codeberg.org/pebbe/errors
This commit is contained in:
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -26,7 +26,7 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
p = e.PanicErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -36,22 +36,29 @@ func exists(filename string) bool {
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://rss.at5.nl/rss", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -59,7 +66,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/AT5/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -71,19 +78,23 @@ func main() {
|
||||
}
|
||||
filename := dirname + "/" + url.PathEscape(basename)
|
||||
|
||||
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))
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = true
|
||||
}()
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -16,7 +16,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
)
|
||||
@@ -51,18 +51,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"AT5\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"AT5\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
|
||||
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
|
||||
)
|
||||
@@ -55,18 +55,15 @@ func main() {
|
||||
x(err)
|
||||
var item Item
|
||||
x(xml.Unmarshal(b, &item), filename)
|
||||
_, err = fp.WriteString(addEnd(fixSpace(item.Title)))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(item.Title))))
|
||||
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
|
||||
x(err)
|
||||
root := doc.Root()
|
||||
pp, err := root.Search(`//body/p | //body/h2`)
|
||||
x(err)
|
||||
for _, p := range pp {
|
||||
_, err = fp.WriteString(addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(p.Content()))))
|
||||
}
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
}
|
||||
}
|
||||
|
||||
@@ -53,8 +53,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -29,7 +29,7 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
p = e.PanicErr
|
||||
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"
|
||||
)
|
||||
@@ -40,22 +40,29 @@ func exists(filename string) bool {
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://www.amsterdam.nl/nieuws/nieuwsoverzicht/?rss=true", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -63,26 +70,31 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(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)
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
doArticle(filename, item.Link, item.Title, t)
|
||||
ok = true
|
||||
}()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -93,58 +105,54 @@ func doArticle(filename string, url string, title string, timestamp time.Time) {
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
doc, err := gokogiri.ParseHtml(body)
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
root := doc.Root()
|
||||
|
||||
fp, err := os.Create(filename + ".txt")
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
_, err = fp.WriteString(addEnd(title))
|
||||
x(err)
|
||||
p(fp.WriteString(addEnd(title)))
|
||||
|
||||
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)
|
||||
p(err)
|
||||
for _, p1 := range pp {
|
||||
p(fp.WriteString(addEnd(p1.Content())))
|
||||
count++
|
||||
}
|
||||
|
||||
ell, err := root.Search(`//div[@id="zone_content"]//div[contains(@class, "tekst")]/child::*`)
|
||||
x(err)
|
||||
p(err)
|
||||
for _, el := range ell {
|
||||
if n := el.Name(); n == "p" || n == "h3" {
|
||||
_, err = fp.WriteString(addEnd(el.Content()))
|
||||
p(fp.WriteString(addEnd(el.Content())))
|
||||
count++
|
||||
x(err)
|
||||
}
|
||||
}
|
||||
|
||||
x(fp.Close())
|
||||
p(fp.Close())
|
||||
|
||||
x(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".debug.html", timestamp, timestamp))
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
108
GG/cmd/gg/gg.go
108
GG/cmd/gg/gg.go
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -29,8 +29,8 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
w = util.WarnErr
|
||||
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"
|
||||
)
|
||||
@@ -48,26 +48,36 @@ func fileDate(filename string) string {
|
||||
s := string(b)
|
||||
i1 := strings.Index(s, "<unixTime>") + 10
|
||||
i2 := strings.Index(s, "</unixTime>")
|
||||
if i2 < i1 {
|
||||
return ""
|
||||
}
|
||||
return s[i1:i2]
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://gemeente.groningen.nl/feed/rss/nieuws", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -75,7 +85,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/GG/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -86,26 +96,28 @@ func main() {
|
||||
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, item.Title, t, needUpdate) {
|
||||
x(os.Remove(filename + ".xml"))
|
||||
}
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\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) bool {
|
||||
func doArticle(filename string, url string, title string, timestamp time.Time, needUpdate bool) (ok bool) {
|
||||
if needUpdate {
|
||||
_ = os.Remove(filename + ".err")
|
||||
_ = os.Remove(filename + ".html")
|
||||
@@ -119,18 +131,18 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
doc, err := gokogiri.ParseHtml(body)
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
root := doc.Root()
|
||||
|
||||
@@ -138,41 +150,37 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
`//div[contains(@class,"component-richtext")]/p` +
|
||||
` | ` +
|
||||
`//div[contains(@class,"component-richtext")]/h2`)
|
||||
x(err)
|
||||
p(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))
|
||||
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")
|
||||
x(err)
|
||||
_, err = fp.Write(body)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
|
||||
return false
|
||||
}
|
||||
|
||||
fp, err := os.Create(filename + ".txt")
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
_, err = fp.WriteString(addEnd(fixSpace(title)))
|
||||
x(err)
|
||||
p(fp.WriteString(addEnd(fixSpace(title))))
|
||||
|
||||
for _, el := range ell {
|
||||
_, err = fp.WriteString(addEnd(fixSpace(el.Content())))
|
||||
x(err)
|
||||
p(fp.WriteString(addEnd(fixSpace(el.Content()))))
