= f
->words
[pos
- 1].hy
!= 0;
451 lwid
+= f
->words
[i
].wid
;
453 lwid
+= f
->words
[i
+ 1].gap
;
454 if (i
+ 1 < pos
&& f
->words
[i
+ 1].str
) {
455 swid
+= f
->words
[i
+ 1].gap
;
458 if (lwid
> llen
+ swid
* n_ssh
/ 100 && i
+ 1 < pos
)
460 cur
= fmt_findcost(f
, i
) + FMT_COST(llen
, lwid
, swid
, nspc
);
462 cur
+= hycost(1 + fmt_hydepth(f
, i
));
463 if (f
->best_pos
[pos
] < 0 || cur
< f
->best
[pos
]) {
464 f
->best_pos
[pos
] = i
;
465 f
->best_dep
[pos
] = f
->best_dep
[i
] + 1;
470 return f
->best
[pos
] + f
->words
[pos
- 1].cost
;
473 static int fmt_bestpos(struct fmt
*f
, int pos
)
475 fmt_findcost(f
, pos
);
476 return MAX(0, f
->best_pos
[pos
]);
479 static int fmt_bestdep(struct fmt
*f
, int pos
)
481 fmt_findcost(f
, pos
);
482 return MAX(0, f
->best_dep
[pos
]);
485 /* return the last filled word */
486 static int fmt_breakparagraph(struct fmt
*f
, int pos
, int br
)
490 long cost
, best_cost
= 0;
491 int llen
= FMT_LLEN(f
);
492 int lwid
= 0; /* current line length */
493 int swid
= 0; /* amount of stretchable spaces */
494 int nspc
= 0; /* number of stretchable spaces */
495 if (f
->fillreq
> 0 && f
->fillreq
<= f
->words_n
) {
496 fmt_findcost(f
, f
->fillreq
);
499 if (pos
> 0 && f
->words
[pos
- 1].wid
>= llen
) {
500 fmt_findcost(f
, pos
);
505 if (f
->words
[i
].hy
) /* the last word is hyphenated */
506 lwid
+= f
->words
[i
].hy
;
508 lwid
+= f
->words
[i
].wid
;
510 lwid
+= f
->words
[i
+ 1].gap
;
511 if (i
+ 1 < pos
&& f
->words
[i
+ 1].str
) {
512 swid
+= f
->words
[i
+ 1].gap
;
515 if (lwid
> llen
&& i
+ 1 < pos
)
517 cost
= fmt_findcost(f
, i
);
518 /* the cost of formatting short lines; should prevent widows */
519 if (br
&& n_pmll
&& lwid
< llen
* n_pmll
/ 100) {
520 int pmll
= llen
* n_pmll
/ 100;
521 cost
+= (long) n_pmllcost
* (pmll
- lwid
) / pmll
;
523 if (best
< 0 || cost
< best_cost
) {
532 /* extract the first nreq formatted lines before the word at pos */
533 static int fmt_head(struct fmt
*f
, int nreq
, int pos
)
535 int best
= pos
; /* best line break for nreq-th line */
536 int prev
, next
; /* best line breaks without hyphenation */
537 if (nreq
<= 0 || fmt_bestdep(f
, pos
) < nreq
)
539 /* finding the optimal line break for nreq-th line */
540 while (best
> 0 && fmt_bestdep(f
, best
) > nreq
)
541 best
= fmt_bestpos(f
, best
);
544 /* finding closest line breaks without hyphenation */
545 while (prev
> 1 && f
->words
[prev
- 1].hy
&&
546 fmt_bestdep(f
, prev
- 1) == nreq
)
548 while (next
< pos
&& f
->words
[next
- 1].hy
&&
549 fmt_bestdep(f
, next
) == nreq
)
551 /* choosing the best of them */
552 if (!f
->words
[prev
- 1].hy
&& !f
->words
[next
- 1].hy
)
553 return fmt_findcost(f
, prev
) <= fmt_findcost(f
, next
) ? prev
: next
;
554 if (!f
->words
[prev
- 1].hy
)
556 if (!f
->words
[next
- 1].hy
)
561 /* break f->words[0..end] into lines according to fmt_bestpos() */
562 static int fmt_break(struct fmt
*f
, int end
)
565 beg
= fmt_bestpos(f
, end
);
567 ret
+= fmt_break(f
, beg
);
568 f
->words
[beg
].gap
= 0;
569 if (fmt_extractline(f
, beg
, end
, 1))
573 return ret
+ (end
- beg
);
576 /* estimated number of lines until traps or the end of a page */
577 static int fmt_safelines(void)
579 int lnht
= MAX(1, n_L
) * n_v
;
580 return (f_nexttrap() + lnht
- 1) / lnht
;
583 /* fill the words collected in the buffer */
584 static int fmt_fillwords(struct fmt
*f
, int br
)
586 int nreq
; /* the number of lines until a trap */
587 int end
; /* the final line ends before this word */
588 int end_head
; /* like end, but only the first nreq lines included */
589 int head
= 0; /* only nreq first lines have been formatted */
590 int llen
; /* line length, taking shrinkable spaces into account */
594 llen
= fmt_wordslen(f
, 0, f
->words_n
) -
595 fmt_spacessum(f
, 0, f
->words_n
) * n_ssh
/ 100;
596 /* not enough words to fill */
597 if ((f
->fillreq
<= 0 || f
->words_n
< f
->fillreq
) && llen
<= FMT_LLEN(f
))
599 nreq
= (n_hy
& HY_LAST
) ? fmt_safelines() : 0;
600 if (nreq
> 0 && nreq
<= fmt_nlines(f
))
602 /* resetting positions */
603 f
->best
= malloc((f
->words_n
+ 1) * sizeof(f
->best
[0]));
604 f
->best_pos
= malloc((f
->words_n
+ 1) * sizeof(f
->best_pos
[0]));
605 f
->best_dep
= malloc((f
->words_n
+ 1) * sizeof(f
->best_dep
[0]));
606 memset(f
->best
, 0, (f
->words_n
+ 1) * sizeof(f
->best
[0]));
607 memset(f
->best_dep
, 0, (f
->words_n
+ 1) * sizeof(f
->best_dep
[0]));
608 for (i
= 0; i
< f
->words_n
+ 1; i
++)
610 end
= fmt_breakparagraph(f
, f
->words_n
, br
);
612 end_head
= fmt_head(f
, nreq
- fmt_nlines(f
), end
);
613 head
= end_head
< end
;
616 /* recursively add lines */
617 n
= end
> 0 ? fmt_break(f
, end
) : 0;
620 fmt_movewords(f
, 0, n
, f
->words_n
);
621 f
->filled
= n
&& !f
->words_n
;
624 if (f
->words_n
) /* apply the new .l and .i */
632 return head
|| n
!= end
;
635 struct fmt
*fmt_alloc(void)
637 struct fmt
*fmt
= xmalloc(sizeof(*fmt
));
638 memset(fmt
, 0, sizeof(*fmt
));
642 void fmt_free(struct fmt
*fmt
)
649 int fmt_wid(struct fmt
*fmt
)
651 return fmt_wordslen(fmt
, 0, fmt
->words_n
) + fmt_wordgap(fmt
);
654 int fmt_morewords(struct fmt
*fmt
)
656 return fmt_morelines(fmt
) || fmt
->words_n
;
659 int fmt_morelines(struct fmt
*fmt
)
661 return fmt
->lines_head
!= fmt
->lines_tail
;
664 /* suppress the last newline */
665 void fmt_suppressnl(struct fmt
*fmt
)