add text highlight in both reflow and non-reflow mode

Makefile

@@ -66,7 +66,10 @@ customupdate: all
 	cp -p README.md COPYING $(KOR_BASE)/{koreader-base,extr} koreader.sh $(LUA_FILES) $(INSTALL_DIR)
 	$(STRIP) --strip-unneeded $(INSTALL_DIR)/koreader-base $(INSTALL_DIR)/extr
 	mkdir $(INSTALL_DIR)/data
-	cp -L koreader-base/$(DJVULIB) $(KOR_BASE)/$(CRELIB) $(KOR_BASE)/$(LUALIB) $(KOR_BASE)/$(K2PDFOPTLIB) $(INSTALL_DIR)/libs
+	cp -L koreader-base/$(DJVULIB) $(KOR_BASE)/$(CRELIB) \
+		$(KOR_BASE)/$(LUALIB) $(KOR_BASE)/$(K2PDFOPTLIB) \
+		$(KOR_BASE)/$(LEPTONICALIB) $(KOR_BASE)/$(TESSERACTLIB) \
+		$(INSTALL_DIR)/libs
 	$(STRIP) --strip-unneeded $(INSTALL_DIR)/libs/*
 	cp -rpL $(KOR_BASE)/data/*.css $(INSTALL_DIR)/data
 	cp -rpL $(KOR_BASE)/fonts $(INSTALL_DIR)

frontend/document/djvudocument.lua

@@ -47,6 +47,25 @@ function validDjvuFile(filename)
 	return true
 end
 
+function DjvuDocument:getPageText(pageno)
+	if self.configurable.text_wrap == 1 then
+		return self.koptinterface:getPageText(self, pageno)
+	else
+		return self._document:getPageText(pageno)
+	end
+end
+
+function DjvuDocument:getOCRWord(pageno, rect)
+	if self.configurable.text_wrap == 1 then
+		return self.koptinterface:getOCRWord(self, pageno, rect)
+	else
+		--local page = self._document:openPage(pageno)
+		--local word = page:getOCRWord(rect)
+		--page:close()
+		--return word
+	end
+end
+
 function DjvuDocument:getUsedBBox(pageno)
 	-- djvu does not support usedbbox, so fake it.
 	local used = {}

frontend/document/document.lua

@@ -184,6 +184,14 @@ function Document:getToc()
 	return self._document:getToc()
 end
 
+function Document:getPageText(pageno)
+	return nil
+end
+
+function Document:getOCRWord(pageno, rect)
+	return nil
+end
+
 function Document:renderPage(pageno, rect, zoom, rotation, gamma, render_mode)
 	local hash = "renderpg|"..self.file.."|"..pageno.."|"..zoom.."|"..rotation.."|"..gamma.."|"..render_mode
 	local page_size = self:getPageDimensions(pageno, zoom, rotation)

frontend/document/koptinterface.lua

@@ -1,9 +1,23 @@
+require "dbg"
 require "cache"
 require "ui/geometry"
 require "ui/device"
 require "ui/reader/readerconfig"
 
-KoptInterface = {}
+KoptInterface = {
+	tessocr_data = "data",
+	ocr_lang = "eng",
+	ocr_type = 0, -- default, for more accuracy use 3
+}
+
+ContextCacheItem = CacheItem:new{}
+
+function ContextCacheItem:onFree()
+	if self.kctx.free then
+		DEBUG("free koptcontext", self.kctx)
+		self.kctx:free()
+	end
+end
 
 function KoptInterface:waitForContext(kc)
 	-- if koptcontext is being processed in background thread
@@ -12,10 +26,13 @@ function KoptInterface:waitForContext(kc)
 		DEBUG("waiting for background rendering")
 		util.usleep(100000)
 	end
+	return kc
 end
 
