func score(tag *goquery.Selection) int {
src, _ := tag.Attr("src")
if src == "" {
src, _ = tag.Attr("data-src")
}
if src == "" {
src, _ = tag.Attr("data-lazy-src")
}
if src == "" {
return -1
}
tagScore := 0
for rule, score := range rules {
if rule.MatchString(src) {
tagScore += score
}
}
alt, exists := tag.Attr("alt")
if exists {
if strings.Contains(alt, "thumbnail") {
tagScore--
}
}
id, exists := tag.Attr("id")
if exists {
if id == "fbPhotoImage" {
tagScore++
}
}
return tagScore
}
func guessSourceURL(s *goquery.Selection, link *url.URL) string {
possibleSrcAttr := []string{"src",
"data-src",
"srcset",
"data-full-size",
"data-original",
"href", // bbc
"pagespeed_lazy_src",
}
var possibleSrcURLs []string
for _, attr := range possibleSrcAttr {
link, exist := s.Attr(attr)
if exist {
Debug("possible image src url under: %v", attr)
possibleSrcURLs = append(possibleSrcURLs, link)
}
}
if len(possibleSrcURLs) == 0 {
return ""
}
validSrcURLs := removeInvalidURLs(possibleSrcURLs, link)
if len(validSrcURLs) == 0 {
return ""
}
indexLongestElement, _ := longestElement(validSrcURLs)
return validSrcURLs[indexLongestElement]
}
// attributeOrDefault reads an attribute and returns it or the default value when it's empty.
func (bow *Browser) attrOrDefault(name, def string, sel *goquery.Selection) string {
a, ok := sel.Attr(name)
if ok {
return a
}
return def
}
// Add street to result and find street information (to WikipediaMoscow.result)
func (parser *WikipediaMoscow) processLink(_ int, s *goquery.Selection, done chan<- *StreetInfo) {
name := strings.TrimSpace(s.Text())
if len(name) == 0 {
done <- parser.getEmptyInfo("")
return
}
href, exists := s.Attr("href")
if !exists {
done <- parser.getEmptyInfo("")
return
}
var (
class string
info *StreetInfo
)
class, exists = s.Attr("class")
if exists && class == "new" {
info = parser.getEmptyInfo(name)
} else if resp, err := http.Get(parser.baseURL + href); err != nil {
info = parser.getEmptyInfo(name)
} else {
streetparser := NewWikipediaStreetParser()
info, err = streetparser.ParseStreetInfo(name, resp.Body)
if err != nil {
info = parser.getEmptyInfo(name)
}
}
done <- info
}
func (this *parser) name(selector string, selection *goquery.Selection) string {
value, exists := selection.Attr(selector)
if exists {
return value
}
return ""
}
// toPage is a helper function that accepts an anchor
// tag referencing a markdown file, parsing the markdown
// file and returning a page to be included in our docs.
func toPage(site *Site, el *goquery.Selection) (*Page, error) {
// follow the link to see if this is a page
// that should be added to our documentation.
href, ok := el.Attr("href")
if !ok || href == "#" {
return nil, nil
}
// read the markdown file, convert to html and
// read into a dom element.
doc, err := toDocument(filepath.Join(site.base, href))
if err != nil {
return nil, err
}
// convert the extension from markdown to
// html, in preparation for type conversion.
href = strings.Replace(href, ".md", ".html", -1)
el.SetAttr("href", href)
page := &Page{}
page.Href = href
page.html, err = doc.Html()
return page, err
}
func (ve *VideoExtractor) getSrc(node *goquery.Selection) string {
value, exists := node.Attr("src")
if exists {
return value
}
return ""
}
func convertTagToJqueryFormat(tag string, s *goquery.Selection) string {
tagitself := tag
pos := strings.Index(tag, " ")
if pos > -1 {
tagitself = tag[0:pos]
} else {
return tag
}
class, found := s.Attr("class")
if found && class != "" {
pos := strings.Index(class, " ")
// leave only a first class from a list
if pos > -1 {
class = class[0:pos]
}
tagitself = tagitself + "." + class
}
return tagitself
}
func (d *Document) classWeight(s *goquery.Selection) int {
weight := 0
if !d.WeightClasses {
return weight
}
class, _ := s.Attr("class")
id, _ := s.Attr("id")
if class != "" {
if negativeRegexp.MatchString(class) {
weight -= 25
}
if positiveRegexp.MatchString(class) {
weight += 25
}
}
if id != "" {
if negativeRegexp.MatchString(id) {
weight -= 25
}
if positiveRegexp.MatchString(id) {
weight += 25
}
}
return weight
}
// Parse from div.tweet
func (tweet *Tweet) Parse(s *goquery.Selection) (err error) {
success := false
attrs := []string{
"data-item-id",
"data-screen-name",
"data-name",
}
data := map[string]string{}
for _, attr := range attrs {
var value string
if value, success = s.Attr(attr); !success {
tweet.Success = 0
err = fmt.Errorf("not having %s attribute", attr)
return
}
data[attr] = value
}
tweet.ItemID = data["data-item-id"]
tweet.ScreenName = data["data-screen-name"]
tweet.Name = data["data-name"]
tweet.Success = 1
// if could get the above attribues, allow the following values to be blank.
