package main import ( "github.com/jbowtie/gokogiri" "github.com/pebbe/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 ( x = util.CheckErr w = util.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, "") return s[i1:i2] } func main() { req, err := http.NewRequest("GET", "https://gemeente.groningen.nl/feed/rss/nieuws", nil) x(err) req.Header.Set("User-Agent", agent) client := &http.Client{} resp, err := client.Do(req) x(err) body, err := io.ReadAll(resp.Body) x(err) x(resp.Body.Close()) var rss Rss x(xml.Unmarshal(body, &rss)) if len(rss.Items) == 0 { x(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) } x(err) year, week := t.ISOWeek() dirname := fmt.Sprintf("/net/corpora/nlnieuws/GG/%d/%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 x(os.MkdirAll(dirname, 0777)) fp, err := os.Create(filename + ".xml") x(err) _, err = fp.WriteString("\n\n") x(err) _, err = fmt.Fprintf(fp, "%d", t.Unix()) x(err) _, err = fp.Write(item.Data) x(err) _, err = fp.WriteString("\n") x(err) x(fp.Close()) x(os.Chtimes(filename+".xml", t, t)) if !doArticle(filename, item.Link, item.Title, t, needUpdate) { x(os.Remove(filename + ".xml")) } } } func doArticle(filename string, url string, title string, timestamp time.Time, needUpdate bool) bool { if needUpdate { _ = os.Remove(filename + ".err") _ = os.Remove(filename + ".html") _ = os.Remove(filename + ".txt") _ = os.Remove(filename + ".skip") } else { if exists(filename+".txt") || exists(filename+".skip") { return true } } time.Sleep(2 * time.Second) req, err := http.NewRequest("GET", url, nil) x(err) req.Header.Set("User-Agent", agent) client := &http.Client{} resp, err := client.Do(req) x(err) body, err := io.ReadAll(resp.Body) x(err) x(resp.Body.Close()) doc, err := gokogiri.ParseHtml(body) x(err) root := doc.Root() ell, err := root.Search( `//div[contains(@class,"component-richtext")]/p` + ` | ` + `//div[contains(@class,"component-richtext")]/h2`) x(err) if len(ell) == 0 { _ = w(fmt.Errorf("empty: %s", url)) fp, err := os.Create(filename + ".err") x(err) _, err = fmt.Fprintf(fp, "empty: %s\n", url) x(err) x(fp.Close()) x(os.Chtimes(filename+".err", timestamp, timestamp)) fp, err = os.Create(filename + ".html") x(err) _, err = fp.Write(body) x(err) x(fp.Close()) x(os.Chtimes(filename+".html", timestamp, timestamp)) return false } fp, err := os.Create(filename + ".txt") x(err) _, err = fp.WriteString(addEnd(fixSpace(title))) x(err) for _, el := range ell { _, err = fp.WriteString(addEnd(fixSpace(el.Content()))) x(err) } x(fp.Close()) x(os.Chtimes(filename+".txt", timestamp, timestamp)) return true } func addEnd(s string) string { s = strings.TrimSpace(s) n := len(s) if n == 0 { return "" } if n > 0 { if strings.ContainsAny(s[n-1:], ".!?") { return s + "\n" } } if n > 1 { s2 := s[n-2:] if s2 == `."` || s2 == `!"` || s2 == `?"` || s2 == `.'` || s2 == `!'` || s2 == `?'` { return s + "\n" } } return s + ".\n" } func fixSpace(s string) string { return strings.Join(strings.Fields(s), " ") }