X-Git-Url: https://git.distorted.org.uk/~mdw/sgt/halibut/blobdiff_plain/f0e51ce1f8e8c03966509153e9ff360684d4dd6f..27bdc5ab5b1f2a90b06fc924346f014e1b272c65:/bk_pdf.c

diff --git a/bk_pdf.c b/bk_pdf.c
index fd1424a..6aba554 100644
--- a/bk_pdf.c
+++ b/bk_pdf.c
@@ -10,30 +10,7 @@
 
 paragraph *pdf_config_filename(char *filename)
 {
-    paragraph *p;
-    wchar_t *ufilename, *up;
-    int len;
-
-    p = mknew(paragraph);
-    memset(p, 0, sizeof(*p));
-    p->type = para_Config;
-    p->next = NULL;
-    p->fpos.filename = "<command line>";
-    p->fpos.line = p->fpos.col = -1;
-
-    ufilename = ufroma_dup(filename);
-    len = ustrlen(ufilename) + 2 + lenof(L"pdf-filename");
-    p->keyword = mknewa(wchar_t, len);
-    up = p->keyword;
-    ustrcpy(up, L"pdf-filename");
-    up = uadv(up);
-    ustrcpy(up, ufilename);
-    up = uadv(up);
-    *up = L'\0';
-    assert(up - p->keyword < len);
-    sfree(ufilename);
-
-    return p;
+    return cmdline_cfg_simple("pdf-filename", filename, NULL);
 }
 
 typedef struct object_Tag object;
@@ -56,12 +33,17 @@ struct objlist_Tag {
 static object *new_object(objlist *list);
 static void objtext(object *o, char const *text);
 static void objstream(object *o, char const *text);
+static void pdf_string(void (*add)(object *, char const *),
+		       object *, char const *);
+static void pdf_string_len(void (*add)(object *, char const *),
+			   object *, char const *, int);
 static void objref(object *o, object *dest);
 
 static void make_pages_node(object *node, object *parent, page_data *first,
 			    page_data *last, object *resources);
 static int make_outline(object *parent, outline_element *start, int n,
 			int open);
+static int pdf_versionid(FILE *fp, word *words);
 
 void pdf_backend(paragraph *sourceform, keywordlist *keywords,
 		 indexdata *idx, void *vdoc) {
@@ -85,7 +67,7 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
 	if (p->type == para_Config && p->parent) {
 	    if (!ustricmp(p->keyword, L"pdf-filename")) {
 		sfree(filename);
-		filename = utoa_dup(uadv(p->keyword));
+		filename = dupstr(adv(p->origkeyword));
 	    }
 	}
     }
@@ -94,7 +76,10 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
     olist.number = 1;
 
     cat = new_object(&olist);
-    outlines = new_object(&olist);
+    if (doc->n_outline_elements > 0)
+	outlines = new_object(&olist);
+    else
+	outlines = NULL;
     pages = new_object(&olist);
     resources = new_object(&olist);
 
@@ -102,11 +87,16 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
      * The catalogue just contains references to the outlines and
      * pages objects.
      */
-    objtext(cat, "<<\n/Type /Catalog\n/Outlines ");
-    objref(cat, outlines);
+    objtext(cat, "<<\n/Type /Catalog");
+    if (outlines) {
+	objtext(cat, "\n/Outlines ");
+	objref(cat, outlines);
+    }
     objtext(cat, "\n/Pages ");
     objref(cat, pages);
-    objtext(cat, "\n/PageMode /UseOutlines\n>>\n");
+    if (outlines)
+	objtext(cat, "\n/PageMode /UseOutlines");
+    objtext(cat, "\n>>\n");
 
     /*
      * Set up the resources dictionary, which mostly means
@@ -194,8 +184,9 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
     for (page = doc->pages; page; page = page->next) {
 	object *opage, *cstr;
 	rect *r;
-	text_fragment *frag;
+	text_fragment *frag, *frag_end;
 	char buf[256];
+	int x, y, lx, ly;
 
 	opage = (object *)page->spare;
 	/*
@@ -235,24 +226,89 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
 	}
 
 	objstream(cstr, "BT\n");
-	for (frag = page->first_text; frag; frag = frag->next) {
-	    char *c;
 
