Refactor the epub parser to it's own module

parser/language.go

@@ -0,0 +1,66 @@
+package parser
+
+import (
+	"io/ioutil"
+	"strings"
+
+	"github.com/meskio/epubgo"
+	"github.com/rainycape/cld2"
+)
+
+func GuessLang(epub *epubgo.Epub, orig_langs []string) []string {
+	spine, err := epub.Spine()
+	if err != nil {
+		return orig_langs
+	}
+
+	var err_spine error
+	err_spine = nil
+	langs := []string{}
+	for err_spine == nil {
+		html, err := spine.Open()
+		err_spine = spine.Next()
+		if err != nil {
+			continue
+		}
+		defer html.Close()
+
+		buff, err := ioutil.ReadAll(html)
+		if err != nil {
+			continue
+		}
+		langs = append(langs, cld2.Detect(string(buff)))
+	}
+
+	lang := commonLang(langs)
+	if lang != "un" && differentLang(lang, orig_langs) {
+		return []string{lang}
+	}
+	return orig_langs
+}
+
+func commonLang(langs []string) string {
+	count := map[string]int{}
+	for _, l := range langs {
+		count[l]++
+	}
+
+	lang := "un"
+	maxcount := 0
+	for l, c := range count {
+		if c > maxcount && l != "un" {
+			lang = l
+			maxcount = c
+		}
+	}
+	return lang
+}
+
+func differentLang(lang string, orig_langs []string) bool {
+	orig_lang := "un"
+	if len(orig_langs) > 0 && len(orig_langs) >= 2 {
+		orig_lang = strings.ToLower(orig_langs[0][0:2])
+	}
+
+	return orig_lang != lang
+}

parser/parser.go

@@ -0,0 +1,100 @@
+package parser
+
+import (
+	"regexp"
+	"strings"
+
+	"git.gitorious.org/go-pkg/epubgo.git"
+)
+
+type MetaData map[string]interface{}
+
+func EpubMetadata(epub *epubgo.Epub) MetaData {
+	metadata := MetaData{}
+	for _, m := range epub.MetadataFields() {
+		data, err := epub.Metadata(m)
+		if err != nil {
+			continue
+		}
+		switch m {
+		case "creator":
+			metadata["author"] = parseAuthr(data)
+		case "description":
+			metadata[m] = parseDescription(data)
+		case "subject":
+			metadata[m] = parseSubject(data)
+		case "date":
+			metadata[m] = parseDate(data)
+		case "language":
+			metadata["lang"] = GuessLang(epub, data)
+		case "title", "contributor", "publisher":
+			metadata[m] = cleanStr(strings.Join(data, ", "))
+		case "identifier":
+			attr, _ := epub.MetadataAttr(m)
+			for i, d := range data {
+				if attr[i]["scheme"] == "ISBN" {
+					metadata["isbn"] = d
+				}
+			}
+		default:
+			metadata[m] = strings.Join(data, ", ")
+		}
+	}
+	return metadata
+}
+
+func cleanStr(str string) string {
+	str = strings.Replace(str, "'", "'", -1)
+	exp, _ := regexp.Compile("&[^;]*;")
+	str = exp.ReplaceAllString(str, "")
+	exp, _ = regexp.Compile("[ ,]*$")
+	str = exp.ReplaceAllString(str, "")
+	return str
+}
+
+func parseAuthr(creator []string) []string {
+	exp1, _ := regexp.Compile("^(.*\\( *([^\\)]*) *\\))*$")
+	exp2, _ := regexp.Compile("^[^:]*: *(.*)$")
+	res := make([]string, len(creator))
+	for i, s := range creator {
+		auth := exp1.FindStringSubmatch(s)
+		if auth != nil {
+			res[i] = cleanStr(strings.Join(auth[2:], ", "))
+		} else {
+			auth := exp2.FindStringSubmatch(s)
+			if auth != nil {
+				res[i] = cleanStr(auth[1])
+			} else {
+				res[i] = cleanStr(s)
+			}
+		}
+	}
+	return res
+}
+
+func parseDescription(description []string) string {
+	str := cleanStr(strings.Join(description, "\n"))
+	str = strings.Replace(str, "</p>", "\n", -1)
+	exp, _ := regexp.Compile("<[^>]*>")
+	str = exp.ReplaceAllString(str, "")
+	str = strings.Replace(str, "&amp;", "&", -1)
+	str = strings.Replace(str, "&lt;", "<", -1)
+	str = strings.Replace(str, "&gt;", ">", -1)
+	str = strings.Replace(str, "\\n", "\n", -1)
+	return str
+}
+
+func parseSubject(subject []string) []string {
+	var res []string
+	for _, s := range subject {
+		res = append(res, strings.Split(s, " / ")...)
+	}
+	return res
+}
+
+func parseDate(date []string) string {
+	if len(date) == 0 {
+		return ""
+	}
+	return strings.Replace(date[0], "Unspecified: ", "", -1)
+}