ISODate -> ISOWeek; exists(dirname + "/lock")

Amsterdam/Makefile

@@ -0,0 +1,4 @@
+all: amsterdam
+
+% : %.go
+	go build $<

Amsterdam/amsterdam.go

@@ -0,0 +1,169 @@
+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
+	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 main() {
+	req, err := http.NewRequest("GET", "https://www.amsterdam.nl/nieuws/nieuwsoverzicht/?rss=true", 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)
+		dirname := fmt.Sprintf("/net/corpora/nlnieuws/amsterdam/%d/%02d", t.Year(), int(t.Month()))
+		if exists(dirname + "/lock") {
+			continue
+		}
+		filename := dirname + "/" + url.PathEscape(strings.TrimPrefix(item.Guid, "https://www.amsterdam.nl/nieuws/"))
+		x(os.MkdirAll(dirname, 0777))
+		fp, err := os.Create(filename + ".xml")
+		x(err)
+		_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n")
+		x(err)
+		_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix())
+		x(err)
+		_, err = fp.Write(item.Data)
+		x(err)
+		_, err = fp.WriteString("</item>\n")
+		x(err)
+		x(fp.Close())
+		x(os.Chtimes(filename+".xml", t, t))
+		doArticle(filename, item.Link, item.Title, t)
+	}
+}
+
+func doArticle(filename string, url string, title string, timestamp time.Time) {
+	if exists(filename + ".txt") {
+		return
+	}
+	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()
+
+	fp, err := os.Create(filename + ".txt")
+	x(err)
+
+	_, err = fp.WriteString(addEnd(title))
+	x(err)
+
+	count := 0
+
+	pp, err := root.Search(`//div[@id="zone_intro"]//div[contains(@class, "inleiding")]/p`)
+	x(err)
+	for _, p := range pp {
+		_, err = fp.WriteString(addEnd(p.Content()))
+		x(err)
+		count++
+	}
+
+	ell, err := root.Search(`//div[@id="zone_content"]//div[contains(@class, "tekst")]/child::*`)
+	x(err)
+	for _, el := range ell {
+		if n := el.Name(); n == "p" || n == "h3" {
+			_, err = fp.WriteString(addEnd(el.Content()))
+			count++
+			x(err)
+		}
+	}
+
+	x(fp.Close())
+
+	x(os.Chtimes(filename+".txt", timestamp, timestamp))
+
+	if count == 0 {
+		fp, err := os.Create(filename + ".debug.html")
+		x(err)
+		_, err = fp.Write(body)
+		x(err)
+		x(fp.Close())
+		x(os.Chtimes(filename+".debug.html", timestamp, timestamp))
+	}
+}
+
+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"
+}