--- get reflow context
-function KoptInterface:getKOPTContext(doc, pageno, bbox)
+--[[
+get reflow context
+--]]
+function KoptInterface:createContext(doc, pageno, bbox)
 	-- Now koptcontext keeps track of its dst bitmap reflowed by libk2pdfopt.
 	-- So there is no need to check background context when creating new context.
 	local kc = KOPTContext.new()
@@ -37,6 +54,7 @@ function KoptInterface:getKOPTContext(doc, pageno, bbox)
 	kc:setLineSpacing(doc.configurable.line_spacing)
 	kc:setWordSpacing(doc.configurable.word_spacing)
 	kc:setBBox(bbox.x0, bbox.y0, bbox.x1, bbox.y1)
+	if Dbg.is_on then kc:setDebug() end
 	return kc
 end
 
@@ -47,17 +65,6 @@ function KoptInterface:getContextHash(doc, pageno, bbox)
 	return doc.file.."|"..pageno.."|"..doc.configurable:hash("|").."|"..bbox_hash.."|"..screen_size_hash
 end
 
-function KoptInterface:logReflowDuration(pageno, dur)
-	local file = io.open("reflowlog.txt", "a+")
-	if file then
-		if file:seek("end") == 0 then -- write the header only once
-			file:write("PAGE\tDUR\n")
-		end
-		file:write(string.format("%s\t%s\n", pageno, dur))
-		file:close()
-	end
-end
-
 function KoptInterface:getAutoBBox(doc, pageno)
 	local bbox = {
 		x0 = 0, y0 = 0,
@@ -68,26 +75,68 @@ function KoptInterface:getAutoBBox(doc, pageno)
 	local cached = Cache:check(hash)
 	if not cached then
 		local page = doc._document:openPage(pageno)
-		local kc = self:getKOPTContext(doc, pageno, bbox)
+		local kc = self:createContext(doc, pageno, bbox)
 		bbox.x0, bbox.y0, bbox.x1, bbox.y1 = page:getAutoBBox(kc)
 		DEBUG("Auto detected bbox", bbox)
 		page:close()
-		Cache:insert(hash, CacheItem:new{ bbox = bbox })
+		Cache:insert(hash, CacheItem:new{ autobbox = bbox })
 		return bbox
 	else
-		return cached.bbox
+		return cached.autobbox
 	end
 end
 
--- get reflowed page dimension from a cached context. wait for background thread
--- if necessary.
-function KoptInterface:getReflowedDim(kctx)
-	self:waitForContext(kctx)
-	return kctx:getPageDim()
+function KoptInterface:getPageText(doc, pageno)
+	local bbox = doc:getPageBBox(pageno)
+	local context_hash = self:getContextHash(doc, pageno, bbox)
+	local hash = "pgtext|"..context_hash
+	local cached = Cache:check(hash)
+	if not cached then
+		local kctx_hash = "kctx|"..context_hash
+		local cached = Cache:check(kctx_hash)
+		if cached then
+			local kc = self:waitForContext(cached.kctx)
+			local fullwidth, fullheight = kc:getPageDim()
+			local text = kc:getWordBoxes(0, 0, fullwidth, fullheight)
+			Cache:insert(hash, CacheItem:new{ pgtext = text })
+			return text
+		end
+	else
+		return cached.pgtext
+	end
 end
 