|
||||
}
|
||||
|
||||
x(fp.Close())
|
||||
p(fp.Close())
|
||||
|
||||
x(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
p(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
|
||||
return true
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -16,7 +16,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
)
|
||||
@@ -51,18 +51,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Gemeente Groningen\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Gemeente Groningen\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -51,8 +51,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/json"
|
||||
"fmt"
|
||||
@@ -19,7 +19,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
|
||||
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
|
||||
)
|
||||
@@ -57,17 +57,13 @@ func main() {
|
||||
var item Item
|
||||
x(json.Unmarshal(b, &item))
|
||||
for _, cat := range item.Cats {
|
||||
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat))
|
||||
x(err)
|
||||
x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
|
||||
}
|
||||
for _, tag := range item.Tags {
|
||||
_, err = fmt.Fprintf(fp, "##META text tag = %s\n", fixSpace(tag))
|
||||
x(err)
|
||||
x(fmt.Fprintf(fp, "##META text tag = %s\n", fixSpace(tag)))
|
||||
}
|
||||
_, err = fp.WriteString(addEnd(fixSpace(item.Title)))
|
||||
x(err)
|
||||
_, err = fp.WriteString(fixSpace(item.Text))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(item.Title))))
|
||||
x(fp.WriteString(fixSpace(item.Text)))
|
||||
x(fp.Close())
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -66,18 +66,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NOS\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NOS\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -27,8 +27,8 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
w = util.WarnErr
|
||||
p = e.PanicErr
|
||||
w = e.WarnErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -45,21 +45,31 @@ func fileDate(filename string) string {
|
||||
s := string(b)
|
||||
i1 := strings.Index(s, "<unixTime>") + 10
|
||||
i2 := strings.Index(s, "</unixTime>")
|
||||
if i2 < i1 {
|
||||
return ""
|
||||
}
|
||||
return s[i1:i2]
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
resp, err := http.Get("https://feeds.nos.nl/nosnieuwsalgemeen")
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -74,7 +84,7 @@ func main() {
|
||||
break
|
||||
}
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/NOS/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -85,22 +95,24 @@ func main() {
|
||||
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"))
|
||||
}
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = doArticle(filename, item.Link, t, needUpdate)
|
||||
}()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -118,15 +130,15 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
s := string(body)
|
||||
|
||||
@@ -147,27 +159,24 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
|
||||
_ = 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))
|
||||
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")
|
||||
x(err)
|
||||
_, err = fp.Write(body)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.WriteString(s))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".json", timestamp, timestamp))
|
||||
return true
|
||||
}
|
||||
|
||||
@@ -53,8 +53,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -56,18 +56,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NU\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NU\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
124
NU/cmd/nu/nu.go
124
NU/cmd/nu/nu.go
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bytes"
|
||||
"encoding/json"
|
||||
@@ -39,8 +39,8 @@ type GItem struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
w = util.WarnErr
|
||||
p = e.PanicErr
|
||||
w = e.WarnErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -57,21 +57,31 @@ func fileDate(filename string) string {
|
||||
s := string(b)
|
||||
i1 := strings.Index(s, "<unixTime>") + 10
|
||||
i2 := strings.Index(s, "</unixTime>")
|
||||
if i2 < i1 {
|
||||
return ""
|
||||
}
|
||||
return s[i1:i2]
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
resp, err := http.Get("https://www.nu.nl/rss")
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -79,7 +89,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/NU/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -90,22 +100,24 @@ func main() {
|
||||
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"))
|
||||
}
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = doArticle(filename, item.Link, t, needUpdate)
|
||||
}()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -126,15 +138,15 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
s := string(body)
|
||||
ok := true
|
||||
@@ -154,38 +166,34 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
|
||||
_ = 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))
|
||||
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")
|
||||
x(err)
|
||||
_, err = fp.Write(body)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.WriteString(s))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".json", timestamp, timestamp))
|
||||
|
||||
var doc Doc
|
||||
if err = json.Unmarshal([]byte(s), &doc); err != nil {
|
||||
_ = w(err, url)
|
||||
|
||||
fp, err2 := os.Create(filename + ".err")
|
||||
x(err2)
|
||||
_, err2 = fmt.Fprintf(fp, "%s: %v\n%s\n", url, err, s)
|
||||
x(err2)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".err", timestamp, timestamp))
|
||||
p(err2)
|
||||
p(fmt.Fprintf(fp, "%s: %v\n%s\n", url, err, s))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".err", timestamp, timestamp))
|
||||
|
||||
return false
|
||||
}
|
||||
@@ -193,8 +201,7 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
|
||||
cats := make([]string, 0)
|
||||
var buffer bytes.Buffer
|
||||
for _, i := range doc.Graph {
|
||||
_, err = buffer.WriteString(html.UnescapeString(i.ArticleBody))
|
||||
x(err)
|
||||
p(buffer.WriteString(html.UnescapeString(i.ArticleBody)))
|
||||
cats = append(cats, i.ArticleSection...)
