nlnieuws/RTVNoord/cmd/rtvnoord/rtvnoord.go

package main

import (
	e "codeberg.org/pebbe/errors"

	u "git.web.rug.nl/p209327/nlnieuws/internal/util"

	"encoding/json"
	"encoding/xml"
	"fmt"
	"html"
	"io"
	"net/http"
	"os"
	"strings"
	"time"
)

type Rss struct {
	XMLName xml.Name `xml:"rss"`
	Items   []ItemT  `xml:"channel>item"`
}

type ItemT struct {
	PubDate  string `xml:"pubDate"`
	UnixTime int64  `xml:"unixTime"`
	Guid     string `xml:"guid"`
	Link     string `xml:"link"`
	Data     []byte `xml:",innerxml"`
}

type Doc struct {
	Type  string   `json:"@type"`
	Cat   string   `json:"articleSection"`
	Tags  []string `json:"keywords"`
	Title string   `json:"headline"`
	Text  string   `json:"articleBody"`
}

type GItem struct {
	ArticleBody    string   `json:"articleBody"`
	ArticleSection []string `json:"articleSection"`
}

var (
	p     = e.PanicErr
	w     = e.WarnErr
	agent = "AhrefsBot/7.0"
)

func exists(filename string) bool {
	_, err := os.Stat(filename)
	return err == nil
}

func fileDate(filename string) string {
	b, err := os.ReadFile(filename)
	if err != nil {
		return ""
	}
	s := string(b)
	i1 := strings.Index(s, "<unixTime>") + 10
	i2 := strings.Index(s, "</unixTime>")
	if i2 < i1 {
		return ""
	}
	return s[i1:i2]
}

func main() {
	defer func() {
		if e.Panicked {
			_ = recover()
			os.Exit(1)
		}
	}()

	myLock := "/net/corpora/nlnieuws/RTVNoord/lock"
	u.MkLock(myLock)
	defer func() {
		_ = os.Remove(myLock)
	}()

	resp, err := http.Get("https://www.rtvnoord.nl/rss/index.xml")
	p(err)
	body, err := io.ReadAll(resp.Body)
	p(err)
	p(resp.Body.Close())

	var rss Rss
	p(xml.Unmarshal(body, &rss))

	if len(rss.Items) == 0 {
		p(fmt.Errorf("len(rss.Items) == 0"))
	}

	for _, item := range rss.Items {
		t, err := time.Parse(time.RFC1123Z, item.PubDate)
		if err != nil {
			t, err = time.Parse(time.RFC1123, item.PubDate)
		}
		p(err)
		year, week := t.ISOWeek()
		dirname := fmt.Sprintf("/net/corpora/nlnieuws/RTVNoord/%d/w%02d", year, week)
		if exists(dirname + "/lock") {
			continue
		}
		guid := strings.TrimPrefix(item.Guid, "https://www.rtvnoord.nl/")
		guid = strings.TrimRight(guid, "/-")
		guid = strings.ReplaceAll(guid, "/", "_")
		filename := dirname + "/" + guid

		ts := fmt.Sprintf("%d", t.Unix())
		needUpdate := fileDate(filename+".xml") != ts

		p(os.MkdirAll(dirname, 0777))
		func() {
			var ok bool
			defer func() {
				if e.Panicked {
					fmt.Fprintln(os.Stderr, "----", filename)
					fmt.Fprintln(os.Stderr, "----", item.Link)
				}
				if !ok {
					_ = os.Remove(filename + ".xml")
				}
			}()
			fp, err := os.Create(filename + ".xml")
			p(err)
			p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
			p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
			p(fp.Write(item.Data))
			p(fp.WriteString("</item>\n"))
			p(fp.Close())
			p(os.Chtimes(filename+".xml", t, t))
			ok = doArticle(filename, item.Link, t, needUpdate)
		}()
	}
}

func doArticle(filename string, url string, timestamp time.Time, needUpdate bool) bool {
	if exists(filename + ".skip") {
		return true
	}
	if needUpdate {
		_ = os.Remove(filename + ".err")
		_ = os.Remove(filename + ".html")
		_ = os.Remove(filename + ".json")
		_ = os.Remove(filename + ".txt")
	} else {
		// voor sommige berichten is geen .txt, alleen .json
		if exists(filename + ".json") {
			return true
		}
	}

	time.Sleep(2 * time.Second)

	req, err := http.NewRequest("GET", url, nil)
	p(err)
	req.Header.Set("User-Agent", agent)

	client := &http.Client{}
	resp, err := client.Do(req)
	p(err)
	body, err := io.ReadAll(resp.Body)
	p(err)
	p(resp.Body.Close())

	s := string(body)
	ok := true
	var doc Doc
	skip := 0
	for {
		var i2 int
		var s2 string
		i1 := strings.Index(s[skip:], `type="application/ld+json"`) + skip
		if i1 < 0 {
			ok = false
			break
		} else {
			i1 += strings.Index(s[i1:], `>`) + 1
			i2 = i1 + strings.Index(s[i1:], `</script>`)
			if i2 < i1 {
				ok = false
				break
			} else {
				s2 = html.UnescapeString(s[i1:i2])
			}
		}
		if err = json.Unmarshal([]byte(s2), &doc); err != nil {
			_ = w(err, url)
			skip = i2
			continue
		}
		if doc.Type == "Organization" {
			skip = i2
			continue
		}
		s = s2
		break
	}
	if !ok {
		_ = w(fmt.Errorf("script jsonld not found: %s", url))

		fp, err := os.Create(filename + ".err")
		p(err)
		p(fmt.Fprintf(fp, "script jsonld not found: %s\n", url))
		p(fp.Close())
		p(os.Chtimes(filename+".err", timestamp, timestamp))

		fp, err = os.Create(filename + ".html")
		p(err)
		p(fp.Write(body))
		p(fp.Close())
		p(os.Chtimes(filename+".html", timestamp, timestamp))

		return false
	}
	fp, err := os.Create(filename + ".json")
	p(err)
	p(fp.WriteString(s))
	p(fp.Close())
	p(os.Chtimes(filename+".json", timestamp, timestamp))

	// text bevat kopjes zonder punt aan het eind
	lines := strings.Split(doc.Text, "\n")
	for i, line := range lines {
		lines[i] = u.AddEnd(u.FixSpace(line))
	}
	text := strings.Join(lines, "") + "\n"

	fp, err = os.Create(filename + ".txt")
	p(err)
	if len(doc.Tags) == 0 {
		p(fmt.Fprintln(fp, "##META text tag ="))
	} else {
		for _, tag := range doc.Tags {
			t := strings.ToLower(u.FixSpace(tag))
			if strings.HasPrefix(t, "br_") {
				continue
			}
			if strings.HasPrefix(t, "tr_") {
				t = t[3:]
			}
			p(fmt.Fprintf(fp, "##META text tag = %s\n", t))
		}
	}
	if doc.Cat == "" {
		p(fmt.Fprintln(fp, "##META text cat ="))
	} else {
		p(fmt.Fprintf(fp, "##META text cat = %s\n", u.FixSpace(doc.Cat)))
	}

	p(fp.WriteString(u.AddEnd(doc.Title)))

	p(fp.WriteString(text))
	p(fp.Close())
	p(os.Chtimes(filename+".txt", timestamp, timestamp))

	return true
}