+	/*
+	 * PDF tracks two separate current positions: the position
+	 * given in the `line matrix' and the position given in the
+	 * `text matrix'. We must therefore track both as well.
+	 * They start off at -1 (unset).
+	 */
+	lx = ly = -1;
+	x = y = -1;
+
+	frag = page->first_text;
+	while (frag) {
+	    /*
+	     * For compactness, I'm going to group text fragments
+	     * into subsequences that use the same font+size. So
+	     * first find the end of this subsequence.
+	     */
+	    for (frag_end = frag;
+		 (frag_end &&
+		  frag_end->fe == frag->fe &&
+		  frag_end->fontsize == frag->fontsize);
+		 frag_end = frag_end->next);
+
+	    /*
+	     * Now select the text fragment, and prepare to display
+	     * the text.
+	     */
 	    objstream(cstr, "/");
 	    objstream(cstr, frag->fe->name);
-	    sprintf(buf, " %d Tf 1 0 0 1 %g %g Tm (", frag->fontsize,
-		    frag->x/4096.0, frag->y/4096.0);
+	    sprintf(buf, " %d Tf ", frag->fontsize);
 	    objstream(cstr, buf);
 
-	    for (c = frag->text; *c; c++) {
-		if (*c == '(' || *c == ')' || *c == '\\')
-		    objstream(cstr, "\\");
-		buf[0] = *c;
-		buf[1] = '\0';
+	    while (frag && frag != frag_end) {
+		/*
+		 * Place the text position for the first piece of
+		 * text.
+		 */
+		if (lx < 0) {
+		    sprintf(buf, "1 0 0 1 %g %g Tm ",
+			    frag->x/4096.0, frag->y/4096.0);
+		} else {
+		    sprintf(buf, "%g %g Td ",
+			    (frag->x - lx)/4096.0, (frag->y - ly)/4096.0);
+		}
 		objstream(cstr, buf);
+		lx = x = frag->x;
+		ly = y = frag->y;
+
+		/*
+		 * See if we're going to use Tj (show a single
+		 * string) or TJ (show an array of strings with
+		 * x-spacings between them). We determine this by
+		 * seeing if there's more than one text fragment in
+		 * sequence with the same y-coordinate.
+		 */
+		if (frag->next && frag->next != frag_end &&
+		    frag->next->y == y) {
+		    /*
+		     * The TJ strategy.
+		     */
+		    objstream(cstr, "[");
+		    while (frag && frag != frag_end && frag->y == y) {
+			if (frag->x != x) {
+			    sprintf(buf, "%g",
+				    (x - frag->x) * 1000.0 /
+				    (4096.0 * frag->fontsize));
+			    objstream(cstr, buf);
+			}
+			pdf_string(objstream, cstr, frag->text);
+			x = frag->x + frag->width;
+			frag = frag->next;
+		    }
+		    objstream(cstr, "]TJ\n");
+		} else
+		{
+		    /*
+		     * The Tj strategy.
+		     */
+		    pdf_string(objstream, cstr, frag->text);
+		    objstream(cstr, "Tj\n");
+		    frag = frag->next;
+		}
 	    }
-
-	    objstream(cstr, ") Tj\n");
 	}
 	objstream(cstr, "ET");
 
@@ -284,18 +340,9 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
 		    objref(annot, (object *)xr->dest.page->spare);
 		    objtext(annot, " /XYZ null null null]\n");
 		} else {
-		    char *p;
-
-		    objtext(annot, "/A <<\n/Type /Action\n/S /URI\n/URI (");
-		    for (p = xr->dest.url; *p; p++) {
-			char c[2];
-			c[0] = *p;
-			c[1] = '\0';
-			if (*p == '(' || *p == ')' || *p == '\\')
-			    objtext(annot, "\\");
-			objtext(annot, c);
-		    }
-		    objtext(annot, ")\n>>\n");
+		    objtext(annot, "/A <<\n/Type /Action\n/S /URI\n/URI ");
+		    pdf_string(objtext, annot, xr->dest.url);
+		    objtext(annot, "\n>>\n");
 		}
 
 		objtext(annot, ">>\n");
@@ -310,7 +357,7 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
     /*
      * Set up the outlines dictionary.
      */
-    {
+    if (outlines) {
 	int topcount;
 	char buf[80];
 
@@ -372,9 +419,13 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
     }
 
     /*
-     * Header
+     * Header. I'm going to put the version IDs in the header as
+     * well, simply in PDF comments.
      */
     fileoff = fprintf(fp, "%%PDF-1.3\n");
+    for (p = sourceform; p; p = p->next)
+	if (p->type == para_VersionID)
+	    fileoff += pdf_versionid(fp, p->words);
 