--- get cached koptcontext for centain page. if context doesn't exist in cache make
--- new context and reflow the src page immediatly.
+function KoptInterface:getOCRWord(doc, pageno, rect)
+	local bbox = doc:getPageBBox(pageno)
+	local context_hash = self:getContextHash(doc, pageno, bbox)
+	local hash = "ocrword|"..context_hash..rect.x..rect.y..rect.w..rect.h
+	local cached = Cache:check(hash)
+	if not cached then
+		local kctx_hash = "kctx|"..context_hash
+		local cached = Cache:check(kctx_hash)
+		if cached then
+			local kc = self:waitForContext(cached.kctx)
+			local fullwidth, fullheight = kc:getPageDim()
+			local ok, word = pcall(
+				kc.getOCRWord, kc,
+				self.tessocr_data,
+				self.ocr_lang,
+				self.ocr_type, 
+				rect.x, rect.y,
+				rect.w, rect.h)
+			Cache:insert(hash, CacheItem:new{ ocrword = word })
+			return word
+		end
+	else
+		return cached.ocrword
+	end
+end
+
+--[[
+get cached koptcontext for centain page. if context doesn't exist in cache make
+new context and reflow the src page immediatly, or wait background thread for 
+reflowed context.
+--]]
 function KoptInterface:getCachedContext(doc, pageno)
 	local bbox = doc:getPageBBox(pageno)
 	local context_hash = self:getContextHash(doc, pageno, bbox)
@@ -95,7 +144,7 @@ function KoptInterface:getCachedContext(doc, pageno)
 	local cached = Cache:check(kctx_hash)
 	if not cached then
 		-- If kctx is not cached, create one and get reflowed bmp in foreground.
-		local kc = self:getKOPTContext(doc, pageno, bbox)
+		local kc = self:createContext(doc, pageno, bbox)
 		local page = doc._document:openPage(pageno)
 		-- reflow page
 		--local secs, usecs = util.gettime()
@@ -107,22 +156,27 @@ function KoptInterface:getCachedContext(doc, pageno)
 		--self:logReflowDuration(pageno, dur)
 		local fullwidth, fullheight = kc:getPageDim()
 		DEBUG("reflowed page", pageno, "fullwidth:", fullwidth, "fullheight:", fullheight)
-		Cache:insert(kctx_hash, CacheItem:new{ kctx = kc })
+		Cache:insert(kctx_hash, ContextCacheItem:new{ kctx = kc })
 		return kc
 	else
-		return cached.kctx
+		-- wait for background thread
+		return self:waitForContext(cached.kctx)
 	end
 end
 
--- get reflowed page dimensions
+--[[
+get reflowed page dimensions
+--]]
 function KoptInterface:getPageDimensions(doc, pageno, zoom, rotation)
-	local kctx = self:getCachedContext(doc, pageno)
-	local fullwidth, fullheight = self:getReflowedDim(kctx)
+	local kc = self:getCachedContext(doc, pageno)
+	local fullwidth, fullheight = kc:getPageDim()
 	return Geom:new{ w = fullwidth, h = fullheight }
 end
 
--- inherited from common document interface
--- render reflowed page into tile cache. 
+--[[
+inherited from common document interface
+render reflowed page into tile cache.
+--]] 
 function KoptInterface:renderPage(doc, pageno, rect, zoom, rotation, render_mode)
 	doc.render_mode = render_mode
 	local bbox = doc:getPageBBox(pageno)
@@ -132,8 +186,8 @@ function KoptInterface:renderPage(doc, pageno, rect, zoom, rotation, render_mode
 	local cached = Cache:check(renderpg_hash)
 	if not cached then
 		-- do the real reflowing if kctx is not been cached yet
-		local kctx = self:getCachedContext(doc, pageno)
-		local fullwidth, fullheight = self:getReflowedDim(kctx)
+		local kc = self:getCachedContext(doc, pageno)
+		local fullwidth, fullheight = kc:getPageDim()
 		if not Cache:willAccept(fullwidth * fullheight / 2) then
 			-- whole page won't fit into cache
 			error("aborting, since we don't have enough cache for this page")
@@ -146,10 +200,8 @@ function KoptInterface:renderPage(doc, pageno, rect, zoom, rotation, render_mode
 			pageno = pageno,
 			bb = Blitbuffer.new(fullwidth, fullheight)
 		}
-		page:rfdraw(kctx, tile.bb)
+		page:rfdraw(kc, tile.bb)
 		page:close()
-		-- free dst bitmap in kctx as soon as we draw the bitmap to blitbuffer
-		kctx:free()
 		Cache:insert(renderpg_hash, tile)
 		return tile
 	else
@@ -157,29 +209,33 @@ function KoptInterface:renderPage(doc, pageno, rect, zoom, rotation, render_mode
 	end
 end
 