|
||||
}
|
||||
text := buffer.String()
|
||||
@@ -224,20 +231,17 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
|
||||
text = strings.Join(lines, "\n") + "\n"
|
||||
|
||||
fp, err := os.Create(filename + ".txt")
|
||||
x(err)
|
||||
p(err)
|
||||
if len(cats) == 0 {
|
||||
_, err := fmt.Fprintln(fp, "##META text cat =")
|
||||
x(err)
|
||||
p(fmt.Fprintln(fp, "##META text cat ="))
|
||||
} else {
|
||||
for _, cat := range cats {
|
||||
_, err := fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat))
|
||||
x(err)
|
||||
p(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
|
||||
}
|
||||
}
|
||||
_, err = fp.WriteString(text)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
p(fp.WriteString(text))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
}
|
||||
|
||||
return true
|
||||
|
||||
@@ -51,8 +51,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -56,18 +56,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NieuwsNL\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NieuwsNL\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"bytes"
|
||||
"encoding/xml"
|
||||
@@ -30,8 +30,8 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
w = util.WarnErr
|
||||
p = e.PanicErr
|
||||
w = e.WarnErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -48,26 +48,36 @@ func fileDate(filename string) string {
|
||||
s := string(b)
|
||||
i1 := strings.Index(s, "<unixTime>") + 10
|
||||
i2 := strings.Index(s, "</unixTime>")
|
||||
if i2 < i1 {
|
||||
return ""
|
||||
}
|
||||
return s[i1:i2]
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://nieuws.nl/sitemap/news.xml", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -75,7 +85,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/NieuwsNL/%d/%02d/%02d", t.Year(), int(t.Month()), t.Day())
|
||||
if exists(dirname + "/lock") {
|
||||
continue
|
||||
@@ -85,22 +95,24 @@ func main() {
|
||||
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, item.Title, t, needUpdate) {
|
||||
x(os.Remove(filename + ".xml"))
|
||||
}
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = doArticle(filename, item.Link, item.Title, t, needUpdate)
|
||||
}()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -118,18 +130,18 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
doc, err := gokogiri.ParseHtml(body)
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
var buf bytes.Buffer
|
||||
fouten := make([]string, 0)
|
||||
@@ -138,10 +150,9 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
|
||||
var cat string
|
||||
aa, err := root.Search(`//a[contains(@class, "articleHeader__info__category")]`)
|
||||
x(err)
|
||||
p(err)
|
||||
if len(aa) == 0 {
|
||||
_, err = fmt.Fprintln(&buf, "##META text cat =")
|
||||
x(err)
|
||||
p(fmt.Fprintln(&buf, "##META text cat ="))
|
||||
_ = w(fmt.Errorf("no cat: %s", url))
|
||||
// geen fout, maar waarschuwing als er meer fouten zijn
|
||||
fouten = append(fouten, fmt.Sprintf("no text: %s\n", url))
|
||||
@@ -149,52 +160,46 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
} else {
|
||||
for _, a := range aa {
|
||||
cat = strings.ReplaceAll(a.Content(), "\n", " ")
|
||||
_, err = fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(cat))
|
||||
x(err)
|
||||
p(fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(cat)))
|
||||
}
|
||||
}
|
||||
|
||||
_, err = buf.WriteString(addEnd(fixSpace(title)))
|
||||
x(err)
|
||||
p(buf.WriteString(addEnd(fixSpace(title))))
|
||||
|
||||
// oud: //div[@id="article-blocks"]//p
|
||||
pp, err := root.Search(`//div[@id="article-blocks"]//div[contains(@class, "paragraph-content")]`)
|
||||
x(err)
|
||||
p(err)
|
||||
if len(pp) == 0 {
|
||||
_ = w(fmt.Errorf("empty: %s", url))
|
||||
// dit is echt fout
|
||||
fouten = append(fouten, fmt.Sprintf("empty: %s\n", url))
|
||||
|
||||
fp, err := os.Create(filename + ".err")
|
||||
x(err)
|
||||
p(err)
|
||||
for _, fout := range fouten {
|
||||
_, err = fp.WriteString(fout)
|
||||
x(err)
|
||||
p(fp.WriteString(fout))
|
||||
}
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".err", timestamp, timestamp))
|
||||
p(fp.Close())
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
|
||||
return false // echt fout
|
||||
}
|
||||
for _, p := range pp {
|
||||
_, err = buf.WriteString(addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
for _, p1 := range pp {
|
||||
p(buf.WriteString(addEnd(fixSpace(p1.Content()))))
|
||||
}
|
||||
|
||||
fp, err := os.Create(filename + ".txt")
|
||||
x(err)
|
||||
_, err = fp.Write(buf.Bytes())
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
p(err)
|
||||
p(fp.Write(buf.Bytes()))
|