tweet.Time, _ = s.Find("._timestamp").Attr("data-time")
tweet.Text = s.Find(".tweet-text").Text()
return
}
func ScrapeNotes(s *goquery.Selection) []string {
notes := []string{}
s.Find("abbr").Not("abbr:first-of-type").Not("abbr.c").Each(func(i int, s *goquery.Selection) {
note, _ := s.Attr("title")
notes = append(notes, note)
})
return notes
}
func ScrapeOrigins(s *goquery.Selection) []string {
origins := []string{}
s.Find("abbr.c").Each(func(i int, s *goquery.Selection) {
origin, _ := s.Attr("title")
origins = append(origins, origin)
})
return origins
}
func ScrapeNotes(s *goquery.Selection) []string {
notes := []string{}
s.Clone().Find("span[title]").First().Remove().End().End().Find("span.d i span.d[title]").Remove().End().Find("span.d[title]").Each(func(i int, s *goquery.Selection) {
note, _ := s.Attr("title")
notes = append(notes, note)
})
return notes
}
func attrToUrl(s *goquery.Selection, attr string) (*url.URL, error) {
link, exists := s.Attr(attr)
if exists {
return url.Parse(link)
}
return nil, errors.New("Attr " + attr + " not found")
}
func (ve *VideoExtractor) getHeight(node *goquery.Selection) int {
value, exists := node.Attr("height")
if exists {
nvalue, _ := strconv.Atoi(value)
return nvalue
}
return 0
}
func getNumbericAttribute(s *goquery.Selection, attr string) int {
a, f := s.Attr(attr)
if f {
ai, _ := strconv.Atoi(a)
return ai
}
return 0
}
func getHeight(sel *goquery.Selection) string {
heightAttr := "height"
height, exist := sel.Attr(heightAttr)
if exist {
return height
}
return ""
}
func getDataFromDOM(s *gq.Selection, arr []string, code string) string {
var dt string
if arr[0] == "text" {
dt = s.Text()
} else {
dt, _ = s.Attr(arr[0])
}
return encode_string(dt, code)
}
func itemURL(s *goquery.Selection) (url string) {
s.Closest("tr").Next().Find("a").Each(func(_ int, s *goquery.Selection) {
href, _ := s.Attr("href")
if strings.HasPrefix(href, "item?id=") {
url = hnURL + href
}
})
return
}
// guessCodeLang returns the code language if supported, otherwise and empty
// string.
func guessCodeLang(sel *goquery.Selection) string {
codeLangAttr, _ := sel.Attr("class")
if codeLangAttr == "" {
codeLangAttr, _ = sel.Attr("lang")
}
return extractCodeLang(codeLangAttr)
}
/**
* This function parses and returns the uri associated with the HTML anchor
* <a href="http://www..."...> tag
* This function assumes that 'href' attribute contains absolute url.
* It returns "" empty string if it can't find href attribute from the
* goquery.Selection parameter.
