1 /* $Id: term_ps.c,v 1.69 2014/11/20 13:56:20 schwarze Exp $ */
3 * Copyright (c) 2010, 2011 Kristaps Dzonsons <kristaps@bsd.lv>
4 * Copyright (c) 2014 Ingo Schwarze <schwarze@openbsd.org>
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.
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.
20 #include <sys/types.h>
31 #include "mandoc_aux.h"
36 /* These work the buffer used by the header and footer. */
37 #define PS_BUFSLOP 128
39 /* Convert PostScript point "x" to an AFM unit. */
40 #define PNT2AFM(p, x) \
41 (size_t)((double)(x) * (1000.0 / (double)(p)->ps->scale))
43 /* Convert an AFM unit "x" to a PostScript points */
44 #define AFM2PNT(p, x) \
45 ((double)(x) / (1000.0 / (double)(p)->ps->scale))
48 unsigned short wx; /* WX in AFM */
52 const char *name; /* FontName in AFM */
53 #define MAXCHAR 95 /* total characters we can handle */
54 struct glyph gly[MAXCHAR]; /* glyph metrics */
59 #define PS_INLINE (1 << 0) /* we're in a word */
60 #define PS_MARGINS (1 << 1) /* we're in the margins */
61 #define PS_NEWPAGE (1 << 2) /* new page, no words yet */
62 #define PS_BACKSP (1 << 3) /* last character was backspace */
63 size_t pscol; /* visible column (AFM units) */
64 size_t psrow; /* visible row (AFM units) */
65 char *psmarg; /* margin buf */
66 size_t psmargsz; /* margin buf size */
67 size_t psmargcur; /* cur index in margin buf */
68 char last; /* last non-backspace seen */
69 enum termfont lastf; /* last set font */
70 enum termfont nextf; /* building next font here */
71 size_t scale; /* font scaling factor */
72 size_t pages; /* number of pages shown */
73 size_t lineheight; /* line height (AFM units) */
74 size_t top; /* body top (AFM units) */
75 size_t bottom; /* body bottom (AFM units) */
76 size_t height; /* page height (AFM units */
77 size_t width; /* page width (AFM units) */
78 size_t lastwidth; /* page width before last ll */
79 size_t left; /* body left (AFM units) */
80 size_t header; /* header pos (AFM units) */
81 size_t footer; /* footer pos (AFM units) */
82 size_t pdfbytes; /* current output byte */
83 size_t pdflastpg; /* byte of last page mark */
84 size_t pdfbody; /* start of body object */
85 size_t *pdfobjs; /* table of object offsets */
86 size_t pdfobjsz; /* size of pdfobjs */
89 static double ps_hspan(const struct termp *,
90 const struct roffsu *);
91 static size_t ps_width(const struct termp *, int);
92 static void ps_advance(struct termp *, size_t);
93 static void ps_begin(struct termp *);
94 static void ps_closepage(struct termp *);
95 static void ps_end(struct termp *);
96 static void ps_endline(struct termp *);
97 static void ps_fclose(struct termp *);
98 static void ps_growbuf(struct termp *, size_t);
99 static void ps_letter(struct termp *, int);
100 static void ps_pclose(struct termp *);
101 static void ps_pletter(struct termp *, int);
102 #if __GNUC__ - 0 >= 4
103 __attribute__((__format__ (__printf__, 2, 3)))
105 static void ps_printf(struct termp *, const char *, ...);
106 static void ps_putchar(struct termp *, char);
107 static void ps_setfont(struct termp *, enum termfont);
108 static void ps_setwidth(struct termp *, int, size_t);
109 static struct termp *pspdf_alloc(const struct mchars *, char *);
110 static void pdf_obj(struct termp *, size_t);
113 * We define, for the time being, three fonts: bold, oblique/italic, and
114 * normal (roman). The following table hard-codes the font metrics for
115 * ASCII, i.e., 32--127.
