174 lines
3.6 KiB
Go
174 lines
3.6 KiB
Go
package main
|
|
|
|
import (
|
|
"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 {
|
|
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"
|
|
)
|
|
|
|
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>")
|
|
return s[i1:i2]
|
|
}
|
|
|
|
func main() {
|
|
resp, err := http.Get("https://feeds.nos.nl/nosnieuwsalgemeen")
|
|
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 {
|
|
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
|
|
}
|
|
}
|
|
x(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
|
|
|
|
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))
|
|
if !doArticle(filename, item.Link, t, needUpdate) {
|
|
x(os.Remove(filename + ".xml"))
|
|
}
|
|
}
|
|
}
|
|
|
|
func doArticle(filename string, url string, timestamp time.Time, needUpdate bool) bool {
|
|
if needUpdate {
|
|
_ = os.Remove(filename + ".err")
|
|
_ = os.Remove(filename + ".html")
|
|
_ = os.Remove(filename + ".skip")
|
|
_ = os.Remove(filename + ".json")
|
|
} else {
|
|
if exists(filename+".json") || 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())
|
|
|
|
s := string(body)
|
|
|
|
ok := true
|
|
i1 := strings.Index(s, `<script type="application/ld+json"`)
|
|
if i1 < 0 {
|
|
ok = false
|
|
} else {
|
|
i1 += strings.Index(s[i1:], `>`) + 1
|
|
i2 := i1 + strings.Index(s[i1:], `</script>`)
|
|
if i2 < i1 {
|
|
ok = false
|
|
} else {
|
|
s = s[i1:i2]
|
|
}
|
|
}
|
|
if !ok {
|
|
_ = w(fmt.Errorf("script jsonld not found: %s", url))
|
|
|
|
fp, err := os.Create(filename + ".err")
|
|
x(err)
|
|
_, err = fmt.Fprintf(fp, "script jsonld not found: %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 + ".json")
|
|
x(err)
|
|
_, err = fp.WriteString(s)
|
|
x(err)
|
|
x(fp.Close())
|
|
x(os.Chtimes(filename+".json", timestamp, timestamp))
|
|
return true
|
|
}
|