# # This file is licensed under the MIT No Attribution license. # Render groff 1.24.1 HTML image regions as selectable SVG objects. Apply to the unmodified 1.24.1 source with patch --fuzz=0 -p1. Requires Ghostscript, MuPDF 1.27.2's mutool, Perl, dvisvgm 3.6 with WOFF2, and svg-text.py installed as groff-svg-text in PATH. This changes only HTML output; ordinary PDF and PostScript output are unaffected. See README.adoc beside this patch. --- a/src/preproc/html/pre-html.cpp +++ b/src/preproc/html/pre-html.cpp @@ -156,14 +156,11 @@ #include "pushback.h" #include "html-strings.h" -#define DEFAULT_LINE_LENGTH 7 // inches wide +#define IMAGE_PAGE_LENGTH 1000 // inches; keep HTML image regions together #define DEFAULT_IMAGE_RES 100 // number of pixels per inch resolution -#define IMAGE_BORDER_PIXELS 0 +#define IMAGE_BORDER_PIXELS 2 #define INLINE_LEADER_CHAR '\\' -// Don't use colour names here! Otherwise there is a dependency on -// a file called 'rgb.txt' which maps names to colours. -#define TRANSPARENT "-background rgb:f/f/f -transparent rgb:f/f/f" #define MIN_ALPHA_BITS 0 #define MAX_ALPHA_BITS 4 @@ -187,21 +184,15 @@ // -1 means closed static const char *psFileName = 0 /* nullptr */; // PostScript // file name -static const char *psPageName = 0 /* nullptr */; // name of file - // containing - // current - // PostScript - // page static const char *regionFileName = 0 /* nullptr */; // name of file // containing // all image // regions static const char *imagePageName = 0 /* nullptr */; // name of - // bitmap image + // PDF region // file // containing - // current page -static const char *image_device = "pnmraw"; + // current region static int image_res = DEFAULT_IMAGE_RES; static int vertical_offset = 0; static char *image_template = 0 /* nullptr */; // image file name @@ -215,11 +206,8 @@ // directory static int textAlphaBits = MAX_ALPHA_BITS; static int graphicAlphaBits = MAX_ALPHA_BITS; -static const char *antiAlias = 0 /* nullptr */; // anti-alias arguments - // to be passed to gs static bool want_progress_report = false; // display page numbers // as they are processed -static int currentPageNo = -1; // current image page number static bool debugging = false; static const char *troffFileName = 0 /* nullptr */; // pre-html // output sent @@ -438,6 +426,8 @@ program_name, s); } close(saved_stdout); + if (status != 0) + fatal("image conversion failed"); } } @@ -613,25 +603,6 @@ } /* - * setupAntiAlias - Set up the antialias string, used when we call gs. - */ - -static void setupAntiAlias(void) -{ - if (textAlphaBits == 0 && graphicAlphaBits == 0) - antiAlias = make_string(" "); - else if (textAlphaBits == 0) - antiAlias = make_string("-dGraphicsAlphaBits=%d ", - graphicAlphaBits); - else if (graphicAlphaBits == 0) - antiAlias = make_string("-dTextAlphaBits=%d ", textAlphaBits); - else - antiAlias = make_string("-dTextAlphaBits=%d" - " -dGraphicsAlphaBits=%d ", textAlphaBits, - graphicAlphaBits); -} - -/* * checkImageDir - Check whether the image directory is available. */ @@ -692,7 +663,7 @@ break; } writeString(image_template); - writeString(".png]"); + writeString(".svg]"); if (is_html) writeString("\\O[0]\\O[3]"); else @@ -825,6 +796,15 @@ // with no translation, for image file creation in the // post-processor. + // The PostScript pass is only an intermediate layout for HTML images. + // Its page length must match grops below; paper boundaries must not + // split a table or equation into incomplete