118 static const struct font fonts[TERMFONT__MAX] = {
410 { "Times-BoldItalic", {
510 pdf_alloc(const struct mchars *mchars, char *outopts)
514 if (NULL != (p = pspdf_alloc(mchars, outopts)))
515 p->type = TERMTYPE_PDF;
521 ps_alloc(const struct mchars *mchars, char *outopts)
525 if (NULL != (p = pspdf_alloc(mchars, outopts)))
526 p->type = TERMTYPE_PS;
531 static struct termp *
532 pspdf_alloc(const struct mchars *mchars, char *outopts)
535 unsigned int pagex, pagey;
536 size_t marginx, marginy, lineheight;
541 p = mandoc_calloc(1, sizeof(struct termp));
543 p->enc = TERMENC_ASCII;
544 p->ps = mandoc_calloc(1, sizeof(struct termp_ps));
546 p->advance = ps_advance;
549 p->endline = ps_endline;
551 p->letter = ps_letter;
552 p->setwidth = ps_setwidth;
560 while (outopts && *outopts)
561 switch (getsubopt(&outopts, UNCONST(toks), &v)) {
569 /* Default to US letter (millimetres). */
575 * The ISO-269 paper sizes can be calculated automatically, but
576 * it would require bringing in -lm for pow() and I'd rather not
577 * do that. So just do it the easy way for now. Since this
578 * only happens once, I'm not terribly concerned.
581 if (pp && strcasecmp(pp, "letter")) {
582 if (0 == strcasecmp(pp, "a3")) {
585 } else if (0 == strcasecmp(pp, "a4")) {
588 } else if (0 == strcasecmp(pp, "a5")) {
591 } else if (0 == strcasecmp(pp, "legal")) {
594 } else if (2 != sscanf(pp, "%ux%u", &pagex, &pagey))
595 fprintf(stderr, "%s: Unknown paper\n", pp);
599 * This MUST be defined before any PNT2AFM or AFM2PNT
600 * calculations occur.
605 /* Remember millimetres -> AFM units. */
607 pagex = PNT2AFM(p, ((double)pagex * 2.834));
608 pagey = PNT2AFM(p, ((double)pagey * 2.834));
610 /* Margins are 1/9 the page x and y. */
612 marginx = (size_t)((double)pagex / 9.0);
613 marginy = (size_t)((double)pagey / 9.0);
615 /* Line-height is 1.4em. */
617 lineheight = PNT2AFM(p, ((double)p->ps->scale * 1.4));
619 p->ps->width = p->ps->lastwidth = (size_t)pagex;
620 p->ps->height = (size_t)pagey;
621 p->ps->header = pagey - (marginy / 2) - (lineheight / 2);
622 p->ps->top = pagey - marginy;
623 p->ps->footer = (marginy / 2) - (lineheight / 2);
624 p->ps->bottom = marginy;
625 p->ps->left = marginx;
626 p->ps->lineheight = lineheight;
628 p->defrmargin = pagex - (marginx * 2);
633 ps_setwidth(struct termp *p, int iop, size_t width)
637 lastwidth = p->ps->width;
639 p->ps->width += width;
641 p->ps->width = width ? width : p->ps->lastwidth;
642 else if (p->ps->width > width)
643 p->ps->width -= width;
646 p->ps->lastwidth = lastwidth;
650 pspdf_free(void *arg)
654 p = (struct termp *)arg;
659 free(p->ps->pdfobjs);
666 ps_printf(struct termp *p, const char *fmt, ...)
674 * If we're running in regular mode, then pipe directly into
675 * vprintf(). If we're processing margins, then push the data
676 * into our growable margin buffer.
679 if ( ! (PS_MARGINS & p->ps->flags)) {
680 len = vprintf(fmt, ap);
682 p->ps->pdfbytes += len < 0 ? 0 : (size_t)len;
687 * XXX: I assume that the in-margin print won't exceed
688 * PS_BUFSLOP (128 bytes), which is reasonable but still an
689 * assumption that will cause pukeage if it's not the case.