*/
func getUri(sel *goquery.Selection) string {
if sel != nil {
str, exists := sel.Attr("href")
if exists {
u, err := url.Parse(str)
checkErr(err)
return u.String()
}
}
return ""
}
func displayDetails(single *goquery.Selection) {
text := strings.TrimSpace(single.Text())
href, _ := single.Attr("href")
length := utf8.RuneCountInString(text)
if ((length > 5) && wordExists(text, "keywords")) || ((length > 5) && wordExists(href, "keywords")) {
if wordExists(text, "products") {
fmt.Println("Link", single.Text(), "--->", href)
}
}
}
func (this *contentExtractor) getNodeGravityScore(node *goquery.Selection) int {
grvScoreString, exists := node.Attr("gravityScore")
if !exists {
return 0
}
grvScore, err := strconv.Atoi(grvScoreString)
if err != nil {
return 0
}
return grvScore
}
func addIngredient(ingredients []data.Ingredient, a *goquery.Selection) []data.Ingredient {
if href, ok := a.Attr("href"); ok {
glog.V(2).Info(" href: " + href)
id, err := strconv.Atoi(strings.Split(href, "/")[2])
if err != nil {
glog.Errorf("Failed to extract id from %s: %v", href, err)
} else {
ingredients = append(ingredients, data.Ingredient{Name: a.Text(), Id: id})
}
}
return ingredients
}
func parseResource(s *goquery.Selection) (_production, _stored, _capacity int) {
productionStr, _ := s.Attr("title")
production, _ := strconv.Atoi(productionStr)
status := s.Text()
split := strings.Split(status, "/")
stored, _ := strconv.Atoi(split[0])
capacity, _ := strconv.Atoi(split[1])
return production, stored, capacity
}
func ParseRecipe(sel *goquery.Selection) (*Recipe, error) {
if _, exists := sel.Attr("itemscope"); !exists {
return nil, ErrMissingItemScope
}
itemtype, exists := sel.Attr("itemtype")
if !exists {
return nil, ErrMissingItemType
}
if itemtype != RecipeSchemaURL {
return nil, ErrWrongItemType
}
recipe := &Recipe{
CreativeWork: CreativeWork{
Thing: Thing{},
},
}
nameSel := sel.Find("[itemprop='name']")
recipe.Name = strings.TrimSpace(nameSel.Text())
authorSel := sel.Find("[itemprop='author']").First()
recipe.Author = strings.TrimSpace(authorSel.Text())
datePublishedSel := sel.Find("[itemprop='datePublished']")
datePublishedText, exists := datePublishedSel.Attr("content")
if !exists {
datePublishedText = datePublishedSel.Text()
}
var err error
if len(datePublishedText) != 0 {
recipe.DatePublished, err = time.Parse("2006-01-02", datePublishedText)
if err != nil {
return nil, err
}
}
nutritionInformationSel := sel.Find(fmt.Sprintf("[itemscope=''][itemtype='%s']", NutritionInformationSchemaURL))
if nutritionInformationSel.Size() > 0 {
recipe.Nutrition, err = ParseNutritionInformation(nutritionInformationSel)
if err != nil {
return nil, err
}
}
imageSel := sel.Find("[itemprop='image']")
recipe.Image, _ = imageSel.Attr("src")
descriptionSel := sel.Find("[itemprop='description']")
recipe.Description = strings.TrimSpace(descriptionSel.Text())
return recipe, nil
}
//stores how many decent nodes are under a parent node
func (this *contentExtractor) updateNodeCount(node *goquery.Selection, addToCount int) {
currentScore := 0
var err error
scoreString, _ := node.Attr("gravityNodes")
if scoreString != "" {
currentScore, err = strconv.Atoi(scoreString)
if err != nil {
currentScore = 0
}
}
newScore := currentScore + addToCount
this.config.parser.setAttr(node, "gravityNodes", strconv.Itoa(newScore))
}
//adds a score to the gravityScore Attribute we put on divs
//we'll get the current score then add the score we're passing in to the current
func (extr *ContentExtractor) updateScore(node *goquery.Selection, addToScore int) {
currentScore := 0
var err error
scoreString, _ := node.Attr("gravityScore")
if scoreString != "" {
currentScore, err = strconv.Atoi(scoreString)
if err != nil {
currentScore = 0
}
}
newScore := currentScore + addToScore
extr.config.parser.setAttr(node, "gravityScore", strconv.Itoa(newScore))
}
// filterLITag finds street names
func filterLITag(_ int, s *goquery.Selection) bool {
_, exists := s.Attr("id")
if exists {
return false
}
_, exists = s.Attr("class")
if exists {
return false
}
return true
}
// attributeToUrl reads an attribute from an element and returns a url.
func (bow *Browser) attrToResolvedUrl(name string, sel *goquery.Selection) (*url.URL, error) {
src, ok := sel.Attr(name)
if !ok {
return nil, errors.NewAttributeNotFound(
"Attribute '%s' not found.", name)
}
ur, err := url.Parse(src)
if err != nil {
return nil, err
}
return bow.ResolveUrl(ur), nil
}
