Faster conversion and cleaner code

pgaskin · Dec 16, 2017 · 0c08274 · 0c08274
1 parent 950cd0e
commit 0c08274
Show file tree

Hide file tree

Showing 2 changed files with 53 additions and 39 deletions.
diff --git a/kepub/content.go b/kepub/content.go
@@ -169,10 +169,12 @@ func addSpans(doc *goquery.Document) error {
 	return nil
 }
 
-// openSelfClosingPs opens self-closing p tags.
-func openSelfClosingPs(html *string) error {
-	re := regexp.MustCompile(`<p[^>/]*/>`)
-	*html = re.ReplaceAllString(*html, `<p></p>`)
+// addKoboStyles adds kobo styles.
+func addKoboStyles(doc *goquery.Document) error {
+	s := doc.Find("head").First().AppendHtml(`<style type="text/css">div#book-inner{margin-top: 0;margin-bottom: 0;}</style>`)
+	if s.Length() != 1 {
+		return fmt.Errorf("could not append kobo styles")
+	}
 	return nil
 }
 
@@ -191,28 +193,34 @@ func smartenPunctuation(html *string) error {
 }
 
 // cleanHTML cleans up html for a kobo epub.
-func cleanHTML(html *string) error {
-	emptyHeadingRe := regexp.MustCompile(`<h\d+>\s*</h\d+>`)
-	*html = emptyHeadingRe.ReplaceAllString(*html, "")
-
-	msPRe := regexp.MustCompile(`\s*<o:p>\s*<\/o:p>`)
-	*html = msPRe.ReplaceAllString(*html, " ")
-
-	msStRe := regexp.MustCompile(`<\/?st1:\w+>`)
-	*html = msStRe.ReplaceAllString(*html, "")
-
-	// unicode replacement chars
-	*html = strings.Replace(*html, "�", "", -1)
+func cleanHTML(doc *goquery.Document) error {
+	// Remove Adobe DRM tags
+	doc.Find(`meta[name="Adept.expected.resource"]`).Remove()
+
+	// Remove empty MS <o:p> tags
+	doc.Find(`o\:p`).FilterFunction(func(_ int, s *goquery.Selection) bool {
+		return strings.Trim(s.Text(), "\t \n") == ""
+	}).Remove()
+
+	// Remove empty headings
+	doc.Find(`h1,h2,h3,h4,h5,h6`).FilterFunction(func(_ int, s *goquery.Selection) bool {
+		return strings.Trim(s.Text(), "\t \n") == ""
+	}).Remove()
+
+	// Remove MS <st1:whatever> tags
+	doc.Find(`*`).FilterFunction(func(_ int, s *goquery.Selection) bool {
+		return strings.HasPrefix(goquery.NodeName(s), "st1:")
+	}).Remove()
+
+	// Open self closing p tags
+	doc.Find(`p`).Each(func(_ int, s *goquery.Selection) {
+		if s.Children().Length() == 0 && strings.Trim(s.Text(), "\n \t") == "" {
+			s.SetHtml("")
+		}
+	})
 
 	// Add type to style tags
-	*html = strings.Replace(*html, `<style>`, `<style type="text/css">`, -1)
-
-	// ADEPT drm tags
-	adeptRe := regexp.MustCompile(`(<meta\s+content=".+"\s+name="Adept.expected.resource"\s+\/>)`)
-	*html = adeptRe.ReplaceAllString(*html, "")
-
-	// Fix commented xml tag
-	*html = strings.Replace(*html, `<!-- ?xml version="1.0" encoding="utf-8"? -->`, `<?xml version="1.0" encoding="utf-8"?>`, 1)
+	doc.Find(`style`).SetAttr("type", "text/css")
 
 	return nil
 }
@@ -232,25 +240,29 @@ func process(content *string) error {
 		return err
 	}
 
-	h, err := doc.Html()
-	if err != nil {
+	if err := addKoboStyles(doc); err != nil {
 		return err
 	}
 
-	if err := openSelfClosingPs(&h); err != nil {
+	if err := cleanHTML(doc); err != nil {
 		return err
 	}
 
-	if err := cleanHTML(&h); err != nil {
+	h, err := doc.Html()
+	if err != nil {
 		return err
 	}
 
 	if err := smartenPunctuation(&h); err != nil {
 		return err
 	}
 
-	// Kobo style fixes
-	h = strings.Replace(h, "</head>", "<style type=\"text/css\">div#book-inner{margin-top: 0;margin-bottom: 0;}</style></head>", 1)
+	// Remove unicode replacement chars
+	h = strings.Replace(h, "�", "", -1)
+
+	// Fix commented xml tag
+	h = strings.Replace(h, `<!-- ?xml version="1.0" encoding="utf-8"? -->`, `<?xml version="1.0" encoding="utf-8"?>`, 1)
+	h = strings.Replace(h, `<!--?xml version="1.0" encoding="utf-8"?-->`, `<?xml version="1.0" encoding="utf-8"?>`, 1)
 
 	*content = h
 

diff --git a/kepub/content_test.go b/kepub/content_test.go
@@ -13,9 +13,17 @@ import (
 )
 
 func TestCleanHTML(t *testing.T) {
-	h := `<meta  content="urn:uuid:asd--asdasd-asdasdas-dasdasd234234"   name="Adept.expected.resource"   />��<st1:asd></st1:asd><o:p></o:p><h1></h1><h3></h3><h2>test</h2><style></style>`
-	cleanHTML(&h)
-	assert.Equal(t, " <h2>test</h2><style type=\"text/css\"></style>", h, "should be equal if cleaned correctly")
+	h := `<html><head></head><body><p /><p>test</p><p /><p  /><p>test</p><meta  content="urn:uuid:asd--asdasd-asdasdas-dasdasd234234"   name="Adept.expected.resource"   /><st1:asd></st1:asd><o:p></o:p><h1></h1><h3></h3><h2>test</h2><p>test</p><style></style></body></html>`
+
+	doc, err := goquery.NewDocumentFromReader(strings.NewReader(h))
+	assert.Nil(t, err, "err should be nil")
+
+	cleanHTML(doc)
+
+	nh, err := doc.Html()
+	assert.Nil(t, err, "err should be nil")
+
+	assert.Equal(t, `<html><head></head><body><p></p><p>test</p><p></p><p></p><p>test</p><h2>test</h2><p>test</p><style type="text/css"></style></body></html>`, nh, "should be equal if cleaned correctly")
 }
 
 func TestSmartenPunctuation(t *testing.T) {
@@ -24,12 +32,6 @@ func TestSmartenPunctuation(t *testing.T) {
 	assert.Equal(t, " &#x2014;   &#x2013;  <!-- test -->", h, "should be equal if smartened correctly")
 }
 
-func TestOpenSelfClosingPs(t *testing.T) {
-	h := `<p>test</p><p /><p  /><p>test</p>`
-	openSelfClosingPs(&h)
-	assert.Equal(t, "<p>test</p><p></p><p></p><p>test</p>", h, "should be equal if reopened correctly")
-}
-
 func TestAddSpans(t *testing.T) {
 	h := `<!DOCTYPE html>
 <html xmlns="http://www.w3.org/1999/xhtml">