||||
p(fp.Close())
|
||||
|
||||
x(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
p(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
|
||||
return true
|
||||
}
|
||||
|
||||
@@ -53,8 +53,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../../metadata 2> err
|
||||
rm err
|
||||
../../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -66,18 +66,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Reporters Online\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Reporters Online\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -26,7 +26,7 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
p = e.PanicErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -36,22 +36,29 @@ func exists(filename string) bool {
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://reportersonline.nl/feed/", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -59,7 +66,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/RO/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -71,19 +78,24 @@ func main() {
|
||||
}
|
||||
filename := dirname + "/" + url.PathEscape(basename)
|
||||
|
||||
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))
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = true
|
||||
}()
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -19,7 +19,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
|
||||
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
|
||||
)
|
||||
@@ -57,11 +57,9 @@ func main() {
|
||||
var item Item
|
||||
x(xml.Unmarshal(b, &item))
|
||||
for _, cat := range item.Cats {
|
||||
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat))
|
||||
x(err)
|
||||
x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
|
||||
}
|
||||
_, err = fp.WriteString(addEnd(fixSpace(item.Title)))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(item.Title))))
|
||||
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
|
||||
x(err)
|
||||
root := doc.Root()
|
||||
@@ -73,10 +71,8 @@ func main() {
|
||||
pp, err := root.Search(`//body//p[not(.//a[contains(@href,"reportersonline.nl/support")])]`)
|
||||
x(err)
|
||||
for _, p := range pp {
|
||||
_, err = fp.WriteString(addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(p.Content()))))
|
||||
}
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
}
|
||||
}
|
||||
|
||||
@@ -53,8 +53,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -66,18 +66,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sargasso\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sargasso\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -26,7 +26,7 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
p = e.PanicErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -36,22 +36,29 @@ func exists(filename string) bool {
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://sargasso.nl/feed/", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -59,7 +66,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/Sargasso/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -71,19 +78,24 @@ func main() {
|
||||
}
|
||||
filename := dirname + "/" + url.PathEscape(basename)
|
||||
|
||||
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))
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</)item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = true
|
||||
}()
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -19,7 +19,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
|
||||
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
|
||||
)
|
||||
@@ -57,21 +57,17 @@ func main() {
|
||||
var item Item
|
||||
x(xml.Unmarshal(b, &item))
|
||||
for _, cat := range item.Cats {
|
||||
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat))
|
||||
x(err)
|
||||
x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
|
||||
}
|
||||
_, err = fp.WriteString(addEnd(fixSpace(item.Title)))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(item.Title))))
|
||||
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
|
||||
x(err)
|
||||
root := doc.Root()
|
||||
pp, err := root.Search(`//body//p`)
|
||||
x(err)
|
||||
for _, p := range pp {
|
||||
_, err = fp.WriteString(addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(p.Content()))))
|
||||
}
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
}
|
||||
}
|
||||
|
||||
@@ -53,8 +53,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -16,7 +16,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
)
|
||||
@@ -51,18 +51,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sikkom\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sikkom\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -29,8 +29,8 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
