From 80a6e0210be92a5d9c0ba5299e876463e7e718de Mon Sep 17 00:00:00 2001
From: HW <hwhilse@googlemail.com>
Date: Wed, 11 Apr 2012 21:12:59 +0200
Subject: [PATCH] add PDF text extraction (for pages)

this will return data in the way that djvu.c does already.
hopefully, this will permit us to re-use the highlighting code
(and factor it out into unireader.lua)
---
 pdf.c | 120 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 119 insertions(+), 1 deletion(-)

diff --git a/pdf.c b/pdf.c
index 87190f14d..7b0a811bf 100644
--- a/pdf.c
+++ b/pdf.c
@@ -311,6 +311,123 @@ static int openPage(lua_State *L) {
 	return 1;
 }
 
+/* get the text of the given page
+ * 
+ * will return text in a Lua table that is modeled after
+ * djvu.c creates this table.
+ *
+ * note that the definition of "line" is somewhat arbitrary
+ * here (for now)
+ *
+ * MuPDFs API provides text as single char information
+ * that is collected in "spans". we use a span as a "line"
+ * in Lua output and segment spans into words by looking
+ * for space characters.
+ *
+ * will return an empty table if we have no text
+ */
+static int getPageText(lua_State *L) {
+	fz_text_span *page_text;
+	fz_text_span *ptr;
+	fz_device *tdev;
+	fz_bbox bbox, linebbox;
+	fz_matrix ctm;
+	int i;
+	int word, line;
+	int len, c;
+	int start;
+	char chars[4]; // max length of UTF-8 encoded rune
+	luaL_Buffer textbuf;
+
+	PdfPage *page = (PdfPage*) luaL_checkudata(L, 1, "pdfpage");
+
+	/* returned coordinates are in centi-point (n * 0.01 pt) */
+	ctm = fz_scale(100, 100);
+
+	page_text = fz_new_text_span(page->doc->context);
+	tdev = fz_new_text_device(page->doc->context, page_text);
+	fz_run_page(page->doc->xref, page->page, tdev, ctm, NULL);
+	fz_free_device(tdev);
+
+	/* table that contains all the lines */
+	lua_newtable(L);
+	ptr = page_text;
+	line = 1;
+	while(ptr) {
+		/* table for the words */
+		lua_newtable(L);
+		word = 1;
+		linebbox = ptr->text[0].bbox; // start with sensible default
+		for(i = 0; i < ptr->len; ) {
+			/* will hold information about a word: */
+			lua_newtable(L);
+
+			luaL_buffinit(L, &textbuf);
+			bbox = ptr->text[i].bbox; // start with sensible default
+			for(; i < ptr->len; i++) {
+				/* check for space characters */
+				if(ptr->text[i].c == ' ' ||
+					ptr->text[i].c == '\t' ||
+					ptr->text[i].c == '\n' ||
+					ptr->text[i].c == '\v' ||
+					ptr->text[i].c == '\f' ||
+					ptr->text[i].c == '\r' ) {
+					// ignore and end word
+					i++;
+					break;
+				}
+				len = runetochar(chars, &ptr->text[i].c);
+				for(c = 0; c < len; c++) {
+					luaL_addchar(&textbuf, chars[c]);
+				}
+				bbox = fz_union_bbox(bbox, ptr->text[i].bbox);
+				linebbox = fz_union_bbox(linebbox, ptr->text[i].bbox);
+			}
+			lua_pushstring(L, "word");
+			luaL_pushresult(&textbuf);
+			lua_settable(L, -3);
+
+			/* bbox for a word: */
+			lua_pushstring(L, "x0");
+			lua_pushinteger(L, bbox.x0);
+			lua_settable(L, -3);
+			lua_pushstring(L, "y0");
+			lua_pushinteger(L, bbox.y0);
+			lua_settable(L, -3);
+			lua_pushstring(L, "x1");
+			lua_pushinteger(L, bbox.x1);
+			lua_settable(L, -3);
+			lua_pushstring(L, "y1");
+			lua_pushinteger(L, bbox.y1);
+			lua_settable(L, -3);
+
+			lua_rawseti(L, -2, word++);
+		}
+
+		/* bbox for a whole line (or in fact, a "span") */
+		lua_pushstring(L, "x0");
+		lua_pushinteger(L, linebbox.x0);
+		lua_settable(L, -3);
+		lua_pushstring(L, "y0");
+		lua_pushinteger(L, linebbox.y0);
+		lua_settable(L, -3);
+		lua_pushstring(L, "x1");
+		lua_pushinteger(L, linebbox.x1);
+		lua_settable(L, -3);
+		lua_pushstring(L, "y1");
+		lua_pushinteger(L, linebbox.y1);
+		lua_settable(L, -3);
+
+		lua_rawseti(L, -2, line++);
+
+		ptr = ptr->next;
+	}
+
+	fz_free_text_span(page->doc->context, page_text);
+
+	return 1;
+}
+
 static int getPageSize(lua_State *L) {
 	fz_matrix ctm;
 	fz_rect bounds;
@@ -323,7 +440,7 @@ static int getPageSize(lua_State *L) {
 	ctm = fz_concat(ctm, fz_rotate(dc->rotate));
 	bbox = fz_transform_rect(ctm, bounds);
 
-    lua_pushnumber(L, bbox.x1-bbox.x0);
+	lua_pushnumber(L, bbox.x1-bbox.x0);
 	lua_pushnumber(L, bbox.y1-bbox.y0);
 
 	return 2;
@@ -456,6 +573,7 @@ static const struct luaL_Reg pdfdocument_meth[] = {
 static const struct luaL_Reg pdfpage_meth[] = {
 	{"getSize", getPageSize},
 	{"getUsedBBox", getUsedBBox},
+	{"getPageText", getPageText},
 	{"close", closePage},
 	{"__gc", closePage},
 	{"draw", drawPage},