text clip mode. split text content streams. text extraction api.

2 files changed, 33 insertions, 115 deletions

diff --git a/test/pdfrip.c b/test/pdfrip.c
index 0837e969..5d4aba39 100644
--- a/test/pdfrip.c
+++ b/test/pdfrip.c
@@ -2,110 +2,14 @@
 #include <mupdf.h>
 
 int showtree = 0;
-int showtext = 0;
 float zoom = 1.0;
 
 void usage()
 {
-	fprintf(stderr, "usage: pdfrip [-dt] [-p password] [-z zoom] file.pdf [pages...]\n");
+	fprintf(stderr, "usage: pdfrip [-d] [-p password] [-z zoom] file.pdf [pages...]\n");
 	exit(1);
 }
 
-enum
-{
-        Bit1    = 7,
-        Bitx    = 6,
-        Bit2    = 5,
-        Bit3    = 4,
-        Bit4    = 3,
-
-        T1      = ((1<<(Bit1+1))-1) ^ 0xFF,     /* 0000 0000 */
-        Tx      = ((1<<(Bitx+1))-1) ^ 0xFF,     /* 1000 0000 */
-        T2      = ((1<<(Bit2+1))-1) ^ 0xFF,     /* 1100 0000 */
-        T3      = ((1<<(Bit3+1))-1) ^ 0xFF,     /* 1110 0000 */
-        T4      = ((1<<(Bit4+1))-1) ^ 0xFF,     /* 1111 0000 */
-
-        Rune1   = (1<<(Bit1+0*Bitx))-1,         /* 0000 0000 0111 1111 */
-        Rune2   = (1<<(Bit2+1*Bitx))-1,         /* 0000 0111 1111 1111 */
-        Rune3   = (1<<(Bit3+2*Bitx))-1,         /* 1111 1111 1111 1111 */
-
-        Maskx   = (1<<Bitx)-1,                  /* 0011 1111 */
-        Testx   = Maskx ^ 0xFF,                 /* 1100 0000 */
-};
-
-void putrune(int c)
-{
-	if (c <= Rune1)
-	{
-		putchar(c);
-		return;
-	}
-
-	if (c <= Rune2)
-	{
-		putchar(T2 | (c >> 1*Bitx));
-		putchar(Tx | (c & Maskx));
-		return;
-	}
-		
-	putchar(T3 | (c >> 2*Bitx));
-	putchar(Tx | ((c >> 1*Bitx) & Maskx));
-	putchar(Tx | (c & Maskx));
-}
-
-/*
- * Dump text nodes as unicode
- */
-void dumptext(fz_node *node)
-{
-	int i, cid, ucs;
-	static fz_point old = { 0, 0 };
-	fz_point p;
-	float dx, dy;
-	fz_vmtx v;
-	fz_hmtx h;
-
-	if (fz_istextnode(node))
-	{
-		fz_textnode *text = (fz_textnode*)node;
-		pdf_font *font = (pdf_font*)text->font;
-		fz_matrix invtrm = fz_invertmatrix(text->trm);
-
-		for (i = 0; i < text->len; i++)
-		{
-			cid = text->els[i].cid;
-			p.x = text->els[i].x;
-			p.y = text->els[i].y;
-			p = fz_transformpoint(invtrm, p);
-			dx = old.x - p.x;
-			dy = old.y - p.y;
-			old = p;
-
-			if (fabs(dy) > 1.6)
-				puts("\n");
-			else if (fabs(dy) > 0.2)
-				putchar('\n');
-			else if (fabs(dx) > 0.2)
-				putchar(' ');
-
-			h = fz_gethmtx(text->font, cid);
-			old.x += h.w / 1000.0;
-
-			if (font->tounicode)
-				ucs = fz_lookupcid(font->tounicode, cid);
-			else if (font->ncidtoucs)
-				ucs = font->cidtoucs[cid];
-			else
-				ucs = cid;
-
-			putrune(ucs);
-		}
-	}
-
-	for (node = node->first; node; node = node->next)
-		dumptext(node);
-}
-
 /*
  * Draw page
  */
@@ -139,14 +43,6 @@ void showpage(pdf_xref *xref, fz_obj *pageobj)
 		printf("endtree\n");
 	}
 