w = util.WarnErr
|
||||
p = e.PanicErr
|
||||
w = e.WarnErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -47,21 +47,31 @@ func fileDate(filename string) string {
|
||||
s := string(b)
|
||||
i1 := strings.Index(s, "<unixTime>") + 10
|
||||
i2 := strings.Index(s, "</unixTime>")
|
||||
if i2 < i1 {
|
||||
return ""
|
||||
}
|
||||
return s[i1:i2]
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
resp, err := http.Get("https://www.sikkom.nl/api/feed/rss")
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -69,7 +79,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/Sikkom/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -80,22 +90,24 @@ func main() {
|
||||
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, item.Title, t, needUpdate) {
|
||||
x(os.Remove(filename + ".xml"))
|
||||
}
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = doArticle(filename, item.Link, item.Title, t, needUpdate)
|
||||
}()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -115,15 +127,15 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
s := string(body)
|
||||
|
||||
@@ -144,69 +156,63 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
|
||||
_ = 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))
|
||||
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")
|
||||
x(err)
|
||||
_, err = fp.Write(body)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.WriteString(s))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".json", timestamp, timestamp))
|
||||
|
||||
doc, err := gokogiri.ParseHtml(body)
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
root := doc.Root()
|
||||
|
||||
pp, err := root.Search(`//div[contains(@class,"article-page__body")]//p`)
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
if len(pp) == 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))
|
||||
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")
|
||||
x(err)
|
||||
_, err = fp.Write(body)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(err)
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
|
||||
return false
|
||||
}
|
||||
|
||||
fp, err = os.Create(filename + ".txt")
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
_, err = fp.WriteString(addEnd(fixSpace(title)))
|
||||
x(err)
|
||||
p(fp.WriteString(addEnd(fixSpace(title))))
|
||||
|
||||
for _, p := range pp {
|
||||
_, err = fp.WriteString(addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
for _, p1 := range pp {
|
||||
p(fp.WriteString(addEnd(fixSpace(p1.Content()))))
|
||||
}
|
||||
|
||||
x(fp.Close())
|
||||
p(fp.Close())
|
||||
return true
|
||||
}
|
||||
|
||||
|
||||
@@ -51,8 +51,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -66,18 +66,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Tzum\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Tzum\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -26,7 +26,7 @@ type ItemT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
p = e.PanicErr
|
||||
agent = "AhrefsBot/7.0"
|
||||
)
|
||||
|
||||
@@ -36,22 +36,29 @@ func exists(filename string) bool {
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
req, err := http.NewRequest("GET", "https://www.tzum.info/feed/", nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -59,7 +66,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.PubDate)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
year, week := t.ISOWeek()
|
||||
dirname := fmt.Sprintf("/net/corpora/nlnieuws/Tzum/%d/%02d", year, week)
|
||||
if exists(dirname + "/lock") {
|
||||
@@ -71,19 +78,24 @@ func main() {
|
||||
}
|
||||
filename := dirname + "/" + url.PathEscape(basename)
|
||||
|
||||
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))
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
ok = true
|
||||
}()
|
||||
}
|
||||
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"encoding/xml"
|
||||
"fmt"
|
||||
@@ -19,7 +19,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
|
||||
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
|
||||
)
|
||||
@@ -57,11 +57,9 @@ func main() {
|
||||
var item Item
|
||||
x(xml.Unmarshal(b, &item))
|
||||
for _, cat := range item.Cats {
|
||||
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat))
|
||||
x(err)
|
||||
x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
|
||||
}
|
||||
_, err = fp.WriteString(addEnd(fixSpace(item.Title)))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(item.Title))))
|
||||
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
|
||||
x(err)
|
||||
root := doc.Root()
|
||||