     /*
      * Body
@@ -413,7 +464,7 @@ void pdf_backend(paragraph *sourceform, keywordlist *keywords,
 
 static object *new_object(objlist *list)
 {
-    object *obj = mknew(object);
+    object *obj = snew(object);
 
     obj->list = list;
 
@@ -539,101 +590,32 @@ static void make_pages_node(object *node, object *parent, page_data *first,
  * means that glyphs are identified by PS strings and hence font
  * encoding can be managed independently of the supplied encoding
  * of the font. However, in the document outline, the PDF spec
- * simply asks for ordinary text strings without mentioning what
- * character set they are supposed to be interpreted in.
- * 
- * Therefore, for the moment, I'm going to assume they're US-ASCII
- * only. If anyone knows better, they should let me know :-/
+ * encodes in either PDFDocEncoding (a custom superset of
+ * ISO-8859-1) or UTF-16BE.
  */
-static int pdf_convert(wchar_t *s, char **result) {
-    int doing = (result != 0);
-    int ok = TRUE;
-    char *p = NULL;
-    int plen = 0, psize = 0;
-
-    for (; *s; s++) {
-	wchar_t c = *s;
-	char outc;
-
-	if (c >= 32 && c <= 126) {
-	    /* Char is OK. */
-	    outc = (char)c;
-	} else {
-	    /* Char is not OK. */
-	    ok = FALSE;
-	    outc = 0xBF;	       /* approximate the good old DEC `uh?' */
-	}
-	if (doing) {
-	    if (plen >= psize) {
-		psize = plen + 256;
-		p = resize(p, psize);
-	    }
-	    p[plen++] = outc;
-	}
-    }
-    if (doing) {
-	p = resize(p, plen+1);
-	p[plen] = '\0';
-	*result = p;
+static char *pdf_outline_convert(wchar_t *s, int *len) {
+    char *ret;
+
+    ret = utoa_careful_dup(s, CS_PDF);
+
+    /*
+     * Very silly special case: if the returned string begins with
+     * FE FF, then the PDF reader will mistake it for a UTF-16BE
+     * string. So in this case we give up on PDFDocEncoding and
+     * encode it in UTF-16 straight away.
+     */
+    if (ret && ret[0] == '\xFE' && ret[1] == '\xFF') {
+	sfree(ret);
+	ret = NULL;
     }
-    return ok;
-}
 
-static void pdf_rdaddwc(rdstringc *rs, word *text) {
-    char *c;
-
-    for (; text; text = text->next) switch (text->type) {
-      case word_HyperLink:
-      case word_HyperEnd:
-      case word_UpperXref:
-      case word_LowerXref:
-      case word_XrefEnd:
-      case word_IndexRef:
-	break;
-
-      case word_Normal:
-      case word_Emph:
-      case word_Code:
-      case word_WeakCode:
-      case word_WhiteSpace:
-      case word_EmphSpace:
-      case word_CodeSpace:
-      case word_WkCodeSpace:
-      case word_Quote:
-      case word_EmphQuote:
-      case word_CodeQuote:
-      case word_WkCodeQuote:
-	assert(text->type != word_CodeQuote &&
-	       text->type != word_WkCodeQuote);
-	if (towordstyle(text->type) == word_Emph &&
-	    (attraux(text->aux) == attr_First ||
-	     attraux(text->aux) == attr_Only))
-	    rdaddc(rs, '_');	       /* FIXME: configurability */
-	else if (towordstyle(text->type) == word_Code &&
-		 (attraux(text->aux) == attr_First ||
-		  attraux(text->aux) == attr_Only))
-	    rdaddc(rs, '\'');	       /* FIXME: configurability */
-	if (removeattr(text->type) == word_Normal) {
-	    if (pdf_convert(text->text, &c))
-		rdaddsc(rs, c);
-	    else
-		pdf_rdaddwc(rs, text->alt);
-	    sfree(c);
-	} else if (removeattr(text->type) == word_WhiteSpace) {
-	    rdaddc(rs, ' ');
-	} else if (removeattr(text->type) == word_Quote) {
-	    rdaddc(rs, '\''); /* FIXME: configurability */
-	}
-	if (towordstyle(text->type) == word_Emph &&
-	    (attraux(text->aux) == attr_Last ||
-	     attraux(text->aux) == attr_Only))
-	    rdaddc(rs, '_');	       /* FIXME: configurability */
-	else if (towordstyle(text->type) == word_Code &&
-		 (attraux(text->aux) == attr_Last ||
-		  attraux(text->aux) == attr_Only))
-	    rdaddc(rs, '\'');	       /* FIXME: configurability */
-	break;
+    if (!ret) {
+	ret = utoa_dup_len(s, CS_UTF16BE, len);
+    } else {
+	*len = strlen(ret);
     }
+
+    return ret;
 }
 
