1 /* $Id: term_ps.c,v 1.70 2014/12/01 08:05:52 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>
30 #include "mandoc_aux.h"
35 /* These work the buffer used by the header and footer. */
36 #define PS_BUFSLOP 128
38 /* Convert PostScript point "x" to an AFM unit. */
39 #define PNT2AFM(p, x) \
40 (size_t)((double)(x) * (1000.0 / (double)(p)->ps->scale))
42 /* Convert an AFM unit "x" to a PostScript points */
43 #define AFM2PNT(p, x) \
44 ((double)(x) / (1000.0 / (double)(p)->ps->scale))
47 unsigned short wx; /* WX in AFM */
51 const char *name; /* FontName in AFM */
52 #define MAXCHAR 95 /* total characters we can handle */
53 struct glyph gly[MAXCHAR]; /* glyph metrics */
58 #define PS_INLINE (1 << 0) /* we're in a word */
59 #define PS_MARGINS (1 << 1) /* we're in the margins */
60 #define PS_NEWPAGE (1 << 2) /* new page, no words yet */
61 #define PS_BACKSP (1 << 3) /* last character was backspace */
62 size_t pscol; /* visible column (AFM units) */
63 size_t psrow; /* visible row (AFM units) */
64 char *psmarg; /* margin buf */
65 size_t psmargsz; /* margin buf size */
66 size_t psmargcur; /* cur index in margin buf */
67 char last; /* last non-backspace seen */
68 enum termfont lastf; /* last set font */
69 enum termfont nextf; /* building next font here */
70 size_t scale; /* font scaling factor */
71 size_t pages; /* number of pages shown */
72 size_t lineheight; /* line height (AFM units) */
73 size_t top; /* body top (AFM units) */
74 size_t bottom; /* body bottom (AFM units) */
75 size_t height; /* page height (AFM units */
76 size_t width; /* page width (AFM units) */
77 size_t lastwidth; /* page width before last ll */
78 size_t left; /* body left (AFM units) */
79 size_t header; /* header pos (AFM units) */
80 size_t footer; /* footer pos (AFM units) */
81 size_t pdfbytes; /* current output byte */
82 size_t pdflastpg; /* byte of last page mark */
83 size_t pdfbody; /* start of body object */
84 size_t *pdfobjs; /* table of object offsets */
85 size_t pdfobjsz; /* size of pdfobjs */
88 static double ps_hspan(const struct termp *,
89 const struct roffsu *);
90 static size_t ps_width(const struct termp *, int);
91 static void ps_advance(struct termp *, size_t);
92 static void ps_begin(struct termp *);
93 static void ps_closepage(struct termp *);
94 static void ps_end(struct termp *);
95 static void ps_endline(struct termp *);
96 static void ps_fclose(struct termp *);
97 static void ps_growbuf(struct termp *, size_t);
98 static void ps_letter(struct termp *, int);
99 static void ps_pclose(struct termp *);
100 static void ps_pletter(struct termp *, int);
101 #if __GNUC__ - 0 >= 4
102 __attribute__((__format__ (__printf__, 2, 3)))
104 static void ps_printf(struct termp *, const char *, ...);
105 static void ps_putchar(struct termp *, char);
106 static void ps_setfont(struct termp *, enum termfont);
107 static void ps_setwidth(struct termp *, int, size_t);
108 static struct termp *pspdf_alloc(const struct mchars *, char *);
109 static void pdf_obj(struct termp *, size_t);
112 * We define, for the time being, three fonts: bold, oblique/italic, and
113 * normal (roman). The following table hard-codes the font metrics for
114 * ASCII, i.e., 32--127.
