p209327/nlnieuws
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

codeberg.org/pebbe/errors

Browse Source
This commit is contained in:
Peter Kleiweg
2026-03-16 13:41:03 +01:00
parent 7b18c51567
commit 78dc580c8d
40 changed files with 706 additions and 692 deletions
Download Patch File Download Diff File Expand all files Collapse all files

57
AT5/cmd/at5/at5.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -26,7 +26,7 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -36,22 +36,29 @@ func exists(filename string) bool {
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
req, err := http.NewRequest("GET", "https://rss.at5.nl/rss", nil) req, err := http.NewRequest("GET", "https://rss.at5.nl/rss", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -59,7 +66,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/AT5/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/AT5/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -71,19 +78,23 @@ func main() {
} }
filename := dirname + "/" + url.PathEscape(basename) filename := dirname + "/" + url.PathEscape(basename)
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
p(fp.WriteString("</item>\n"))
p(fp.Close())
p(os.Chtimes(filename+".xml", t, t))
ok = true
}()
} }
} }

19
AT5/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -16,7 +16,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
) )
@@ -51,18 +51,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"AT5\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"AT5\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

11
AT5/cmd/xml2txt/xml2txt.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`) reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
) )
@@ -55,18 +55,15 @@ func main() {
x(err) x(err)
var item Item var item Item
x(xml.Unmarshal(b, &item), filename) x(xml.Unmarshal(b, &item), filename)
_, err = fp.WriteString(addEnd(fixSpace(item.Title))) x(fp.WriteString(addEnd(fixSpace(item.Title))))
x(err)
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`)) doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
x(err) x(err)
root := doc.Root() root := doc.Root()
pp, err := root.Search(`//body/p | //body/h2`) pp, err := root.Search(`//body/p | //body/h2`)
x(err) x(err)
for _, p := range pp { for _, p := range pp {
_, err = fp.WriteString(addEnd(fixSpace(p.Content()))) x(fp.WriteString(addEnd(fixSpace(p.Content()))))
x(err)
} }
x(err)
x(fp.Close()) x(fp.Close())
} }
} }

3
AT5/txt2corpus.sh
View File

@@ -53,8 +53,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

98
Amsterdam/amsterdam.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -29,7 +29,7 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
agent = "AhrefsBot/7.0" 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" // 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() { 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) req, err := http.NewRequest("GET", "https://www.amsterdam.nl/nieuws/nieuwsoverzicht/?rss=true", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -63,26 +70,31 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) 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())) dirname := fmt.Sprintf("/net/corpora/nlnieuws/amsterdam/%d/%02d", t.Year(), int(t.Month()))
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
continue continue
} }
filename := dirname + "/" + url.PathEscape(strings.TrimPrefix(item.Guid, "https://www.amsterdam.nl/nieuws/")) filename := dirname + "/" + url.PathEscape(strings.TrimPrefix(item.Guid, "https://www.amsterdam.nl/nieuws/"))
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
doArticle(filename, item.Link, item.Title, t) 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) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
doc, err := gokogiri.ParseHtml(body) doc, err := gokogiri.ParseHtml(body)
x(err) p(err)
root := doc.Root() root := doc.Root()
fp, err := os.Create(filename + ".txt") fp, err := os.Create(filename + ".txt")
x(err) p(err)
_, err = fp.WriteString(addEnd(title)) p(fp.WriteString(addEnd(title)))
x(err)
count := 0 count := 0
pp, err := root.Search(`//div[@id="zone_intro"]//div[contains(@class, "inleiding")]/p`) pp, err := root.Search(`//div[@id="zone_intro"]//div[contains(@class, "inleiding")]/p`)
x(err) p(err)
for _, p := range pp { for _, p1 := range pp {
_, err = fp.WriteString(addEnd(p.Content())) p(fp.WriteString(addEnd(p1.Content())))
x(err)
count++ count++
} }
ell, err := root.Search(`//div[@id="zone_content"]//div[contains(@class, "tekst")]/child::*`) ell, err := root.Search(`//div[@id="zone_content"]//div[contains(@class, "tekst")]/child::*`)
x(err) p(err)
for _, el := range ell { for _, el := range ell {
if n := el.Name(); n == "p" || n == "h3" { if n := el.Name(); n == "p" || n == "h3" {
_, err = fp.WriteString(addEnd(el.Content())) p(fp.WriteString(addEnd(el.Content())))
count++ 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 { if count == 0 {
fp, err := os.Create(filename + ".debug.html") fp, err := os.Create(filename + ".debug.html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".debug.html", timestamp, timestamp))
x(os.Chtimes(filename+".debug.html", timestamp, timestamp))
} }
} }