 static int make_outline(object *parent, outline_element *items, int n,
@@ -642,15 +624,14 @@ static int make_outline(object *parent, outline_element *items, int n,
     int level, totalcount = 0;
     outline_element *itemp;
     object *curr, *prev = NULL, *first = NULL, *last = NULL;
-    para_data *pdata;
 
     assert(n > 0);
 
     level = items->level;
 
     while (n > 0) {
-	rdstringc rs = {0, 0, NULL};
-	char *p;
+	char *title;
+	int titlelen;
 
 	/*
 	 * Here we expect to be sitting on an item at the given
@@ -659,33 +640,18 @@ static int make_outline(object *parent, outline_element *items, int n,
 	 */
 	assert(items->level == level);
 
-	if (level == 1 && items->para->kwtext) {
-	    pdf_rdaddwc(&rs, items->para->kwtext);
-	    rdaddsc(&rs, ": ");
-	} else if (level > 1 && items->para->kwtext2) {
-	    pdf_rdaddwc(&rs, items->para->kwtext2);
-	    rdaddsc(&rs, " ");
-	}
-	pdf_rdaddwc(&rs, items->para->words);
+	title = pdf_outline_convert(items->pdata->outline_title, &titlelen);
 
 	totalcount++;
 	curr = new_object(parent->list);
 	if (!first) first = curr;
 	last = curr;
-	objtext(curr, "<<\n/Title (");
-	for (p = rs.text; p < rs.text+rs.pos; p++) {
-	    char c[2];
-	    if (*p == '\\' || *p == '(' || *p == ')')
-		objtext(curr, "\\");
-	    c[0] = *p;
-	    c[1] = '\0';
-	    objtext(curr, c);
-	}
-	objtext(curr, ")\n/Parent ");
+	objtext(curr, "<<\n/Title ");
+	pdf_string_len(objtext, curr, title, titlelen);
+	objtext(curr, "\n/Parent ");
 	objref(curr, parent);
 	objtext(curr, "\n/Dest [");
-	pdata = (para_data *)items->para->private_data;
-	objref(curr, (object *)pdata->first->page->spare);
+	objref(curr, (object *)items->pdata->first->page->spare);
 	objtext(curr, " /XYZ null null null]\n");
 	if (prev) {
 	    objtext(curr, "/Prev ");
@@ -727,3 +693,75 @@ static int make_outline(object *parent, outline_element *items, int n,
 
     return totalcount;
 }
+
+static int pdf_versionid(FILE *fp, word *words)
+{
+    int ret;
+
+    ret = fprintf(fp, "%% ");
+
+    for (; words; words = words->next) {
+	char *text;
+	int type;
+
+	switch (words->type) {
+	  case word_HyperLink:
+	  case word_HyperEnd:
+	  case word_UpperXref:
+	  case word_LowerXref:
+	  case word_XrefEnd:
+	  case word_IndexRef:
+	    continue;
+	}
+
+	type = removeattr(words->type);
+
+	switch (type) {
+	  case word_Normal:
+	    text = utoa_dup(words->text, CS_ASCII);
+	    break;
+	  case word_WhiteSpace:
+	    text = dupstr(" ");
+	    break;
+	  case word_Quote:
+	    text = dupstr("'");
+	    break;
+	}
+
+	fputs(text, fp);
+	ret += strlen(text);
+	sfree(text);
+    }
+
+    ret += fprintf(fp, "\n");
+
+    return ret;
+}
+
+static void pdf_string_len(void (*add)(object *, char const *),
+			   object *o, char const *str, int len)
+{
+    char const *p;
+
+    add(o, "(");
+    for (p = str; len > 0; p++, len--) {
+	char c[10];
+	if (*p < ' ' || *p > '~') {
+	    sprintf(c, "\\%03o", 0xFF & (int)*p);
+	} else {
+	    int n = 0;
+	    if (*p == '\\' || *p == '(' || *p == ')')
+		c[n++] = '\\';
+	    c[n++] = *p;
+	    c[n] = '\0';
+	}
+	add(o, c);
+    }
+    add(o, ")");
+}
+
+static void pdf_string(void (*add)(object *, char const *),
+		       object *o, char const *str)
+{
+    pdf_string_len(add, o, str, strlen(str));
+}