117 static const struct font fonts[TERMFONT__MAX] = {
409 { "Times-BoldItalic", {
509 pdf_alloc(const struct mchars *mchars, char *outopts)
513 if (NULL != (p = pspdf_alloc(mchars, outopts)))
514 p->type = TERMTYPE_PDF;
520 ps_alloc(const struct mchars *mchars, char *outopts)
524 if (NULL != (p = pspdf_alloc(mchars, outopts)))
525 p->type = TERMTYPE_PS;
530 static struct termp *
531 pspdf_alloc(const struct mchars *mchars, char *outopts)
534 unsigned int pagex, pagey;
535 size_t marginx, marginy, lineheight;
540 p = mandoc_calloc(1, sizeof(struct termp));
542 p->enc = TERMENC_ASCII;
543 p->ps = mandoc_calloc(1, sizeof(struct termp_ps));
545 p->advance = ps_advance;
548 p->endline = ps_endline;
550 p->letter = ps_letter;
551 p->setwidth = ps_setwidth;
559 while (outopts && *outopts)
560 switch (getsubopt(&outopts, UNCONST(toks), &v)) {
568 /* Default to US letter (millimetres). */
574 * The ISO-269 paper sizes can be calculated automatically, but
575 * it would require bringing in -lm for pow() and I'd rather not
576 * do that. So just do it the easy way for now. Since this
577 * only happens once, I'm not terribly concerned.
580 if (pp && strcasecmp(pp, "letter")) {
581 if (0 == strcasecmp(pp, "a3")) {
584 } else if (0 == strcasecmp(pp, "a4")) {
587 } else if (0 == strcasecmp(pp, "a5")) {
590 } else if (0 == strcasecmp(pp, "legal")) {
593 } else if (2 != sscanf(pp, "%ux%u", &pagex, &pagey))
594 fprintf(stderr, "%s: Unknown paper\n", pp);
598 * This MUST be defined before any PNT2AFM or AFM2PNT
599 * calculations occur.
604 /* Remember millimetres -> AFM units. */
606 pagex = PNT2AFM(p, ((double)pagex * 2.834));
607 pagey = PNT2AFM(p, ((double)pagey * 2.834));
609 /* Margins are 1/9 the page x and y. */
611 marginx = (size_t)((double)pagex / 9.0);
612 marginy = (size_t)((double)pagey / 9.0);
614 /* Line-height is 1.4em. */
616 lineheight = PNT2AFM(p, ((double)p->ps->scale * 1.4));
618 p->ps->width = p->ps->lastwidth = (size_t)pagex;
619 p->ps->height = (size_t)pagey;
620 p->ps->header = pagey - (marginy / 2) - (lineheight / 2);
621 p->ps->top = pagey - marginy;
622 p->ps->footer = (marginy / 2) - (lineheight / 2);
623 p->ps->bottom = marginy;
624 p->ps->left = marginx;
625 p->ps->lineheight = lineheight;
627 p->defrmargin = pagex - (marginx * 2);
632 ps_setwidth(struct termp *p, int iop, size_t width)
636 lastwidth = p->ps->width;
638 p->ps->width += width;
640 p->ps->width = width ? width : p->ps->lastwidth;
641 else if (p->ps->width > width)
642 p->ps->width -= width;
645 p->ps->lastwidth = lastwidth;
649 pspdf_free(void *arg)
653 p = (struct termp *)arg;
658 free(p->ps->pdfobjs);
665 ps_printf(struct termp *p, const char *fmt, ...)
673 * If we're running in regular mode, then pipe directly into
674 * vprintf(). If we're processing margins, then push the data
675 * into our growable margin buffer.
678 if ( ! (PS_MARGINS & p->ps->flags)) {
679 len = vprintf(fmt, ap);
681 p->ps->pdfbytes += len < 0 ? 0 : (size_t)len;
686 * XXX: I assume that the in-margin print won't exceed
687 * PS_BUFSLOP (128 bytes), which is reasonable but still an
688 * assumption that will cause pukeage if it's not the case.