--- inherited from common document interface
--- render reflowed page into cache in background thread. this method returns immediatly
--- leaving the precache flag on in context. subsequent usage of this context should 
--- wait for the precache flag off by calling self:waitForContext(kctx)
+--[[
+inherited from common document interface
+render reflowed page into cache in background thread. this method returns immediatly
+leaving the precache flag on in context. subsequent usage of this context should 
+wait for the precache flag off by calling self:waitForContext(kctx)
+--]]
 function KoptInterface:hintPage(doc, pageno, zoom, rotation, gamma, render_mode)
 	local bbox = doc:getPageBBox(pageno)
 	local context_hash = self:getContextHash(doc, pageno, bbox)
 	local kctx_hash = "kctx|"..context_hash
 	local cached = Cache:check(kctx_hash)
 	if not cached then
-		local kc = self:getKOPTContext(doc, pageno, bbox)
+		local kc = self:createContext(doc, pageno, bbox)
 		local page = doc._document:openPage(pageno)
 		DEBUG("hinting page", pageno, "in background")
 		-- reflow will return immediately and running in background thread
 		kc:setPreCache()
 		page:reflow(kc, 0)
 		page:close()
-		Cache:insert(kctx_hash, CacheItem:new{ kctx = kc })
+		Cache:insert(kctx_hash, ContextCacheItem:new{ kctx = kc })
 	end
 end
 
--- inherited from common document interface
--- draw cached tile pixels into target blitbuffer
+--[[
+inherited from common document interface
+draw cached tile pixels into target blitbuffer.
+--]]
 function KoptInterface:drawPage(doc, target, x, y, rect, pageno, zoom, rotation, render_mode)
 	local tile = self:renderPage(doc, pageno, rect, zoom, rotation, render_mode)
 	--DEBUG("now painting", tile, rect)
@@ -189,3 +245,17 @@ function KoptInterface:drawPage(doc, target, x, y, rect, pageno, zoom, rotation,
 		rect.y - tile.excerpt.y,
 		rect.w, rect.h)
 end
+
+--[[
+helper functions
+--]]
+function KoptInterface:logReflowDuration(pageno, dur)
+	local file = io.open("reflowlog.txt", "a+")
+	if file then
+		if file:seek("end") == 0 then -- write the header only once
+			file:write("PAGE\tDUR\n")
+		end
+		file:write(string.format("%s\t%s\n", pageno, dur))
+		file:close()
+	end
+end

frontend/document/pdfdocument.lua

@@ -44,6 +44,28 @@ function PdfDocument:unlock(password)
 	return self:_readMetadata()
 end
 
+function PdfDocument:getPageText(pageno)
+	if self.configurable.text_wrap == 1 then
+		return self.koptinterface:getPageText(self, pageno)
+	else
+		local page = self._document:openPage(pageno)
+		local text = page:getPageText()
+		page:close()
+		return text
+	end
+end
+
+function PdfDocument:getOCRWord(pageno, rect)
+	if self.configurable.text_wrap == 1 then
+		return self.koptinterface:getOCRWord(self, pageno, rect)
+	else
+		--local page = self._document:openPage(pageno)
+		--local word = page:getOCRWord(rect)
+		--page:close()
+		--return word
+	end
+end
+
 function PdfDocument:getUsedBBox(pageno)
 	local hash = "pgubbox|"..self.file.."|"..pageno
 	local cached = Cache:check(hash)

frontend/ui/reader/readerdictionary.lua

@@ -0,0 +1,24 @@
+require "ui/device"
+
+ReaderDictionary = EventListener:new{}
+
+function ReaderDictionary:init()
+	local dev_mod = Device:getModel()
+	if dev_mod == "KindlePaperWhite" or dev_mod == "KindleTouch" then
+		require "liblipclua"
+		self.lipc_handle = lipc.init("com.github.koreader.dictionary")
+	end
+end
+
+function ReaderDictionary:onLookupWord(word)
+	DEBUG("lookup word:", word)
+	if self.lipc_handle then
+		-- start indicator depends on pillow being enabled
+		self.lipc_handle:set_string_property(
+			"com.lab126.booklet.kpvbooklet.dict", "lookup", word)
+		local definitions = self.lipc_handle:get_string_property(
+			"com.lab126.booklet.kpvbooklet.word", word)
+		DEBUG("definitions of word:", word, definitions)
+	end
+	return true
+end

