Autodetect ISBN numbers in the text if not found in metadata

isbn_detect.go

+package liber
+
+import (
+	"io"
+	"io/ioutil"
+	"regexp"
+	"strconv"
+	"strings"
+
+	"git.autistici.org/ale/liber/Godeps/_workspace/src/github.com/meskio/epubgo"
+)
+
+var isbnRx = regexp.MustCompile(`(?:ISBN(?:-10|-13)?\s*:?\s*)?((?:(?:[\d]-?){9}|(?:[\d]-?){12})[\dxX])(?:[^-\d]|$)`)
+
+func findISBNInEpub(epub *epubgo.Epub) []string {
+	var isbn []string
+
+	spine, err := epub.Spine()
+	if err != nil {
+		return nil
+	}
+
+	for spine.Next() == nil {
+		r, err := spine.Open()
+		if err != nil {
+			continue
+		}
+		if found := findISBNInPage(r); len(found) > 0 {
+			isbn = append(isbn, found...)
+		}
+		r.Close()
+	}
+
+	return isbn
+}
+
+func findISBNInPage(r io.Reader) []string {
+	data, err := ioutil.ReadAll(r)
+	if err != nil {
+		return nil
+	}
+	var result []string
+	for _, m := range isbnRx.FindAllSubmatch(data, -1) {
+		if len(m) > 1 {
+			isbn := string(m[1])
+			if validateIsbn(isbn) {
+				result = append(result, isbn)
+			}
+		}
+	}
+	return result
+}
+
+func validateIsbn10(isbn string) bool {
+	var sum int
+	var multiply int = 10
+	for i, v := range isbn {
+		if v == '-' {
+			continue
+		}
+		digitString := string(v)
+
+		if i == 9 && digitString == "X" {
+			digitString = "10"
+		}
+
+		digit, err := strconv.Atoi(digitString)
+		if err != nil {
+			return false
+		}
+		sum = sum + (multiply * digit)
+		multiply--
+	}
+
+	return sum%11 == 0
+}
+
+func validateIsbn13(isbn string) bool {
+	var sum int
+	for i, v := range isbn {
+		var multiply int
+		if i%2 == 0 {
+			multiply = 1
+		} else {
+			multiply = 3
+		}
+
+		digit, err := strconv.Atoi(string(v))
+		if err != nil {
+			return false
+		}
+		sum = sum + (multiply * digit)
+	}
+
+	return sum%10 == 0
+}
+
+func validateIsbn(isbn string) bool {
+	isbn = strings.Replace(isbn, "-", "", -1)
+	switch len(isbn) {
+	case 10:
+		return validateIsbn10(isbn)
+	case 13:
+		return validateIsbn13(isbn)
+	default:
+		return false
+	}
+}