108
GG/cmd/gg/gg.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -29,8 +29,8 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
w = util.WarnErr w = e.WarnErr
// agent = "AhrefsBot/7.0" // 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" 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) s := string(b)
i1 := strings.Index(s, "<unixTime>") + 10 i1 := strings.Index(s, "<unixTime>") + 10
i2 := strings.Index(s, "</unixTime>") i2 := strings.Index(s, "</unixTime>")
if i2 < i1 {
return ""
}
return s[i1:i2] return s[i1:i2]
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
req, err := http.NewRequest("GET", "https://gemeente.groningen.nl/feed/rss/nieuws", nil) req, err := http.NewRequest("GET", "https://gemeente.groningen.nl/feed/rss/nieuws", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -75,7 +85,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/GG/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/GG/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -86,26 +96,28 @@ func main() {
ts := fmt.Sprintf("%d", t.Unix()) ts := fmt.Sprintf("%d", t.Unix())
needUpdate := fileDate(filename+".xml") != ts needUpdate := fileDate(filename+".xml") != ts
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
if !doArticle(filename, item.Link, item.Title, t, needUpdate) { p(fp.WriteString("</item>\n"))
x(os.Remove(filename + ".xml")) 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 { if needUpdate {
_ = os.Remove(filename + ".err") _ = os.Remove(filename + ".err")
_ = os.Remove(filename + ".html") _ = 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) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
doc, err := gokogiri.ParseHtml(body) doc, err := gokogiri.ParseHtml(body)
x(err) p(err)
root := doc.Root() 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")]/p` +
` | ` + ` | ` +
`//div[contains(@class,"component-richtext")]/h2`) `//div[contains(@class,"component-richtext")]/h2`)
x(err) p(err)
if len(ell) == 0 { if len(ell) == 0 {
_ = w(fmt.Errorf("empty: %s", url)) _ = w(fmt.Errorf("empty: %s", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
_, err = fmt.Fprintf(fp, "empty: %s\n", url) p(fmt.Fprintf(fp, "empty: %s\n", url))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err := os.Create(filename + ".txt") fp, err := os.Create(filename + ".txt")
x(err) p(err)
_, err = fp.WriteString(addEnd(fixSpace(title))) p(fp.WriteString(addEnd(fixSpace(title))))
x(err)
for _, el := range ell { for _, el := range ell {
_, err = fp.WriteString(addEnd(fixSpace(el.Content()))) p(fp.WriteString(addEnd(fixSpace(el.Content()))))
x(err)
} }
x(fp.Close()) p(fp.Close())
x(os.Chtimes(filename+".txt", timestamp, timestamp)) p(os.Chtimes(filename+".txt", timestamp, timestamp))
return true return true
} }

19
GG/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -16,7 +16,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
) )
@@ -51,18 +51,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Gemeente Groningen\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Gemeente Groningen\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

3
GG/txt2corpus.sh
View File

@@ -51,8 +51,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

16
NOS/cmd/json2txt/json2txt.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/json" "encoding/json"
"fmt" "fmt"
@@ -19,7 +19,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`) reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
) )
@@ -57,17 +57,13 @@ func main() {
var item Item var item Item
x(json.Unmarshal(b, &item)) x(json.Unmarshal(b, &item))
for _, cat := range item.Cats { for _, cat := range item.Cats {
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)) x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
x(err)
} }
for _, tag := range item.Tags { for _, tag := range item.Tags {
_, err = fmt.Fprintf(fp, "##META text tag = %s\n", fixSpace(tag)) x(fmt.Fprintf(fp, "##META text tag = %s\n", fixSpace(tag)))
x(err)
} }
_, err = fp.WriteString(addEnd(fixSpace(item.Title))) x(fp.WriteString(addEnd(fixSpace(item.Title))))
x(err) x(fp.WriteString(fixSpace(item.Text)))
_, err = fp.WriteString(fixSpace(item.Text))
x(err)
x(fp.Close()) x(fp.Close())
} }
} }