692 ps_growbuf(p, PS_BUFSLOP);
694 pos = (int)p->ps->psmargcur;
695 vsnprintf(&p->ps->psmarg[pos], PS_BUFSLOP, fmt, ap);
699 p->ps->psmargcur = strlen(p->ps->psmarg);
703 ps_putchar(struct termp *p, char c)
707 /* See ps_printf(). */
709 if ( ! (PS_MARGINS & p->ps->flags)) {
717 pos = (int)p->ps->psmargcur++;
718 p->ps->psmarg[pos++] = c;
719 p->ps->psmarg[pos] = '\0';
723 pdf_obj(struct termp *p, size_t obj)
728 if ((obj - 1) >= p->ps->pdfobjsz) {
729 p->ps->pdfobjsz = obj + 128;
730 p->ps->pdfobjs = mandoc_reallocarray(p->ps->pdfobjs,
731 p->ps->pdfobjsz, sizeof(size_t));
734 p->ps->pdfobjs[(int)obj - 1] = p->ps->pdfbytes;
735 ps_printf(p, "%zu 0 obj\n", obj);
739 ps_closepage(struct termp *p)
745 * Close out a page that we've already flushed to output. In
746 * PostScript, we simply note that the page must be showed. In
747 * PDF, we must now create the Length, Resource, and Page node
748 * for the page contents.
751 assert(p->ps->psmarg && p->ps->psmarg[0]);
752 ps_printf(p, "%s", p->ps->psmarg);
754 if (TERMTYPE_PS != p->type) {
755 ps_printf(p, "ET\n");
757 len = p->ps->pdfbytes - p->ps->pdflastpg;
758 base = p->ps->pages * 4 + p->ps->pdfbody;
760 ps_printf(p, "endstream\nendobj\n");
762 /* Length of content. */
763 pdf_obj(p, base + 1);
764 ps_printf(p, "%zu\nendobj\n", len);
766 /* Resource for content. */
767 pdf_obj(p, base + 2);
768 ps_printf(p, "<<\n/ProcSet [/PDF /Text]\n");
769 ps_printf(p, "/Font <<\n");
770 for (i = 0; i < (int)TERMFONT__MAX; i++)
771 ps_printf(p, "/F%d %d 0 R\n", i, 3 + i);
772 ps_printf(p, ">>\n>>\n");
775 pdf_obj(p, base + 3);
776 ps_printf(p, "<<\n");
777 ps_printf(p, "/Type /Page\n");
778 ps_printf(p, "/Parent 2 0 R\n");
779 ps_printf(p, "/Resources %zu 0 R\n", base + 2);
780 ps_printf(p, "/Contents %zu 0 R\n", base);
781 ps_printf(p, ">>\nendobj\n");
783 ps_printf(p, "showpage\n");
786 p->ps->psrow = p->ps->top;
787 assert( ! (PS_NEWPAGE & p->ps->flags));
788 p->ps->flags |= PS_NEWPAGE;
792 ps_end(struct termp *p)
794 size_t i, xref, base;
797 * At the end of the file, do one last showpage. This is the
798 * same behaviour as groff(1) and works for multiple pages as
802 if ( ! (PS_NEWPAGE & p->ps->flags)) {
803 assert(0 == p->ps->flags);
804 assert('\0' == p->ps->last);
808 if (TERMTYPE_PS == p->type) {
809 ps_printf(p, "%%%%Trailer\n");
810 ps_printf(p, "%%%%Pages: %zu\n", p->ps->pages);
811 ps_printf(p, "%%%%EOF\n");
816 ps_printf(p, "<<\n/Type /Pages\n");
817 ps_printf(p, "/MediaBox [0 0 %zu %zu]\n",
818 (size_t)AFM2PNT(p, p->ps->width),
819 (size_t)AFM2PNT(p, p->ps->height));
821 ps_printf(p, "/Count %zu\n", p->ps->pages);
822 ps_printf(p, "/Kids [");
824 for (i = 0; i < p->ps->pages; i++)
825 ps_printf(p, " %zu 0 R", i * 4 + p->ps->pdfbody + 3);
827 base = (p->ps->pages - 1) * 4 + p->ps->pdfbody + 4;
829 ps_printf(p, "]\n>>\nendobj\n");
831 ps_printf(p, "<<\n");
832 ps_printf(p, "/Type /Catalog\n");
833 ps_printf(p, "/Pages 2 0 R\n");
834 ps_printf(p, ">>\n");
835 xref = p->ps->pdfbytes;
836 ps_printf(p, "xref\n");
837 ps_printf(p, "0 %zu\n", base + 1);
838 ps_printf(p, "0000000000 65535 f \n");
840 for (i = 0; i < base; i++)
841 ps_printf(p, "%.10zu 00000 n \n",
842 p->ps->pdfobjs[(int)i]);
844 ps_printf(p, "trailer\n");
845 ps_printf(p, "<<\n");
846 ps_printf(p, "/Size %zu\n", base + 1);
847 ps_printf(p, "/Root %zu 0 R\n", base);
848 ps_printf(p, "/Info 1 0 R\n");
849 ps_printf(p, ">>\n");
850 ps_printf(p, "startxref\n");
851 ps_printf(p, "%zu\n", xref);
852 ps_printf(p, "%%%%EOF\n");
856 ps_begin(struct termp *p)
861 * Print margins into margin buffer. Nothing gets output to the
862 * screen yet, so we don't need to initialise the primary state.