-	if (showtext)
-	{
-		printf("---begin text dump---\n");
-		dumptext(page->tree->root);
-		printf("\n---end text dump---\n");
-	}
-
-	else
 	{
 		fz_pixmap *pix;
 		fz_renderer *gc;
@@ -191,14 +87,13 @@ int main(int argc, char **argv)
 
 	char *password = "";
 
-	while ((c = getopt(argc, argv, "dtz:p:")) != -1)
+	while ((c = getopt(argc, argv, "dz:p:")) != -1)
 	{
 		switch (c)
 		{
 		case 'p': password = optarg; break;
 		case 'z': zoom = atof(optarg); break;
 		case 'd': ++showtree; break;
-		case 't': ++showtext; break;
 		default: usage();
 		}
 	}
@@ -226,8 +121,8 @@ int main(int argc, char **argv)
 	if (error) fz_abort(error);
 
 	outlines = nil;
-	error = pdf_loadoutlinetree(&outlines, xref);
-	if (error) { fz_warn(error->msg); fz_droperror(error); }
+//	error = pdf_loadoutlinetree(&outlines, xref);
+//	if (error) { fz_warn(error->msg); fz_droperror(error); }
 
 	if (optind == argc)
 	{
diff --git a/test/x11pdf.c b/test/x11pdf.c
index 58655f8a..3d424b1c 100644
--- a/test/x11pdf.c
+++ b/test/x11pdf.c
@@ -29,6 +29,9 @@ static int rotate = 0;
 static int pageno = 1;
 static int count = 0;
 
+static pdf_page *page = nil;
+static fz_obj *pageobj = nil;
+
 static int hist[256];
 static int histlen = 0;
 
@@ -126,16 +129,13 @@ static void xtitle(char *s)
 static void showpage(void)
 {
 	fz_error *error;
-	pdf_page *page;
 	fz_matrix ctm;
 	fz_rect bbox;
+	fz_obj *obj;
 	char s[256];
-	fz_obj *pageobj;
 
 	assert(pageno > 0 && pageno <= pdf_getpagecount(pages));
 
-	pageobj = pdf_getpageobject(pages, pageno - 1);
-
 	XDefineCursor(xdpy, xwin, xcwait);
 	XFlush(xdpy);
 
@@ -143,6 +143,14 @@ static void showpage(void)
 		fz_droppixmap(image);
 	image = nil;
 
+	obj = pdf_getpageobject(pages, pageno - 1);
+	if (obj == pageobj)
+		goto Lskipload;
+	pageobj = obj;
+
+	if (page)
+		pdf_droppage(page);
+
 	sprintf(s, "Loading page %d", pageno);
 	XSetForeground(xdpy, xgc, BlackPixel(xdpy, xscr));
 	XDrawString(xdpy, xwin, xgc, 10, 20, s, strlen(s));
@@ -152,6 +160,8 @@ static void showpage(void)
 	if (error)
 		fz_abort(error);
 
+Lskipload:
+
 	sprintf(s, "Rendering...");
 	XSetForeground(xdpy, xgc, BlackPixel(xdpy, xscr));
 	XDrawString(xdpy, xwin, xgc, 10, 30, s, strlen(s));
@@ -168,8 +178,6 @@ static void showpage(void)
 	if (error)
 		fz_abort(error);
 
-	pdf_droppage(page);
-
 	XDefineCursor(xdpy, xwin, xcarrow);
 	XFlush(xdpy);
 
@@ -206,6 +214,20 @@ static void pdfopen(char *filename, char *password)
 	image = nil;
 }
 
+static void dumptext()
+{
+	fz_error *error;
+	pdf_textline *line;
+
+	error = pdf_loadtextfromtree(&line, page->tree);
+	if (error)
+		fz_abort(error);
+
+	pdf_debugtextline(line);
+
+	pdf_droptextline(line);
+}
+
 static void handlekey(int c)
 {
     int oldpage = pageno;
@@ -223,6 +245,7 @@ static void handlekey(int c)
 	case 'd': fz_debugglyphcache(rast->cache); break;
 	case 'a': rotate -= 5; break;
 	case 's': rotate += 5; break;
+	case 'x': dumptext(); break;
 
     case 'b':
 		pageno--;