691 ps_growbuf(p, PS_BUFSLOP);
693 pos = (int)p->ps->psmargcur;
694 vsnprintf(&p->ps->psmarg[pos], PS_BUFSLOP, fmt, ap);
698 p->ps->psmargcur = strlen(p->ps->psmarg);
702 ps_putchar(struct termp *p, char c)
706 /* See ps_printf(). */
708 if ( ! (PS_MARGINS & p->ps->flags)) {
716 pos = (int)p->ps->psmargcur++;
717 p->ps->psmarg[pos++] = c;
718 p->ps->psmarg[pos] = '\0';
722 pdf_obj(struct termp *p, size_t obj)
727 if ((obj - 1) >= p->ps->pdfobjsz) {
728 p->ps->pdfobjsz = obj + 128;
729 p->ps->pdfobjs = mandoc_reallocarray(p->ps->pdfobjs,
730 p->ps->pdfobjsz, sizeof(size_t));
733 p->ps->pdfobjs[(int)obj - 1] = p->ps->pdfbytes;
734 ps_printf(p, "%zu 0 obj\n", obj);
738 ps_closepage(struct termp *p)
744 * Close out a page that we've already flushed to output. In
745 * PostScript, we simply note that the page must be showed. In
746 * PDF, we must now create the Length, Resource, and Page node
747 * for the page contents.
750 assert(p->ps->psmarg && p->ps->psmarg[0]);
751 ps_printf(p, "%s", p->ps->psmarg);
753 if (TERMTYPE_PS != p->type) {
754 ps_printf(p, "ET\n");
756 len = p->ps->pdfbytes - p->ps->pdflastpg;
757 base = p->ps->pages * 4 + p->ps->pdfbody;
759 ps_printf(p, "endstream\nendobj\n");
761 /* Length of content. */
762 pdf_obj(p, base + 1);
763 ps_printf(p, "%zu\nendobj\n", len);
765 /* Resource for content. */
766 pdf_obj(p, base + 2);
767 ps_printf(p, "<<\n/ProcSet [/PDF /Text]\n");
768 ps_printf(p, "/Font <<\n");
769 for (i = 0; i < (int)TERMFONT__MAX; i++)
770 ps_printf(p, "/F%d %d 0 R\n", i, 3 + i);
771 ps_printf(p, ">>\n>>\n");
774 pdf_obj(p, base + 3);
775 ps_printf(p, "<<\n");
776 ps_printf(p, "/Type /Page\n");
777 ps_printf(p, "/Parent 2 0 R\n");
778 ps_printf(p, "/Resources %zu 0 R\n", base + 2);
779 ps_printf(p, "/Contents %zu 0 R\n", base);
780 ps_printf(p, ">>\nendobj\n");
782 ps_printf(p, "showpage\n");
785 p->ps->psrow = p->ps->top;
786 assert( ! (PS_NEWPAGE & p->ps->flags));
787 p->ps->flags |= PS_NEWPAGE;
791 ps_end(struct termp *p)
793 size_t i, xref, base;
796 * At the end of the file, do one last showpage. This is the
797 * same behaviour as groff(1) and works for multiple pages as
801 if ( ! (PS_NEWPAGE & p->ps->flags)) {
802 assert(0 == p->ps->flags);
803 assert('\0' == p->ps->last);
807 if (TERMTYPE_PS == p->type) {
808 ps_printf(p, "%%%%Trailer\n");
809 ps_printf(p, "%%%%Pages: %zu\n", p->ps->pages);
810 ps_printf(p, "%%%%EOF\n");
815 ps_printf(p, "<<\n/Type /Pages\n");
816 ps_printf(p, "/MediaBox [0 0 %zu %zu]\n",
817 (size_t)AFM2PNT(p, p->ps->width),
818 (size_t)AFM2PNT(p, p->ps->height));
820 ps_printf(p, "/Count %zu\n", p->ps->pages);
821 ps_printf(p, "/Kids [");
823 for (i = 0; i < p->ps->pages; i++)
824 ps_printf(p, " %zu 0 R", i * 4 + p->ps->pdfbody + 3);
826 base = (p->ps->pages - 1) * 4 + p->ps->pdfbody + 4;
828 ps_printf(p, "]\n>>\nendobj\n");
830 ps_printf(p, "<<\n");
831 ps_printf(p, "/Type /Catalog\n");
832 ps_printf(p, "/Pages 2 0 R\n");
833 ps_printf(p, ">>\n");
834 xref = p->ps->pdfbytes;
835 ps_printf(p, "xref\n");
836 ps_printf(p, "0 %zu\n", base + 1);
837 ps_printf(p, "0000000000 65535 f \n");
839 for (i = 0; i < base; i++)
840 ps_printf(p, "%.10zu 00000 n \n",
841 p->ps->pdfobjs[(int)i]);
843 ps_printf(p, "trailer\n");
844 ps_printf(p, "<<\n");
845 ps_printf(p, "/Size %zu\n", base + 1);
846 ps_printf(p, "/Root %zu 0 R\n", base);
847 ps_printf(p, "/Info 1 0 R\n");
848 ps_printf(p, ">>\n");
849 ps_printf(p, "startxref\n");
850 ps_printf(p, "%zu\n", xref);
851 ps_printf(p, "%%%%EOF\n");
855 ps_begin(struct termp *p)
860 * Print margins into margin buffer. Nothing gets output to the
861 * screen yet, so we don't need to initialise the primary state.