866 assert(p->ps->psmargsz);
867 p->ps->psmarg[0] = '\0';
870 /*p->ps->pdfbytes = 0;*/
871 p->ps->psmargcur = 0;
872 p->ps->flags = PS_MARGINS;
873 p->ps->pscol = p->ps->left;
874 p->ps->psrow = p->ps->header;
876 ps_setfont(p, TERMFONT_NONE);
878 (*p->headf)(p, p->argf);
881 p->ps->pscol = p->ps->left;
882 p->ps->psrow = p->ps->footer;
884 (*p->footf)(p, p->argf);
887 p->ps->flags &= ~PS_MARGINS;
889 assert(0 == p->ps->flags);
890 assert(p->ps->psmarg);
891 assert('\0' != p->ps->psmarg[0]);
894 * Print header and initialise page state. Following this,
895 * stuff gets printed to the screen, so make sure we're sane.
898 if (TERMTYPE_PS == p->type) {
899 ps_printf(p, "%%!PS-Adobe-3.0\n");
900 ps_printf(p, "%%%%DocumentData: Clean7Bit\n");
901 ps_printf(p, "%%%%Orientation: Portrait\n");
902 ps_printf(p, "%%%%Pages: (atend)\n");
903 ps_printf(p, "%%%%PageOrder: Ascend\n");
904 ps_printf(p, "%%%%DocumentMedia: "
905 "Default %zu %zu 0 () ()\n",
906 (size_t)AFM2PNT(p, p->ps->width),
907 (size_t)AFM2PNT(p, p->ps->height));
908 ps_printf(p, "%%%%DocumentNeededResources: font");
910 for (i = 0; i < (int)TERMFONT__MAX; i++)
911 ps_printf(p, " %s", fonts[i].name);
913 ps_printf(p, "\n%%%%EndComments\n");
915 ps_printf(p, "%%PDF-1.1\n");
917 ps_printf(p, "<<\n");
918 ps_printf(p, ">>\n");
919 ps_printf(p, "endobj\n");
921 for (i = 0; i < (int)TERMFONT__MAX; i++) {
922 pdf_obj(p, (size_t)i + 3);
923 ps_printf(p, "<<\n");
924 ps_printf(p, "/Type /Font\n");
925 ps_printf(p, "/Subtype /Type1\n");
926 ps_printf(p, "/Name /F%d\n", i);
927 ps_printf(p, "/BaseFont /%s\n", fonts[i].name);
928 ps_printf(p, ">>\n");
932 p->ps->pdfbody = (size_t)TERMFONT__MAX + 3;
933 p->ps->pscol = p->ps->left;
934 p->ps->psrow = p->ps->top;
935 p->ps->flags |= PS_NEWPAGE;
936 ps_setfont(p, TERMFONT_NONE);
940 ps_pletter(struct termp *p, int c)
945 * If we haven't opened a page context, then output that we're
946 * in a new page and make sure the font is correctly set.
