kkscrap-go/controllers/scraper/wordpress/parser.go

package wordpress

import (
	"encoding/json"
	"fmt"
	"kkscrap-go/controllers/scraper/common"
	"kkscrap-go/controllers/scraper/util"
	"kkscrap-go/model"
	"strings"

	"github.com/PuerkitoBio/goquery"
)

func Parse(body string, item *model.ItemInfo) {
	p := getProduct(body)
	// if p.Image != "" {
	// 	item.Images = append(item.Images, p.Image)
	// }
	item.Images = append(item.Images, getImages(body)...)
	if p.Offers != nil && len(p.Offers) > 0 {
		item.SalesPrice = util.GetFloat32(p.Offers[0].Price)
		item.Currency = p.Offers[0].Pricecurrency
	}
	item.Sku = fmt.Sprintf("%v", p.Sku)
	item.ShortDesc = p.Description
	item.OriginDesc = getProductDescription(body)
	item.TextDesc = getTextDesc(body)
	item.ItemName = p.Name
	item.Language = common.GetLanguage(body)
	item.Emails = common.GetEmails(body)
	item.Options = append(item.Options, getOptions(body))
	cats := getCategories(body)
	catss := []string{}
	if strings.Contains(cats, ",") {
		catss = strings.Split(cats, ",")
	} else if strings.Contains(cats, "/") {
		catss = strings.Split(cats, "/")
	}

	for _, v := range catss {
		item.ItemCategory = append(item.ItemCategory, strings.TrimSpace(v))
	}

	return
}

func getProductDescription(body string) string {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(body))
	util.CheckError(err)
	s := doc.Find("div.woocommerce-Tabs-panel--description")
	//s.Each(func(i int, selection *goquery.Selection) {
	//	log.Println(selection.Html())
	//})
	html, err := s.Html()
	if err != nil {
		return ""
	}
	return html
}

func getTextDesc(body string) string {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(body))
	util.CheckError(err)
	text := doc.Find("div.woocommerce-tabs").Text()
	return text
}

func getProduct(body string) (ret WordPressProduct) {
	jsonStr := getJson(body)
	if idx := strings.Index(jsonStr, "@graph"); idx >= 0 {
		p := WordPressItem{}
		json.Unmarshal([]byte(jsonStr), &p)
		ret = p.Graph[1]
	} else {
		json.Unmarshal([]byte(jsonStr), &ret)
	}
	return
}

func getJson(body string) string {
	idx := strings.Index(body, "<script type=\"application/ld+json\">")
	if idx < 0 {
		return ""
	}
	body = body[idx+len("<script type=\"application/ld+json\">"):]
	idx = strings.Index(body, "</script>")
	body = body[:idx]
	return body
}

func getImages(body string) (ret []string) {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(body))
	util.CheckError(err)
	//doc.Find("img.wp-post-image").Each(func(i int, s *goquery.Selection) {
	//	if src, ok := s.Attr("src"); ok {
	//		width, _ := s.Attr("width")
	//		height, _ := s.Attr("height")
	//		w, _ := strconv.ParseInt(width, 10, 64)
	//		h, _ := strconv.ParseInt(height, 10, 64)
	//		//log.Println(src, width, height)
	//		ret = append(ret, model.Image{
	//			Path:   src,
	//			Width:  int(w),
	//			Height: int(h),
	//		})
	//	}
	//})
	//sort.Slice(ret, func(i, j int) bool {
	//	return ret[i].Width*ret[i].Height > ret[j].Width*ret[j].Height
	//})
	//ret = ret[:1]

	// re := regexp.MustCompile(`-\d+x\d+\.`)
	doc.Find("figure.woocommerce-product-gallery__wrapper div a").Each(func(i int, s *goquery.Selection) {
		if src, ok := s.Attr("href"); ok {
			ret = append(ret, src)
		}
	})

	return
}

func getPrice(body string) (ret string) {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(body))
	util.CheckError(err)
	s := doc.Find("span.woocommerce-Price-amount.amount")
	s.Each(func(i int, selection *goquery.Selection) {
		if i == 0 {
			ret = selection.Text()
		}
	})
	//log.Println(s.Nodes[0].FirstChild.FirstChild.Data)
	//log.Println(s.Nodes[0].FirstChild.NextSibling.Data)
	return
}

func getCategories(body string) (ret string) {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(body))
	util.CheckError(err)
	s := doc.Find("span.posted_in")
	ret = s.Text()
	idx := strings.Index(ret, ":")
	if idx > 0 {
		ret = strings.TrimSpace(ret[idx+1:])
	}
	//log.Println(s.Nodes[0].FirstChild.FirstChild.Data)
	//log.Println(s.Nodes[0].FirstChild.NextSibling.Data)
	return
}

func getOptions(body string) (ret model.Option) {
	doc, err := goquery.NewDocumentFromReader(strings.NewReader(body))
	util.CheckError(err)
	s := doc.Find("body select")

	s.Each(func(i int, selection *goquery.Selection) {
		skippedFistOption := false
		v, _ := selection.Attr("name")
		if v != "rating" {
			selection.Find("option").Each(func(i int, selection *goquery.Selection) {
				if skippedFistOption {
					ret.Choices = append(ret.Choices, model.Choice{
						Name: selection.Text(),
					})
				} else {
					skippedFistOption = true
				}
			})
		}
	})
	return
}