865 assert(p->ps->psmargsz);
866 p->ps->psmarg[0] = '\0';
869 /*p->ps->pdfbytes = 0;*/
870 p->ps->psmargcur = 0;
871 p->ps->flags = PS_MARGINS;
872 p->ps->pscol = p->ps->left;
873 p->ps->psrow = p->ps->header;
875 ps_setfont(p, TERMFONT_NONE);
877 (*p->headf)(p, p->argf);
880 p->ps->pscol = p->ps->left;
881 p->ps->psrow = p->ps->footer;
883 (*p->footf)(p, p->argf);
886 p->ps->flags &= ~PS_MARGINS;
888 assert(0 == p->ps->flags);
889 assert(p->ps->psmarg);
890 assert('\0' != p->ps->psmarg[0]);
893 * Print header and initialise page state. Following this,
894 * stuff gets printed to the screen, so make sure we're sane.
897 if (TERMTYPE_PS == p->type) {
898 ps_printf(p, "%%!PS-Adobe-3.0\n");
899 ps_printf(p, "%%%%DocumentData: Clean7Bit\n");
900 ps_printf(p, "%%%%Orientation: Portrait\n");
901 ps_printf(p, "%%%%Pages: (atend)\n");
902 ps_printf(p, "%%%%PageOrder: Ascend\n");
903 ps_printf(p, "%%%%DocumentMedia: "
904 "Default %zu %zu 0 () ()\n",
905 (size_t)AFM2PNT(p, p->ps->width),
906 (size_t)AFM2PNT(p, p->ps->height));
907 ps_printf(p, "%%%%DocumentNeededResources: font");
909 for (i = 0; i < (int)TERMFONT__MAX; i++)
910 ps_printf(p, " %s", fonts[i].name);
912 ps_printf(p, "\n%%%%EndComments\n");
914 ps_printf(p, "%%PDF-1.1\n");
916 ps_printf(p, "<<\n");
917 ps_printf(p, ">>\n");
918 ps_printf(p, "endobj\n");
920 for (i = 0; i < (int)TERMFONT__MAX; i++) {
921 pdf_obj(p, (size_t)i + 3);
922 ps_printf(p, "<<\n");
923 ps_printf(p, "/Type /Font\n");
924 ps_printf(p, "/Subtype /Type1\n");
925 ps_printf(p, "/Name /F%d\n", i);
926 ps_printf(p, "/BaseFont /%s\n", fonts[i].name);
927 ps_printf(p, ">>\n");
931 p->ps->pdfbody = (size_t)TERMFONT__MAX + 3;
932 p->ps->pscol = p->ps->left;
933 p->ps->psrow = p->ps->top;
934 p->ps->flags |= PS_NEWPAGE;
935 ps_setfont(p, TERMFONT_NONE);
939 ps_pletter(struct termp *p, int c)
944 * If we haven't opened a page context, then output that we're
945 * in a new page and make sure the font is correctly set.