949 if (PS_NEWPAGE & p->ps->flags) {
950 if (TERMTYPE_PS == p->type) {
951 ps_printf(p, "%%%%Page: %zu %zu\n",
952 p->ps->pages + 1, p->ps->pages + 1);
953 ps_printf(p, "/%s %zu selectfont\n",
954 fonts[(int)p->ps->lastf].name,
957 pdf_obj(p, p->ps->pdfbody +
959 ps_printf(p, "<<\n");
960 ps_printf(p, "/Length %zu 0 R\n",
961 p->ps->pdfbody + 1 + p->ps->pages * 4);
962 ps_printf(p, ">>\nstream\n");
964 p->ps->pdflastpg = p->ps->pdfbytes;
965 p->ps->flags &= ~PS_NEWPAGE;
969 * If we're not in a PostScript "word" context, then open one
970 * now at the current cursor.
973 if ( ! (PS_INLINE & p->ps->flags)) {
974 if (TERMTYPE_PS != p->type) {
975 ps_printf(p, "BT\n/F%d %zu Tf\n",
976 (int)p->ps->lastf, p->ps->scale);
977 ps_printf(p, "%.3f %.3f Td\n(",
978 AFM2PNT(p, p->ps->pscol),
979 AFM2PNT(p, p->ps->psrow));
981 ps_printf(p, "%.3f %.3f moveto\n(",
982 AFM2PNT(p, p->ps->pscol),
983 AFM2PNT(p, p->ps->psrow));
984 p->ps->flags |= PS_INLINE;
987 assert( ! (PS_NEWPAGE & p->ps->flags));
990 * We need to escape these characters as per the PostScript
991 * specification. We would also escape non-graphable characters
992 * (like tabs), but none of them would get to this point and
993 * it's superfluous to abort() on them.
1002 ps_putchar(p, '\\');
1008 /* Write the character and adjust where we are on the page. */
1010 f = (int)p->ps->lastf;
1012 if (c <= 32 || c - 32 >= MAXCHAR)
1015 ps_putchar(p, (char)c);
1017 p->ps->pscol += (size_t)fonts[f].gly[c].wx;
1021 ps_pclose(struct termp *p)
1025 * Spit out that we're exiting a word context (this is a
1026 * "partial close" because we don't check the last-char buffer
1030 if ( ! (PS_INLINE & p->ps->flags))
1033 if (TERMTYPE_PS != p->type) {
1034 ps_printf(p, ") Tj\nET\n");
1036 ps_printf(p, ") show\n");
1038 p->ps->flags &= ~PS_INLINE;
1042 ps_fclose(struct termp *p)
1046 * Strong closure: if we have a last-char, spit it out after
1047 * checking that we're in the right font mode. This will of
1048 * course open a new scope, if applicable.
1050 * Following this, close out any scope that's open.
1053 if (p->ps->last != '\0') {
1054 assert( ! (p->ps->flags & PS_BACKSP));
1055 if (p->ps->nextf != p->ps->lastf) {
1057 ps_setfont(p, p->ps->nextf);
1059 p->ps->nextf = TERMFONT_NONE;
1060 ps_pletter(p, p->ps->last);
1064 if ( ! (PS_INLINE & p->ps->flags))
1071 ps_letter(struct termp *p, int arg)
1076 c = arg >= 128 || arg <= 0 ? '?' : arg;
1079 * When receiving a backspace, merely flag it.
1080 * We don't know yet whether it is
1081 * a font instruction or an overstrike.
1085 assert(p->ps->last != '\0');
1086 assert( ! (p->ps->flags & PS_BACKSP));
1087 p->ps->flags |= PS_BACKSP;
1092 * Decode font instructions.
1095 if (p->ps->flags & PS_BACKSP) {
1096 if (p->ps->last == '_') {
1097 switch (p->ps->nextf) {
1101 p->ps->nextf = TERMFONT_BI;
1104 p->ps->nextf = TERMFONT_UNDER;
1107 p->ps->flags &= ~PS_BACKSP;
1110 if (p->ps->last == c) {
1111 switch (p->ps->nextf) {
1114 case TERMFONT_UNDER:
1115 p->ps->nextf = TERMFONT_BI;
1118 p->ps->nextf = TERMFONT_BOLD;
1120 p->ps->flags &= ~PS_BACKSP;
1125 * This is not a font instruction, but rather
1126 * the next character. Prepare for overstrike.