@@ -70,11 +68,9 @@ func main() {
|
||||
for _, p := range pp {
|
||||
s := p.Content()
|
||||
if !strings.Contains(s, "verscheen eerst op Tzum.") {
|
||||
_, err = fp.WriteString(addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
x(fp.WriteString(addEnd(fixSpace(p.Content()))))
|
||||
}
|
||||
}
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
}
|
||||
}
|
||||
|
||||
@@ -53,8 +53,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -1,7 +1,7 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"bufio"
|
||||
"encoding/xml"
|
||||
@@ -18,7 +18,7 @@ type Item struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
escape = html.EscapeString
|
||||
data = make(map[string][]string)
|
||||
location *time.Location
|
||||
@@ -56,18 +56,13 @@ func main() {
|
||||
i += strings.Index(s[i:], "<")
|
||||
fp, err := os.Create("xml/" + filename + ".tmp")
|
||||
x(err)
|
||||
_, err = fp.WriteString(s[:i])
|
||||
x(err)
|
||||
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"VRT\"/>\n")
|
||||
x(err)
|
||||
x(fp.WriteString(s[:i]))
|
||||
x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"VRT\"/>\n"))
|
||||
for _, m := range data[base] {
|
||||
_, err = fp.WriteString(" " + m + "\n")
|
||||
x(err)
|
||||
x(fp.WriteString(" " + m + "\n"))
|
||||
}
|
||||
_, err = fp.WriteString(" </metadata>\n ")
|
||||
x(err)
|
||||
_, err = fp.WriteString(stripMeta(s[i:]))
|
||||
x(err)
|
||||
x(fp.WriteString(" </metadata>\n "))
|
||||
x(fp.WriteString(stripMeta(s[i:])))
|
||||
x(fp.Close())
|
||||
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
|
||||
}
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
package main
|
||||
|
||||
import (
|
||||
e "codeberg.org/pebbe/errors"
|
||||
"github.com/jbowtie/gokogiri"
|
||||
"github.com/pebbe/util"
|
||||
|
||||
"bytes"
|
||||
"encoding/xml"
|
||||
@@ -43,8 +43,8 @@ type LinkT struct {
|
||||
}
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
w = util.WarnErr
|
||||
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"
|
||||
)
|
||||
@@ -62,21 +62,31 @@ func fileDate(filename string) string {
|
||||
s := string(b)
|
||||
i1 := strings.Index(s, "<unixTime>") + 10
|
||||
i2 := strings.Index(s, "</unixTime>")
|
||||
if i2 < i1 {
|
||||
return ""
|
||||
}
|
||||
return s[i1:i2]
|
||||
}
|
||||
|
||||
func main() {
|
||||
defer func() {
|
||||
if e.Panicked() {
|
||||
_ = recover()
|
||||
os.Exit(1)
|
||||
}
|
||||
}()
|
||||
|
||||
resp, err := http.Get("https://www.vrt.be/vrtnws/nl.rss.headlines.xml")
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
var rss Rss
|
||||
x(xml.Unmarshal(body, &rss))
|
||||
p(xml.Unmarshal(body, &rss))
|
||||
|
||||
if len(rss.Items) == 0 {
|
||||
x(fmt.Errorf("len(rss.Items) == 0"))
|
||||
p(fmt.Errorf("len(rss.Items) == 0"))
|
||||
}
|
||||
|
||||
for _, item := range rss.Items {
|
||||
@@ -84,7 +94,7 @@ func main() {
|
||||
if err != nil {
|
||||
t, err = time.Parse(time.RFC1123, item.Published)
|
||||
}
|
||||
x(err)
|
||||
p(err)
|
||||
t2, err := time.Parse(time.RFC3339Nano, item.Updated)
|
||||
if err != nil {
|
||||
t2, _ = time.Parse(time.RFC1123, item.Updated)
|
||||
@@ -102,29 +112,31 @@ func main() {
|
||||
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))
|
||||
p(os.MkdirAll(dirname, 0777))
|
||||
func() {
|
||||
var ok bool
|
||||
defer func() {
|
||||
if !ok {
|
||||
_ = os.Remove(filename + ".xml")
|
||||
}
|
||||
}()
|
||||
fp, err := os.Create(filename + ".xml")
|
||||
p(err)
|
||||
p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
|
||||
p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
|
||||
p(fp.Write(item.Data))
|
||||
p(fp.WriteString("</item>\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".xml", t, t))
|
||||
|
||||
var link string
|
||||
for _, l := range item.Link {
|
||||
if l.Type == "text/html" {
|
||||
link = l.Href
|
||||
var link string
|
||||
for _, l := range item.Link {
|
||||
if l.Type == "text/html" {
|
||||
link = l.Href
|
||||
}
|
||||
}
|
||||
}
|
||||
if !doArticle(filename, link, item.Title.Text, item.Nstag, item.Nslabeltag, t, needUpdate) {
|
||||
x(os.Remove(filename + ".xml"))
|
||||
}
|
||||
ok = doArticle(filename, link, item.Title.Text, item.Nstag, item.Nslabeltag, t, needUpdate)
|
||||
}()
|
||||
}
|
||||
}
|
||||
|
||||
@@ -142,15 +154,15 @@ func doArticle(filename string, url string, title string, tags []string, labels
|
||||
time.Sleep(2 * time.Second)
|
||||
|
||||
req, err := http.NewRequest("GET", url, nil)
|
||||
x(err)
|
||||
p(err)
|
||||
req.Header.Set("User-Agent", agent)
|
||||
|
||||
client := &http.Client{}
|
||||
resp, err := client.Do(req)
|
||||
x(err)
|
||||
p(err)