frontend/ui/reader/readerhighlight.lua

@@ -0,0 +1,125 @@
+
+ReaderHighlight = InputContainer:new{}
+
+function ReaderHighlight:init()
+	if Device:hasKeyboard() then
+		self.key_events = {
+			ShowToc = {
+				{ "." },
+				doc = _("highlight text") },
+		}
+	end
+end
+
+function ReaderHighlight:initGesListener()
+	self.ges_events = {
+		Tap = {
+			GestureRange:new{
+				ges = "tap",
+				range = Geom:new{
+					x = 0, y = 0,
+					w = Screen:getWidth(),
+					h = Screen:getHeight()
+				}
+			}
+		},
+		Hold = {
+			GestureRange:new{
+				ges = "hold",
+				range = Geom:new{
+					x = 0, y = 0,
+					w = Screen:getWidth(),
+					h = Screen:getHeight()
+				}
+			}
+		},
+	}
+end
+
+function ReaderHighlight:onSetDimensions(dimen)
+	-- update listening according to new screen dimen
+	if Device:isTouchDevice() then
+		self:initGesListener()
+	end
+end
+
+function ReaderHighlight:onTap(arg, ges)
+	if self.view.highlight.rect then
+		self.view.highlight.rect = nil
+		UIManager:setDirty(self.dialog, "partial")
+		return true
+	end
+end
+
+function ReaderHighlight:onHold(arg, ges)
+	self.pos = self.view:screenToPageTransform(ges.pos)
+	DEBUG("hold position in page", self.pos)
+	local text = self.ui.document:getPageText(self.pos.page)
+	--DEBUG("page text", text)
+	
+	if not text or #text == 0 then
+		DEBUG("no text extracted")
+		return true
+	end
+	
+	self.word_info = self:getWordFromText(text, self.pos)
+	DEBUG("hold word info in page", self.word_info)
+	if self.word_info then
+		local screen_rect = self.view:pageToScreenTransform(self.pos.page, self.word_info.box)
+		DEBUG("highlight word rect", screen_rect)
+		if screen_rect then
+			self.view.highlight.rect = screen_rect
+			UIManager:setDirty(self.dialog, "partial")
+			-- if we extracted text directly
+			if self.word_info.word then
+				self.ui:handleEvent(Event:new("LookupWord", self.word_info.word))
+			-- or we will do OCR
+			else
+				UIManager:scheduleIn(0.1, function()
+					local word_box = self.word_info.box
+					word_box.x = word_box.x - math.floor(word_box.h * 0.02)
+					word_box.y = word_box.y - math.floor(word_box.h * 0.02)
+					word_box.w = word_box.w + math.floor(word_box.h * 0.04)
+					word_box.h = word_box.h + math.floor(word_box.h * 0.04)
+					local word = self.ui.document:getOCRWord(self.pos.page, word_box)
+					DEBUG("OCRed word:", word)
+					self.ui:handleEvent(Event:new("LookupWord", word))
+				end)
+			end
+		end
+	end
+	return true
+end
+
+function ReaderHighlight:getWordFromText(text, pos)
+	local function ges_inside(x0, y0, x1, y1)
+		local x, y = pos.x, pos.y
+		if x0 ~= nil and y0 ~= nil and x1 ~= nil and y1 ~= nil then
+			if x0 <= x and y0 <= y and x1 >= x and y1 >= y then
+				return true
+			end
+		end
+		return false
+	end
+	
+	for i = 1, #text do
+		local l = text[i]
+		if ges_inside(l.x0, l.y0, l.x1, l.y1) then
+			--DEBUG("line box", l.x0, l.y0, l.x1, l.y1)
+			for j = 1, #text[i] do
+				local w = text[i][j]
+				if ges_inside(w.x0, w.y0, w.x1, w.y1) then
+					local box = Geom:new{
+						x = w.x0, y = w.y0, 
+						w = w.x1 - w.x0,
+						h = w.y1 - w.y0,
+					}
+					return {
+						word = w.word,
+						box = box,
+					}
+				end -- end if inside word box
+			end -- end for each word
+		end -- end if inside line box
+	end -- end for each line
+end