1129 savecol = p->ps->pscol;
1134 * We found the next character, so the font instructions
1135 * for the previous one are complete.
1136 * Use them and print it.
1139 if (p->ps->last != '\0') {
1140 if (p->ps->nextf != p->ps->lastf) {
1142 ps_setfont(p, p->ps->nextf);
1144 p->ps->nextf = TERMFONT_NONE;
1145 ps_pletter(p, p->ps->last);
1149 * Do not print the current character yet because font
1150 * instructions might follow; only remember it.
1151 * For the first character, nothing else is done.
1152 * The final character will get printed from ps_fclose().
1158 * For an overstrike, back up to the previous position.
1161 if (savecol != SIZE_MAX) {
1163 p->ps->pscol = savecol;
1164 p->ps->flags &= ~PS_BACKSP;
1169 ps_advance(struct termp *p, size_t len)
1173 * Advance some spaces. This can probably be made smarter,
1174 * i.e., to have multiple space-separated words in the same
1175 * scope, but this is easier: just close out the current scope
1176 * and readjust our column settings.
1180 p->ps->pscol += len;
1184 ps_endline(struct termp *p)
1187 /* Close out any scopes we have open: we're at eoln. */
1192 * If we're in the margin, don't try to recalculate our current
1193 * row. XXX: if the column tries to be fancy with multiple
1194 * lines, we'll do nasty stuff.
1197 if (PS_MARGINS & p->ps->flags)
1202 p->ps->pscol = p->ps->left;
1204 /* If we haven't printed anything, return. */
1206 if (PS_NEWPAGE & p->ps->flags)
1210 * Put us down a line. If we're at the page bottom, spit out a
1211 * showpage and restart our row.
1214 if (p->ps->psrow >= p->ps->lineheight + p->ps->bottom) {
1215 p->ps->psrow -= p->ps->lineheight;
1223 ps_setfont(struct termp *p, enum termfont f)
1226 assert(f < TERMFONT__MAX);
1230 * If we're still at the top of the page, let the font-setting
1231 * be delayed until we actually have stuff to print.
1234 if (PS_NEWPAGE & p->ps->flags)
1237 if (TERMTYPE_PS == p->type)
1238 ps_printf(p, "/%s %zu selectfont\n",
1239 fonts[(int)f].name, p->ps->scale);
1241 ps_printf(p, "/F%d %zu Tf\n",
1242 (int)f, p->ps->scale);
1246 ps_width(const struct termp *p, int c)
1249 if (c <= 32 || c - 32 >= MAXCHAR)
1254 return((size_t)fonts[(int)TERMFONT_NONE].gly[c].wx);
1258 ps_hspan(const struct termp *p, const struct roffsu *su)
1263 * All of these measurements are derived by converting from the
1264 * native measurement to AFM units.
1269 * Traditionally, the default unit is fixed to the
1270 * output media. So this would refer to the point. In
1271 * mandoc(1), however, we stick to the default terminal
1272 * scaling unit so that output is the same regardless
1275 r = PNT2AFM(p, su->scale * 72.0 / 240.0);
1278 r = PNT2AFM(p, su->scale * 72.0 / 2.54);
1282 fonts[(int)TERMFONT_NONE].gly[109 - 32].wx;
1286 fonts[(int)TERMFONT_NONE].gly[110 - 32].wx;
1289 r = PNT2AFM(p, su->scale * 72.0);
1293 fonts[(int)TERMFONT_NONE].gly[109 - 32].wx / 100.0;
1296 r = PNT2AFM(p, su->scale * 12.0);
1299 r = PNT2AFM(p, su->scale * 1.0);
1302 r = su->scale * p->ps->lineheight;
1313 ps_growbuf(struct termp *p, size_t sz)
1315 if (p->ps->psmargcur + sz <= p->ps->psmargsz)
1318 if (sz < PS_BUFSLOP)
1321 p->ps->psmargsz += sz;
1322 p->ps->psmarg = mandoc_realloc(p->ps->psmarg, p->ps->psmargsz);