|
||||
body, err := io.ReadAll(resp.Body)
|
||||
x(err)
|
||||
x(resp.Body.Close())
|
||||
p(err)
|
||||
p(resp.Body.Close())
|
||||
|
||||
/*
|
||||
s := string(body)
|
||||
@@ -171,80 +183,71 @@ func doArticle(filename string, url string, title string, tags []string, labels
|
||||
_ = 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))
|
||||
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")
|
||||
x(err)
|
||||
_, err = fp.Write(body)
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.WriteString(s))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".json", timestamp, timestamp))
|
||||
*/
|
||||
|
||||
var buf bytes.Buffer
|
||||
|
||||
doc, err := gokogiri.ParseHtml(body)
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
root := doc.Root()
|
||||
|
||||
lnn, err := root.Search(`//head/link[@rel="canonical"]/@href`)
|
||||
x(err)
|
||||
p(err)
|
||||
for _, ln := range lnn {
|
||||
if strings.Contains(ln.String(), "/liveblog/") {
|
||||
fp, err := os.Create(filename + ".skip")
|
||||
x(err)
|
||||
_, err = fp.WriteString("liveblog\n")
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".skip", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.WriteString("liveblog\n"))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".skip", timestamp, timestamp))
|
||||
return true
|
||||
}
|
||||
}
|
||||
|
||||
if len(tags) == 0 {
|
||||
_, err = fmt.Fprintln(&buf, "##META text cat =")
|
||||
x(err)
|
||||
p(fmt.Fprintln(&buf, "##META text cat ="))
|
||||
} else {
|
||||
for _, tag := range tags {
|
||||
_, err = fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(tag))
|
||||
x(err)
|
||||
p(fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(tag)))
|
||||
}
|
||||
}
|
||||
if len(labels) == 0 {
|
||||
_, err = fmt.Fprintln(&buf, "##META text label =")
|
||||
x(err)
|
||||
p(fmt.Fprintln(&buf, "##META text label ="))
|
||||
} else {
|
||||
for _, label := range labels {
|
||||
_, err = fmt.Fprintf(&buf, "##META text label = %s\n", fixSpace(label))
|
||||
x(err)
|
||||
p(fmt.Fprintf(&buf, "##META text label = %s\n", fixSpace(label)))
|
||||
}
|
||||
}
|
||||
|
||||
_, err = buf.WriteString(addEnd(fixSpace(title)))
|
||||
x(err)
|
||||
p(err)
|
||||
|
||||
fouten := make([]string, 0)
|
||||
|
||||
found := false
|
||||
pp, err := root.Search(`//div[@data-sentry-component="ArticleHeading"]//*[contains(@class,"prose-article-body-r")]`)
|
||||
x(err)
|
||||
for _, p := range pp {
|
||||
_, err = fmt.Fprint(&buf, addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
p(err)
|
||||
for _, p1 := range pp {
|
||||
p(fmt.Fprint(&buf, addEnd(fixSpace(p1.Content()))))
|
||||
found = true
|
||||
}
|
||||
if !found {
|
||||
@@ -257,10 +260,9 @@ func doArticle(filename string, url string, title string, tags []string, labels
|
||||
`//div[@data-sentry-component="ArticleText"]//*[contains(@class,"prose-article-body-r")]` +
|
||||
` | ` +
|
||||
`//div[@data-sentry-component="ArticleTitle"]//h2`)
|
||||
x(err)
|
||||
for _, p := range pp {
|
||||
_, err = fmt.Fprint(&buf, addEnd(fixSpace(p.Content())))
|
||||
x(err)
|
||||
p(err)
|
||||
for _, p1 := range pp {
|
||||
p(fmt.Fprint(&buf, addEnd(fixSpace(p1.Content()))))
|
||||
found = true
|
||||
}
|
||||
if !found {
|
||||
@@ -270,31 +272,27 @@ func doArticle(filename string, url string, title string, tags []string, labels
|
||||
|
||||
if len(fouten) > 0 {
|
||||
fp, err := os.Create(filename + ".err")
|
||||
x(err)
|
||||
p(err)
|
||||
for _, fout := range fouten {
|
||||
_, err = fp.WriteString(fout)
|
||||
x(err)
|
||||
p(fp.WriteString(fout))
|
||||
}
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".err", timestamp, timestamp))
|
||||
p(fp.Close())
|
||||
p(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))
|
||||
p(err)
|
||||
p(fp.Write(body))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".html", timestamp, timestamp))
|
||||
|
||||
return false
|
||||
}
|
||||
|
||||
fp, err := os.Create(filename + ".txt")
|
||||
x(err)
|
||||
_, err = fp.Write(buf.Bytes())
|
||||
x(err)
|
||||
x(fp.Close())
|
||||
x(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
p(err)
|
||||
p(fp.Write(buf.Bytes()))
|
||||
p(fp.Close())
|
||||
p(os.Chtimes(filename+".txt", timestamp, timestamp))
|
||||
|
||||
return true
|
||||
}
|
||||
|
||||
@@ -51,8 +51,7 @@ cd out
|
||||
mkdir xml
|
||||
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
|
||||
|
||||
../../../metadata 2> err
|
||||
rm err
|
||||
../../../metadata
|
||||
|
||||
cd xml
|
||||
alto -o $corpus.data.dz *.xml 2> /dev/null
|
||||
|
||||
@@ -23,7 +23,7 @@ Dit programma geeft wel de juiste uitvoer.