image regions. + if (device_format_selector == DEVICE_FORMAT(IMAGE_OUTPUT_FILTER)) { + const char *s = make_string(".pl %di\n", IMAGE_PAGE_LENGTH); + writeString(s); + free(const_cast(s)); + } + int idx = 0; char_block *element = head; @@ -907,9 +887,7 @@ void add(int x1, int y1, int x2, int y2, int page, int res, int maxx, char *name); void createImages(void); - int createPage(int pageno); void createImage(imageItem *i); - int getMaxX(int pageno); }; /* @@ -935,58 +913,6 @@ } /* - * createPage - Create image of page `pageno` from PostScript file. - */ - -int imageList::createPage(int pageno) -{ - if (currentPageNo == pageno) - return 0; - - if (currentPageNo >= 1) { - /* - * We need to unlink the files which change each time a new page is - * processed. The final unlink is done by xtmpfile when - * pre-grohtml exits. - */ - unlink(imagePageName); - unlink(psPageName); - } - - if (want_progress_report) { - fprintf(stderr, "[%d] ", pageno); - fflush(stderr); - } - - if (debugging) - fprintf(stderr, "%s: debug: creating page %d\n", program_name, - pageno); - - const char *s = make_string("ps2ps -sPageList=%d %s %s", - pageno, psFileName, psPageName); - html_system(s, 1); - assert(strlen(image_gen) > 0); - s = make_string("echo showpage | " - "%s%s -q -dBATCH -dSAFER " - "-dDEVICEHEIGHTPOINTS=792 " - "-dDEVICEWIDTHPOINTS=%d -dFIXEDMEDIA=true " - "-sDEVICE=%s -r%d %s " - "-sOutputFile=%s %s -", - image_gen, - EXE_EXT, - (getMaxX(pageno) * image_res) / postscriptRes, - image_device, - image_res, - antiAlias, - imagePageName, - psPageName); - html_system(s, 1); - free(const_cast(s)); - currentPageNo = pageno; - return 0; -} - -/* * min - Return the minimum of two numbers. */ @@ -1011,63 +937,67 @@ } /* - * getMaxX - Return the largest right-hand position for any image - * on `pageno`. - */ - -int imageList::getMaxX(int pageno) -{ - imageItem *h = head; - int x = postscriptRes * DEFAULT_LINE_LENGTH; - - while (h != 0 /* nullptr */) { - if (h->pageNo == pageno) - x = max(h->X2, x); - h = h->next; - } - return x; -} - -/* - * createImage - Generate a minimal PNG file from the set of page - * images. + * createImage - Extract one vector region, preserving fonts and lines. + * + * Render directly from the original PostScript: ps2ps clips to its + * paper size before the old raster crop can reach wide image regions. + * A small PDF canvas and translated origin retain only this region; + * no bitmap of the long intermediate page is allocated. dvisvgm embeds + * fonts in SVG, keeping the text selectable without external font files. */ void imageList::createImage(imageItem *i) { if (i->X1 != -1) { - int x1 = max(min(i->X1, i->X2) * image_res / postscriptRes + int x1 = max(int(min(i->X1, i->X2) * double(image_res) / postscriptRes) - IMAGE_BORDER_PIXELS, 0); int y1 = max(image_res * vertical_offset / 72 - + min(i->Y1, i->Y2) * image_res / postscriptRes + + int(min(i->Y1, i->Y2) * double(image_res) / postscriptRes) - IMAGE_BORDER_PIXELS, 0); - int x2 = max(i->X1, i->X2) * image_res / postscriptRes + int x2 = int(max(i->X1, i->X2) * double(image_res) / postscriptRes) + IMAGE_BORDER_PIXELS; int y2 = image_res * vertical_offset / 72 - + max(i->Y1, i->Y2) * image_res / postscriptRes + + int(max(i->Y1, i->Y2) * double(image_res) / postscriptRes) + 1 + IMAGE_BORDER_PIXELS; - if (createPage(i->pageNo) == 0) { - const char *s = make_string("pamcut%s %d %d %d %d < %s " - "| pnmcrop%s " PNMTOOLS_QUIET - "| pnmtopng%s " PNMTOOLS_QUIET " %s" - "> %s", - EXE_EXT, - x1, y1, x2 - x1 + 1, y2 - y1 + 1, - imagePageName, - EXE_EXT, - EXE_EXT, - TRANSPARENT, - i->imageName); - html_system(s, 0); - free(const_cast(s)); - } - else { - fprintf(stderr, "%s: failed to generate image of page %d\n", - program_name, i->pageNo); - fflush(stderr); - } + // Match the former PNG display size while retaining vector paths. + // -r72 avoids PDF interpreter coordinate limits on the virtual page. + assert(strlen(image_gen) > 0); + const char *s = make_string( + "%s%s -q -dBATCH -dNOPAUSE -dSAFER " + "-dFirstPage=%d -dLastPage=%d -dFIXEDMEDIA=true -r72 " + "-dDEVICEWIDTHPOINTS=%d -dDEVICEHEIGHTPOINTS=%d " + "-sDEVICE=pdfwrite -sOutputFile=%s " + "-c '<> setpagedevice' -f %s", + image_gen, EXE_EXT, i->pageNo, i->pageNo, + x2 - x1 + 1, y2 - y1 + 1, imagePageName, -x1, + (y2 + 1) - IMAGE_PAGE_LENGTH * double(image_res), + image_res / 72.0, psFileName); + html_system(s, 1); + free(const_cast(s)); + // Convert PDF points to CSS pixels without changing the viewBox. + // Reading stdin also accepts the extensionless temporary PDF name. + s = make_string("DVISVGM_PDF_PROC=mutool dvisvgm --pdf --stdin " + "--font-format=woff2 --zoom=0.75 --verbosity=3 " + "--output=%s < %s", i->imageName, imagePageName); + html_system(s, 1); + free(const_cast(s)); + // Keep the order of embedded font declarations reproducible. + // Preserve spaces and fractional character widths in the browser. + s = make_string("perl -0777 -i -pe " + "'s~(?:\\@font-face\\{[^}]*\\}\\n?)+~" + "join(\"\\n\", sort split /\\n/, $&) . \"\\n\"~ge;" + "s~(]*>\\n)~$1text { white-space: pre; " + "text-rendering: geometricPrecision; }\\n~' %s", + i->imageName); + html_system(s, 1); + free(const_cast(s)); + s = make_string("groff-svg-text %s %s", + imagePageName, i->imageName); + html_system(s, 1); + free(const_cast(s)); #if defined(DEBUGGING) } else { @@ -1536,9 +1466,10 @@ argv = addRegDef(argc, argv, s.contents()); argc++; - // Override local settings and produce a letter-size PostScript page - // file. - argv = addRegDef(argc, argv, "-P-pletter"); + // Match the long virtual page used by the image-layout pass. + const char *paper = make_string("-P-p%di,8.5i", IMAGE_PAGE_LENGTH); + argv = addRegDef(argc, argv, paper); + free(const_cast(paper)); argc++; if (dialect == xhtml) { @@ -1752,18 +1683,11 @@ psFileName = DEBUG_FILE("prehtml-ps"); regionFileName = DEBUG_FILE("prehtml-region"); imagePageName = DEBUG_FILE("prehtml-page"); - psPageName = DEBUG_FILE("prehtml-psn"); troffFileName = DEBUG_FILE("prehtml-troff"); htmlFileName = DEBUG_FILE("prehtml-html"); FILE *f; - // psPageName contains a single page of PostScript. - f = xtmpfile(&psPageName, PS_TEMPLATE_LONG, PS_TEMPLATE_SHORT); - if (0 /* nullptr */ == f) - sys_fatal("xtmpfile"); - fclose(f); - - // imagePageName contains a bitmap image of a single PostScript page. + // imagePageName contains a single cropped PDF image region. f = xtmpfile(&imagePageName, PAGE_TEMPLATE_LONG, PAGE_TEMPLATE_SHORT); if (0 /* nullptr */ == f) sys_fatal("xtmpfile"); @@ -1839,7 +1763,6 @@ if (postscriptRes < 1) // TODO: what's a more sane minimum value? fatal("'res' directive missing or invalid in file '%1'", devps_desc); - setupAntiAlias(); checkImageDir(); makeFileName(); bool have_file_operand = false; --- a/src/devices/grohtml/post-html.cpp +++ b/src/devices/grohtml/post-html.cpp @@ -2606,10 +2606,36 @@ while ((filename != 0 /* nullptr */) && (' ' == filename[0])) filename++; if (exists(filename)) { - *s += string(""; - if (xhtml == dialect) - *s += ""; + const char *suffix = strrchr(filename, '.'); + if (suffix && strcmp(suffix, ".svg") == 0) { + string size; + // Reserve the space before the object loads to keep layout stable. + // The generated SVG viewBox is measured in CSS pixels. + FILE *image = fopen(filename, "r"); + if (image) { + char line[1024]; + while (fgets(line, sizeof line, image)) { + const char *box = strstr(line, "viewBox="); + if (box) { + int width, height; + if (sscanf(box + 9, "%*d %*d %d %d", &width, &height) == 2 + && width > 0 && height > 0) + size = string(" width=\"") + as_string(width) + + "\" height=\"" + as_string(height) + "\""; + break; + } + } + fclose(image); + } + *s += string("Image " + filename + ""; + } + else { + *s += string(""; + if (xhtml == dialect) + *s += ""; + } } return *s; } --- a/src/roff/troff/mtsm.cpp +++ b/src/roff/troff/mtsm.cpp @@ -397,3 +397,5 @@ void mtsm::inherit(statem *s, int reset_bool) if (!stack.empty()) { - statem top = stack.top(); + // Consume inherited breaks in the saved state, so they are not + // repeated for every glyph in a diversion. + statem &top = stack.top(); if (s->units_values[MTSM_IN].is_known --- a/tmac/www.tmac.in +++ b/tmac/www.tmac.in @@ -1266,7 +1266,7 @@ . ie '\*[.T]'html' \{\ . www-make-unique-name . nr www-drop-width (100u * \\n[.v]u * 3u / \\n[.l]u) -. MPIMG -L \\*[www-unique-name].png \\n[www-drop-width]% +. MPIMG -L \\*[www-unique-name].svg \\n[www-drop-width]% . \} . el \{\ . ie r ps4html \{\ @@ -1278,7 +1278,7 @@ . bp . ev www-DC . vs 320p -. nop \O[5i\\*[www-unique-name].png]\O[1] +. nop \O[5i\\*[www-unique-name].svg]\O[1] . nop \m[\\*[www-dropcolor]]\s[160]\O[3]\\$1\O[4] . nop \O[2]\O[0] . br @@ -1304,7 +1304,7 @@ .\" supplementary macros used by other macro sets .\" .\" here are some tags specially for -Tps or -Thtml when invoked by -.\" pre-html to generate PNG images from postscript. +.\" pre-html to generate SVG images from PostScript. . .\" -------------------------------------------------------------------- .\" HTML-DO-IMAGE - tell troff to issue an image marker which can be @@ -1312,9 +1312,9 @@ .\" .de HTML-DO-IMAGE . if r ps4html \ -. nop \O[5\\$2\\$1.png]\O[1]\O[3] +. nop \O[5\\$2\\$1.svg]\O[1]\O[3] . if \\n[www-html] \ -. nop \O[5\\$2\\$1.png]\O[0]\O[3] +. nop \O[5\\$2\\$1.svg]\O[0]\O[3] .. . .\" -------------------------------------------------------------------- @@ -1396,9 +1396,9 @@ . ie r xhtml \{\ . if !(\\n[.O] == 0) \{\ . if r ps4html \ -. nop \O[5\\$2\\$1.png]\O[1]\O[3] +. nop \O[5\\$2\\$1.svg]\O[1]\O[3] . if \\n[www-html] \ -. nop \O[5\\$2\\$1.png]\O[0]\O[3] +. nop \O[5\\$2\\$1.svg]\O[0]\O[3] . \} . \} . el \ --- a/tmac/an.tmac +++ b/tmac/an.tmac @@ -1399,7 +1399,6 @@ . el .nr 3usekeeps 1 . if \\n[an*is-output-html] \{\ . nr an-TS-ll \\n[.l] -. ll 1000n . \} . HTML-IMAGE .. @@ -1428,7 +1427,6 @@ .de1 EQ . if \\n[an*is-output-html] \{\ . nr an-EQ-ll \\n[.l] -. ll 1000n . \} . HTML-IMAGE .. --- a/tmac/s.tmac +++ b/tmac/s.tmac @@ -1788,6 +1788,9 @@ .\" **************************** .\" Table of contents generation. .de @XS +.\" Close the HTML heading before diverting a printed contents entry. +.if \\n[s@devtag-needs-end-of-heading] .DEVTAG-EO-H +.nr s@devtag-needs-end-of-heading 0 .da toc*div .ev h .ie \\n[.$] .XA "\\$1" @@ -1936,6 +1939,8 @@ . \} . \} . rm eqn*msg +. if '\*(.T'html' .nr eqn*type 3 +. if r ps4html .nr eqn*type 3 .\} .di eqn*div .in 0