948 if (PS_NEWPAGE & p->ps->flags) {
949 if (TERMTYPE_PS == p->type) {
950 ps_printf(p, "%%%%Page: %zu %zu\n",
951 p->ps->pages + 1, p->ps->pages + 1);
952 ps_printf(p, "/%s %zu selectfont\n",
953 fonts[(int)p->ps->lastf].name,
956 pdf_obj(p, p->ps->pdfbody +
958 ps_printf(p, "<<\n");
959 ps_printf(p, "/Length %zu 0 R\n",
960 p->ps->pdfbody + 1 + p->ps->pages * 4);
961 ps_printf(p, ">>\nstream\n");
963 p->ps->pdflastpg = p->ps->pdfbytes;
964 p->ps->flags &= ~PS_NEWPAGE;
968 * If we're not in a PostScript "word" context, then open one
969 * now at the current cursor.
972 if ( ! (PS_INLINE & p->ps->flags)) {
973 if (TERMTYPE_PS != p->type) {
974 ps_printf(p, "BT\n/F%d %zu Tf\n",
975 (int)p->ps->lastf, p->ps->scale);
976 ps_printf(p, "%.3f %.3f Td\n(",
977 AFM2PNT(p, p->ps->pscol),
978 AFM2PNT(p, p->ps->psrow));
980 ps_printf(p, "%.3f %.3f moveto\n(",
981 AFM2PNT(p, p->ps->pscol),
982 AFM2PNT(p, p->ps->psrow));
983 p->ps->flags |= PS_INLINE;
986 assert( ! (PS_NEWPAGE & p->ps->flags));
989 * We need to escape these characters as per the PostScript
990 * specification. We would also escape non-graphable characters
991 * (like tabs), but none of them would get to this point and
992 * it's superfluous to abort() on them.
1001 ps_putchar(p, '\\');
1007 /* Write the character and adjust where we are on the page. */
1009 f = (int)p->ps->lastf;
1011 if (c <= 32 || c - 32 >= MAXCHAR)
1014 ps_putchar(p, (char)c);
1016 p->ps->pscol += (size_t)fonts[f].gly[c].wx;
1020 ps_pclose(struct termp *p)
1024 * Spit out that we're exiting a word context (this is a
1025 * "partial close" because we don't check the last-char buffer
1029 if ( ! (PS_INLINE & p->ps->flags))
1032 if (TERMTYPE_PS != p->type) {
1033 ps_printf(p, ") Tj\nET\n");
1035 ps_printf(p, ") show\n");
1037 p->ps->flags &= ~PS_INLINE;
1041 ps_fclose(struct termp *p)
1045 * Strong closure: if we have a last-char, spit it out after
1046 * checking that we're in the right font mode. This will of
1047 * course open a new scope, if applicable.
1049 * Following this, close out any scope that's open.
1052 if (p->ps->last != '\0') {
1053 assert( ! (p->ps->flags & PS_BACKSP));
1054 if (p->ps->nextf != p->ps->lastf) {
1056 ps_setfont(p, p->ps->nextf);
1058 p->ps->nextf = TERMFONT_NONE;
1059 ps_pletter(p, p->ps->last);
1063 if ( ! (PS_INLINE & p->ps->flags))
1070 ps_letter(struct termp *p, int arg)
1075 c = arg >= 128 || arg <= 0 ? '?' : arg;
1078 * When receiving a backspace, merely flag it.
1079 * We don't know yet whether it is
1080 * a font instruction or an overstrike.
1084 assert(p->ps->last != '\0');
1085 assert( ! (p->ps->flags & PS_BACKSP));
1086 p->ps->flags |= PS_BACKSP;
1091 * Decode font instructions.
1094 if (p->ps->flags & PS_BACKSP) {
1095 if (p->ps->last == '_') {
1096 switch (p->ps->nextf) {
1100 p->ps->nextf = TERMFONT_BI;
1103 p->ps->nextf = TERMFONT_UNDER;
1106 p->ps->flags &= ~PS_BACKSP;
1109 if (p->ps->last == c) {
1110 switch (p->ps->nextf) {
1113 case TERMFONT_UNDER:
1114 p->ps->nextf = TERMFONT_BI;
1117 p->ps->nextf = TERMFONT_BOLD;
1119 p->ps->flags &= ~PS_BACKSP;
1124 * This is not a font instruction, but rather
1125 * the next character. Prepare for overstrike.