|
||||
*/
|
||||
|
||||
import (
|
||||
"github.com/pebbe/util"
|
||||
e "codeberg.org/pebbe/errors"
|
||||
|
||||
"fmt"
|
||||
"os"
|
||||
@@ -32,7 +32,7 @@ import (
|
||||
)
|
||||
|
||||
var (
|
||||
x = util.CheckErr
|
||||
x = e.ExitErr
|
||||
)
|
||||
|
||||
func main() {
|
||||
|
||||
9
go.mod
9
go.mod
@@ -1,8 +1,7 @@
|
||||
module nlnieuws
|
||||
|
||||
go 1.25.0
|
||||
go 1.26.1
|
||||
|
||||
require (
|
||||
github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5
|
||||
github.com/pebbe/util v0.10.0
|
||||
)
|
||||
require github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5
|
||||
|
||||
require codeberg.org/pebbe/errors v0.3.0
|
||||
|
||||
4
go.sum
4
go.sum
@@ -1,4 +1,4 @@
|
||||
codeberg.org/pebbe/errors v0.3.0 h1:031dKFUvGzXxsb+ig7cKNpohHeQ38ghXDZrd7KehdbU=
|
||||
codeberg.org/pebbe/errors v0.3.0/go.mod h1:O7PPxUJM1bWRHq11CRK3wqVaH/3NnRaSVZvh3UhzDCY=
|
||||
github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5 h1:tQbR4RKFBFi0+Ll69dXejKKUbQVNaOAT2fjlDvSAfx4=
|
||||
github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5/go.mod h1:kQE2lxPgVKe0JsBZMFFfMm5zBDCuRhaHFKOBzZeCLiw=
|
||||
github.com/pebbe/util v0.10.0 h1:6GAxH8fo6HGOUq+JcPVsMuyvLS2mPwkrAx3/DCoYy1Y=
|
||||
github.com/pebbe/util v0.10.0/go.mod h1:ynWl/SFX4+Seb9fpjVlYevr1f4TP7FrCmyZHiBCg69Q=
|
||||
|
||||
13
namen.sh
13
namen.sh
@@ -15,6 +15,7 @@ overige opties:
|
||||
-n int : max aantal resultaten
|
||||
-p : gebruik pager
|
||||
-s : tel hits één keer per bericht
|
||||
-v : verbose
|
||||
"
|
||||
exit
|
||||
}
|
||||
@@ -23,7 +24,8 @@ SINGLE=0
|
||||
SELECT=0
|
||||
USEPAGER=0
|
||||
LIMIT=0
|
||||
while getopts 'sin:p' opt
|
||||
VERBOSE=0
|
||||
while getopts 'sin:pv' opt
|
||||
do
|
||||
case "$opt" in
|
||||
i)
|
||||
@@ -38,6 +40,9 @@ do
|
||||
s)
|
||||
SINGLE=1
|
||||
;;
|
||||
v)
|
||||
VERBOSE=1
|
||||
;;
|
||||
*)
|
||||
usage
|
||||
;;
|
||||
@@ -66,7 +71,11 @@ else
|
||||
fi
|
||||
|
||||
search () {
|
||||
eval "alto $@ '$EXPR' '$TEMPLATE' | $SORT | uniq -c | sort -nr 2> /dev/null $TAIL"
|
||||
if [ $VERBOSE = 1 ]
|
||||
then
|
||||
echo "alto $@ '$EXPR' '$TEMPLATE' | $SORT | uniq -c | sort -nr 2> /dev/null$TAIL"
|
||||
fi
|
||||
eval "alto $@ '$EXPR' '$TEMPLATE' | $SORT | uniq -c | sort -nr 2> /dev/null$TAIL"
|
||||
}
|
||||
|
||||
for i in "$@"
|
||||
|
||||
Reference in New Issue
Block a user