isbn_detect_test.go

+package liber
+
+import (
+	"reflect"
+	"strings"
+	"testing"
+)
+
+var isbnTestData = []struct {
+	input    string
+	expected []string
+}{
+	{"not-an-isbn", nil},
+	{"123456780-330-28498-312345678", nil},
+
+	// Valid ISBN numbers.
+	{"ISBN: 0-330-28498-3", []string{"0-330-28498-3"}},
+	{"ISBN : 1-58182-008-9", []string{"1-58182-008-9"}},
+	{"ISBN-10: 2-226-05257-7", []string{"2-226-05257-7"}},
+	{"ISBN 3-7965-1900-8", []string{"3-7965-1900-8"}},
+	{"ISBN 4-19-830127-1", []string{"4-19-830127-1"}},
+	{"ISBN 5-85270-001-0", []string{"5-85270-001-0"}},
+	{"ISBN 978-600-119-125-1", []string{"978-600-119-125-1"}},
+	{"978-601-7151-13-3", []string{"978-601-7151-13-3"}},
+	{"ISBN-13: 978-602-8328-22-7", []string{"978-602-8328-22-7"}},
+	{"ISBN 978-603-500-045-1", []string{"978-603-500-045-1"}},
+	{"ISBN 605-384-057-2", []string{"605-384-057-2"}},
+	{"ISBN 978-606-8126-35-7", []string{"978-606-8126-35-7"}},
+	{"ISBN 978-607-455-035-1", []string{"978-607-455-035-1"}},
+	{"ISBN: 978-608-203-023-4", []string{"978-608-203-023-4"}},
+	{"ISBN 978-612-45165-9-7", []string{"978-612-45165-9-7"}},
+	{"ISBN 978-614-404-018-8", []string{"978-614-404-018-8"}},
+	{"ISBN 978-615-5014-99-4", []string{"978-615-5014-99-4"}},
+	{"ISBN 7-301-10299-2", []string{"7-301-10299-2"}},
+	{"ISBN 80-85983-44-3", []string{"80-85983-44-3"}},
+	{"81-7215-399-6", []string{"81-7215-399-6"}},
+	{"82-530-0983-6", []string{"82-530-0983-6"}},
+	{"eISBN : 83-08-01587-5", []string{"83-08-01587-5"}},
+	{"eISBN : 84-86546-08-7", []string{"84-86546-08-7"}},
+	{"ISBN 85-7531-015-1", []string{"85-7531-015-1"}},
+	{"ISBN 86-341-0846-5", []string{"86-341-0846-5"}},
+	{"ISBN 87-595-2277-1", []string{"87-595-2277-1"}},
+	{"88-04-47328-2", []string{"88-04-47328-2"}},
+	{"ISBN 90-5691-187-2", []string{"90-5691-187-2"}},
+	{"ISBN 91-1-811692-2", []string{"91-1-811692-2"}},
+	{"ISBN 92-67-10370-9", []string{"92-67-10370-9"}},
+	{"ISBN 93-5025-214-7", []string{"93-5025-214-7"}},
+	{"ISBN 950-04-0442-7", []string{"950-04-0442-7"}},
+	{"ISBN 951-0-11369-7", []string{"951-0-11369-7"}},
+	{"ISBN 952-471-294-6", []string{"952-471-294-6"}},
+	{"ISBN 953-157-105-8", []string{"953-157-105-8"}},
+	{"ISBN 954-430-603-X", []string{"954-430-603-X"}},
+	{"ISBN 955-20-3051-X", []string{"955-20-3051-X"}},
+	{"ISBN 956-7291-48-9", []string{"956-7291-48-9"}},
+	{"ISBN 957-01-7429-3", []string{"957-01-7429-3"}},
+	{"ISBN 958-04-6278-X", []string{"958-04-6278-X"}},
+	{"ISBN 959-10-0363-3", []string{"959-10-0363-3"}},
+	{"ISBN 961-6403-23-0", []string{"961-6403-23-0"}},
+	{"ISBN 962-04-0195-6", []string{"962-04-0195-6"}},
+	{"ISBN 978-988-00-3827-3", []string{"978-988-00-3827-3"}},
+	{"ISBN: 978-9928400529", []string{"978-9928400529"}},
+	{"ISBN: 978-9929801646", []string{"978-9929801646"}},
+	{"ISBN: 978-9930943106", []string{"978-9930943106"}},
+
+	// Multiple results.
+	{"ISBN: 0-330-28498-3 and ISBN : 1-58182-008-9",
+		[]string{"0-330-28498-3", "1-58182-008-9"}},
+
+	// Invalid ISBN numbers (wrong check digit).
+	{"ISBN 961-6403-23-1", nil},
+	{"ISBN 962-04-0195-1", nil},
+	{"ISBN: 978-9928400521", nil},
+	{"ISBN: 978-9929801641", nil},
+	{"ISBN: 978-993094310X", nil},
+}
+
+func TestISBN_Detect(t *testing.T) {
+	for _, td := range isbnTestData {
+		isbn := findISBNInPage(strings.NewReader(td.input))
+		if !reflect.DeepEqual(isbn, td.expected) {
+			t.Errorf("input: %s, got: %v, expected: %v", td.input, isbn, td.expected)
+		}
+	}
+}

metadata.go

@@ -3,6 +3,7 @@ package liber
 import (
 	"errors"
 	"fmt"
+	"log"
 	"path/filepath"
 	"regexp"
 	"strings"
@@ -192,6 +193,14 @@ func parseEpub(filename string) (*Metadata, error) {
 		}
 	}
 
+	// If we haven't found an ISBN, look for it in the book text.
+	if len(m.ISBN) == 0 {
+		if isbn := findISBNInEpub(e); len(isbn) > 0 {
+			log.Printf("found ISBN in book text: %s", strings.Join(isbn, ", "))
+			m.ISBN = isbn
+		}
+	}
+
 	m.Sources = []MetadataSource{{
 		Name: "epub",
 		ID:   filename,