1128 savecol = p->ps->pscol;
1133 * We found the next character, so the font instructions
1134 * for the previous one are complete.
1135 * Use them and print it.
1138 if (p->ps->last != '\0') {
1139 if (p->ps->nextf != p->ps->lastf) {
1141 ps_setfont(p, p->ps->nextf);
1143 p->ps->nextf = TERMFONT_NONE;
1144 ps_pletter(p, p->ps->last);
1148 * Do not print the current character yet because font
1149 * instructions might follow; only remember it.
1150 * For the first character, nothing else is done.
1151 * The final character will get printed from ps_fclose().
1157 * For an overstrike, back up to the previous position.
1160 if (savecol != SIZE_MAX) {
1162 p->ps->pscol = savecol;
1163 p->ps->flags &= ~PS_BACKSP;
1168 ps_advance(struct termp *p, size_t len)
1172 * Advance some spaces. This can probably be made smarter,
1173 * i.e., to have multiple space-separated words in the same
1174 * scope, but this is easier: just close out the current scope
1175 * and readjust our column settings.
1179 p->ps->pscol += len;
1183 ps_endline(struct termp *p)
1186 /* Close out any scopes we have open: we're at eoln. */
1191 * If we're in the margin, don't try to recalculate our current
1192 * row. XXX: if the column tries to be fancy with multiple
1193 * lines, we'll do nasty stuff.
1196 if (PS_MARGINS & p->ps->flags)
1201 p->ps->pscol = p->ps->left;
1203 /* If we haven't printed anything, return. */
1205 if (PS_NEWPAGE & p->ps->flags)
1209 * Put us down a line. If we're at the page bottom, spit out a
1210 * showpage and restart our row.
1213 if (p->ps->psrow >= p->ps->lineheight + p->ps->bottom) {
1214 p->ps->psrow -= p->ps->lineheight;
1222 ps_setfont(struct termp *p, enum termfont f)
1225 assert(f < TERMFONT__MAX);
1229 * If we're still at the top of the page, let the font-setting
1230 * be delayed until we actually have stuff to print.
1233 if (PS_NEWPAGE & p->ps->flags)
1236 if (TERMTYPE_PS == p->type)
1237 ps_printf(p, "/%s %zu selectfont\n",
1238 fonts[(int)f].name, p->ps->scale);
1240 ps_printf(p, "/F%d %zu Tf\n",
1241 (int)f, p->ps->scale);
1245 ps_width(const struct termp *p, int c)
1248 if (c <= 32 || c - 32 >= MAXCHAR)
1253 return((size_t)fonts[(int)TERMFONT_NONE].gly[c].wx);
1257 ps_hspan(const struct termp *p, const struct roffsu *su)
1262 * All of these measurements are derived by converting from the
1263 * native measurement to AFM units.
1268 * Traditionally, the default unit is fixed to the
1269 * output media. So this would refer to the point. In
1270 * mandoc(1), however, we stick to the default terminal
1271 * scaling unit so that output is the same regardless
1274 r = PNT2AFM(p, su->scale * 72.0 / 240.0);
1277 r = PNT2AFM(p, su->scale * 72.0 / 2.54);
1281 fonts[(int)TERMFONT_NONE].gly[109 - 32].wx;
1285 fonts[(int)TERMFONT_NONE].gly[110 - 32].wx;
1288 r = PNT2AFM(p, su->scale * 72.0);
1292 fonts[(int)TERMFONT_NONE].gly[109 - 32].wx / 100.0;
1295 r = PNT2AFM(p, su->scale * 12.0);
1298 r = PNT2AFM(p, su->scale * 1.0);
1301 r = su->scale * p->ps->lineheight;
1312 ps_growbuf(struct termp *p, size_t sz)
1314 if (p->ps->psmargcur + sz <= p->ps->psmargsz)
1317 if (sz < PS_BUFSLOP)
1320 p->ps->psmargsz += sz;
1321 p->ps->psmarg = mandoc_realloc(p->ps->psmarg, p->ps->psmargsz);