frontend/ui/reader/readerkopt.lua

@@ -41,3 +41,11 @@ end
 function ReaderKoptListener:onFineTuningFontSize(delta)
 	self.document.configurable.font_size = self.document.configurable.font_size + delta
 end
+
+function ReaderKoptListener:onZoomUpdate(zoom)
+	-- an exceptional case is reflow mode 
+	if self.document.configurable.text_wrap == 1 then
+		self.view.state.zoom = 1.0
+	end
+end
+

frontend/ui/reader/readerview.lua

@@ -16,6 +16,10 @@ ReaderView = OverlapGroup:new{
 		bbox = nil,
 	},
 	outer_page_color = 0,
+	-- hightlight
+	highlight = {
+		drawer = "marker" -- show as inverted block instead of underline
+	},
 	-- PDF/DjVu continuous paging
 	page_scroll = nil,
 	page_bgcolor = 0,
@@ -96,6 +100,11 @@ function ReaderView:paintTo(bb, x, y)
 			self.dim_area.w, self.dim_area.h
 		)
 	end
+	
+	-- draw highlight
+	if self.highlight.rect then
+		self:drawHightlight(bb, x, y, self.highlight.rect)
+	end
 
 	-- paint dogear
 	if self.dogear_visible then
@@ -113,6 +122,32 @@ function ReaderView:paintTo(bb, x, y)
 	self.ui:handleEvent(Event:new("StopActivityIndicator"))
 end
 
+--[[
+Given coordinates on the screen return position in original page
+]]--
+function ReaderView:screenToPageTransform(pos)
+	if self.ui.document.info.has_pages then
+		if self.page_scroll then
+			return self:getScrollPagePosition(pos)
+		else
+			return self:getSinglePagePosition(pos)
+		end
+	end
+end
+
+--[[
+Given rectangle in original page return rectangle on the screen
+]]--
+function ReaderView:pageToScreenTransform(page, rect)
+	if self.ui.document.info.has_pages then
+		if self.page_scroll then
+			return self:getScrollPageRect(page, rect)
+		else
+			return self:getSinglePageRect(rect)
+		end
+	end
+end
+
 function ReaderView:drawPageBackground(bb, x, y)
 	bb:paintRect(x, y, self.dimen.w, self.dimen.h, self.page_bgcolor)
 end
@@ -155,6 +190,40 @@ function ReaderView:drawScrollPages(bb, x, y)
 	end)
 end
 
+function ReaderView:getScrollPagePosition(pos)
+	local x_s, y_s = pos.x, pos.y
+	local x_p, y_p = nil, nil
+	for _, state in ipairs(self.page_states) do
+		if y_s < state.visible_area.h + state.offset.y then
+			y_p = (state.visible_area.y + y_s - state.offset.y) / state.zoom
+			x_p = (state.visible_area.x + x_s - state.offset.x) / state.zoom
+			return {
+				x = x_p,
+				y = y_p,
+				page = state.page,
+				zoom = state.zoom,
+				rotation = state.rotation,
+			}
+		else
+			y_s = y_s - state.visible_area.h - self.page_gap.height
+		end
+	end
+end
+
+function ReaderView:getScrollPageRect(page, rect_p)
+	local rect_s = Geom:new{}
+	for _, state in ipairs(self.page_states) do
+		if page == state.page and state.visible_area:contains(rect_p) then
+			rect_s.x = rect_s.x + state.offset.x + rect_p.x*state.zoom - state.visible_area.x
+			rect_s.y = rect_s.y + state.offset.y + rect_p.y*state.zoom - state.visible_area.y
+			rect_s.w = rect_p.w * state.zoom
+			rect_s.h = rect_p.h * state.zoom
+			return rect_s
+		end
+		rect_s.y = rect_s.y + state.visible_area.h + self.page_gap.height
+	end
+end
+
 function ReaderView:drawPageGap(bb, x, y)
 	if self.scroll_mode == "vertical" then
 		bb:paintRect(x, y, self.dimen.w, self.page_gap.height, self.page_gap.color)
@@ -179,6 +248,28 @@ function ReaderView:drawSinglePage(bb, x, y)
 	end)
 end
 
