Merge pull request #4 from TheBookPeople/headings

Add support for headings

Author: k3a

html2text.go

@@ -9,6 +9,7 @@ import (
 var badTagnamesRE = regexp.MustCompile(`^(head|script|style|a)($|\s*)`)
 var linkTagRE = regexp.MustCompile(`a.*href=('([^']*?)'|"([^"]*?)")`)
 var badLinkHrefRE = regexp.MustCompile(`#|javascript:`)
+var headersRE = regexp.MustCompile(`^(\/)?h[1-6]`)
 
 func parseHTMLEntity(entName string) (string, bool) {
 	entName = strings.ToLower(entName)
@@ -160,7 +161,12 @@ func HTML2Text(html string) string {
 			shouldOutput = true
 			tagName := strings.ToLower(html[tagStart:i])
 
-			if tagName == "br" || tagName == "br/" {
+			if headersRE.MatchString(tagName) {
+				if canPrintNewline {
+					outBuf.WriteString("\r\n\r\n")
+				}
+				canPrintNewline = false
+			} else if tagName == "br" || tagName == "br/" {
 				// new line
 				outBuf.WriteString("\r\n")
 			} else if tagName == "p" || tagName == "/p" {

html2text_test.go

@@ -26,6 +26,21 @@ func TestHTML2Text(t *testing.T) {
 			So(HTML2Text(`<p>two</p><p>paragraphs</p>`), ShouldEqual, "two\r\nparagraphs")
 		})
 
+		Convey("Headings", func() {
+			So(HTML2Text("<h1>First</h1>main text"), ShouldEqual, "First\r\n\r\nmain text")
+			So(HTML2Text("First<h2>Second</h2>next section"), ShouldEqual, "First\r\n\r\nSecond\r\n\r\nnext section")
+			So(HTML2Text("<h2>Second</h2>next section"), ShouldEqual, "Second\r\n\r\nnext section")
+			So(HTML2Text("Second<h3>Third</h3>next section"), ShouldEqual, "Second\r\n\r\nThird\r\n\r\nnext section")
+			So(HTML2Text("<h3>Third</h3>next section"), ShouldEqual, "Third\r\n\r\nnext section")
+			So(HTML2Text("Third<h4>Fourth</h4>next section"), ShouldEqual, "Third\r\n\r\nFourth\r\n\r\nnext section")
+			So(HTML2Text("<h4>Fourth</h4>next section"), ShouldEqual, "Fourth\r\n\r\nnext section")
+			So(HTML2Text("Fourth<h5>Fifth</h5>next section"), ShouldEqual, "Fourth\r\n\r\nFifth\r\n\r\nnext section")
+			So(HTML2Text("<h5>Fifth</h5>next section"), ShouldEqual, "Fifth\r\n\r\nnext section")
+			So(HTML2Text("Fifth<h6>Sixth</h6>next section"), ShouldEqual, "Fifth\r\n\r\nSixth\r\n\r\nnext section")
+			So(HTML2Text("<h6>Sixth</h6>next section"), ShouldEqual, "Sixth\r\n\r\nnext section")
+			So(HTML2Text("<h7>Not Header</h7>next section"), ShouldEqual, "Not Headernext section")
+		})
+
 		Convey("HTML entities", func() {
 			So(HTML2Text(`two&nbsp;&nbsp;spaces`), ShouldEqual, "two  spaces")
 			So(HTML2Text(`&copy; 2017 K3A`), ShouldEqual, "© 2017 K3A")