package main import ( "html" e "codeberg.org/pebbe/errors" "encoding/xml" "fmt" "io" "net/http" "net/url" "os" "path/filepath" "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"` } 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, "") + 10 i2 := strings.Index(s, "") if i2 < i1 { return "" } return s[i1:i2] } func main() { defer func() { if e.Panicked { _ = recover() os.Exit(1) } }() myLock := "/net/corpora/nlnieuws/NOS/lock" mkLock(myLock) defer func() { _ = os.Remove(myLock) }() resp, err := http.Get("https://feeds.nos.nl/nosnieuwsalgemeen") 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 { var t time.Time for _, format := range []string{ "Mon, 2 Jan 2006 15:04:05 -0700", "Mon, 2 Jan 2006 15:04:05 MST", time.RFC1123, time.RFC1123Z} { t, err = time.Parse(format, item.PubDate) if err == nil { break } } p(err) year, week := t.ISOWeek() dirname := fmt.Sprintf("/net/corpora/nlnieuws/NOS/%d/%02d", year, week) if exists(dirname + "/lock") { continue } filename := dirname + "/" + url.PathEscape(strings.TrimPrefix(item.Guid, "https://nos.nl/l/")) 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("\n\n")) p(fmt.Fprintf(fp, "%d", t.Unix())) p(fp.Write(item.Data)) p(fp.WriteString("\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") } else { 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 i1 := strings.Index(s, ``) if i2 < i1 { ok = false } else { s = html.UnescapeString(s[i1:i2]) } } 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)) return true } func mkLock(filename string) { pid := os.Getpid() link := fmt.Sprintf("%s.%d", filepath.Base(filename), pid) p(os.Symlink(link, filename)) name, err := os.Readlink(filename) p(err) if name != link { p(fmt.Errorf("wrong lock name %q, should be %q", name, link)) } }