package main import ( e "codeberg.org/pebbe/errors" "github.com/jbowtie/gokogiri" u "git.web.rug.nl/p209327/nlnieuws/internal/util" "encoding/xml" "fmt" "io" "net/http" "net/url" "os" "strings" "time" ) type Rss struct { XMLName xml.Name `xml:"rss"` Items []ItemT `xml:"channel>item"` } type ItemT struct { Title string `xml:"title"` 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" agent = "Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/145.0.0.0 Safari/537.36" ) 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/GG/lock" u.MkLock(myLock) defer func() { _ = os.Remove(myLock) }() req, err := http.NewRequest("GET", "https://gemeente.groningen.nl/feed/rss/nieuws", 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()) 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/GG/%d/w%02d", year, week) if exists(dirname + "/lock") { continue } filename := dirname + "/" + url.PathEscape(item.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("\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, item.Title, t, needUpdate) }() } } func doArticle(filename string, url string, title string, timestamp time.Time, needUpdate bool) (ok bool) { if exists(filename + ".skip") { return true } if needUpdate { _ = os.Remove(filename + ".err") _ = os.Remove(filename + ".html") _ = os.Remove(filename + ".txt") } else { if exists(filename + ".txt") { 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()) body = u.HtmlFix(body) doc, err := gokogiri.ParseHtml(body) p(err) root := doc.Root() svgs, err := root.Search(`//svg`) p(err) for _, svg := range svgs { svg.Remove() } ell, err := root.Search( `//div[contains(@class,"component-richtext")]/p` + ` | ` + `//div[contains(@class,"component-richtext")]/h2`) p(err) if len(ell) == 0 { _ = w(fmt.Errorf("empty: %s", url)) fp, err := os.Create(filename + ".err") p(err) p(fmt.Fprintf(fp, "empty: %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 + ".txt") p(err) p(fp.WriteString(u.AddEnd(u.FixSpace(title)))) for _, el := range ell { p(fp.WriteString(u.AddEnd(u.FixSpace(el.Content())))) } p(fp.Close()) p(os.Chtimes(filename+".txt", timestamp, timestamp)) return true }