+function ReaderView:getSinglePagePosition(pos)
+	local x_s, y_s = pos.x, pos.y
+	return {
+		x = (self.visible_area.x + x_s - self.state.offset.x) / self.state.zoom,
+		y = (self.visible_area.y + y_s - self.state.offset.y) / self.state.zoom,
+		page = self.state.page,
+		zoom = self.state.zoom,
+		rotation = self.state.rotation,
+	}
+end
+
+function ReaderView:getSinglePageRect(rect_p)
+	local rect_s = Geom:new{}
+	if self.visible_area:contains(rect_p) then
+		rect_s.x = self.state.offset.x + rect_p.x * self.state.zoom - self.visible_area.x
+		rect_s.y = self.state.offset.y + rect_p.y * self.state.zoom - self.visible_area.y
+		rect_s.w = rect_p.w * self.state.zoom
+		rect_s.h = rect_p.h * self.state.zoom
+		return rect_s
+	end
+end
+
 function ReaderView:drawPageView(bb, x, y)
 	self.ui.document:drawCurrentViewByPage(
 		bb,
@@ -197,6 +288,26 @@ function ReaderView:drawScrollView(bb, x, y)
 		self.state.pos)
 end
 
+function ReaderView:drawHightlight(bb, x, y, rect)
+	-- slightly enlarge the highlight box
+	-- for better viewing experience
+	local x = rect.x - rect.h * 0.01
+	local y = rect.y - rect.h * 0.01
+	local w = rect.w + rect.h * 0.02
+	local h = rect.h + rect.h * 0.02
+	
+	self.highlight.drawer = self.highlight.drawer or "underscore"
+	if self.highlight.drawer == "underscore" then
+		self.highlight.line_width = self.highlight.line_width or 2
+		self.highlight.line_color = self.highlight.line_color or 5
+		bb:paintRect(x, y+h-1, w,
+			self.highlight.line_width,
+			self.highlight.line_color)
+	elseif self.highlight.drawer == "marker" then
+		bb:invertRect(x, y, w, h)
+	end
+end
+
 function ReaderView:getPageArea(page, zoom, rotation)
 	if self.use_bbox then
 		return self.ui.document:getUsedBBoxDimensions(page, zoom, rotation)

frontend/ui/readerui.lua

@@ -14,8 +14,10 @@ require "ui/reader/readercropping"
 require "ui/reader/readerkopt"
 require "ui/reader/readercopt"
 require "ui/reader/readerhinting"
+require "ui/reader/readerhighlight"
 require "ui/reader/readerscreenshot"
 require "ui/reader/readerfrontlight"
+require "ui/reader/readerdictionary"
 require "ui/reader/readerhyphenation"
 require "ui/reader/readeractivityindicator"
 
@@ -96,6 +98,22 @@ function ReaderUI:init()
 		ui = self
 	}
 	table.insert(self, reader_bm)
+	-- text highlight 
+	local highlight = ReaderHighlight:new{
+		dialog = self.dialog,
+		view = self[1],
+		ui = self,
+		document = self.document,
+	}
+	table.insert(self, highlight)
+	-- dictionary 
+	local dict = ReaderDictionary:new{
+		dialog = self.dialog,
+		view = self[1],
+		ui = self,
+		document = self.document,
+	}
+	table.insert(self, dict)
 	-- screenshot controller
 	local reader_ss = ReaderScreenshot:new{
 		dialog = self.dialog,

koreader.sh

@@ -9,6 +9,9 @@ test -e $PROC_FIVEWAY && echo unlock > $PROC_FIVEWAY
 # we're always starting from our working directory
 cd /mnt/us/koreader/
 
+# export trained OCR data directory
+export TESSDATA_PREFIX="data"
+
 # bind-mount system fonts
 if ! grep /mnt/us/koreader/fonts/host /proc/mounts; then
 	mount -o bind /usr/java/lib/fonts /mnt/us/koreader/fonts/host