19
NOS/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -66,18 +66,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NOS\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NOS\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

97
NOS/cmd/nos/nos.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -27,8 +27,8 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
w = util.WarnErr w = e.WarnErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -45,21 +45,31 @@ func fileDate(filename string) string {
s := string(b) s := string(b)
i1 := strings.Index(s, "<unixTime>") + 10 i1 := strings.Index(s, "<unixTime>") + 10
i2 := strings.Index(s, "</unixTime>") i2 := strings.Index(s, "</unixTime>")
if i2 < i1 {
return ""
}
return s[i1:i2] return s[i1:i2]
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
resp, err := http.Get("https://feeds.nos.nl/nosnieuwsalgemeen") resp, err := http.Get("https://feeds.nos.nl/nosnieuwsalgemeen")
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -74,7 +84,7 @@ func main() {
break break
} }
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/NOS/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/NOS/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -85,22 +95,24 @@ func main() {
ts := fmt.Sprintf("%d", t.Unix()) ts := fmt.Sprintf("%d", t.Unix())
needUpdate := fileDate(filename+".xml") != ts needUpdate := fileDate(filename+".xml") != ts
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
if !doArticle(filename, item.Link, t, needUpdate) { p(fp.WriteString("</item>\n"))
x(os.Remove(filename + ".xml")) 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) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
s := string(body) 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)) _ = w(fmt.Errorf("script jsonld not found: %s", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
_, err = fmt.Fprintf(fp, "script jsonld not found: %s\n", url) p(fmt.Fprintf(fp, "script jsonld not found: %s\n", url))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err := os.Create(filename + ".json") fp, err := os.Create(filename + ".json")
x(err) p(err)
_, err = fp.WriteString(s) p(fp.WriteString(s))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".json", timestamp, timestamp))
x(os.Chtimes(filename+".json", timestamp, timestamp))
return true return true
} }

3
NOS/txt2corpus.sh
View File

@@ -53,8 +53,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
NU/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -56,18 +56,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NU\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NU\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

124
NU/cmd/nu/nu.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bytes" "bytes"
"encoding/json" "encoding/json"
@@ -39,8 +39,8 @@ type GItem struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
w = util.WarnErr w = e.WarnErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -57,21 +57,31 @@ func fileDate(filename string) string {
s := string(b) s := string(b)
i1 := strings.Index(s, "<unixTime>") + 10 i1 := strings.Index(s, "<unixTime>") + 10
i2 := strings.Index(s, "</unixTime>") i2 := strings.Index(s, "</unixTime>")
if i2 < i1 {
return ""
}
return s[i1:i2] return s[i1:i2]
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
resp, err := http.Get("https://www.nu.nl/rss") resp, err := http.Get("https://www.nu.nl/rss")
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -79,7 +89,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/NU/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/NU/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -90,22 +100,24 @@ func main() {
ts := fmt.Sprintf("%d", t.Unix()) ts := fmt.Sprintf("%d", t.Unix())
needUpdate := fileDate(filename+".xml") != ts needUpdate := fileDate(filename+".xml") != ts
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
if !doArticle(filename, item.Link, t, needUpdate) { p(fp.WriteString("</item>\n"))
x(os.Remove(filename + ".xml")) 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) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
s := string(body) s := string(body)
ok := true 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)) _ = w(fmt.Errorf("script jsonld not found: %s", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
_, err = fmt.Fprintf(fp, "script jsonld not found: %s\n", url) p(fmt.Fprintf(fp, "script jsonld not found: %s\n", url))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err := os.Create(filename + ".json") fp, err := os.Create(filename + ".json")
x(err) p(err)
_, err = fp.WriteString(s) p(fp.WriteString(s))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".json", timestamp, timestamp))
x(os.Chtimes(filename+".json", timestamp, timestamp))
var doc Doc var doc Doc
if err = json.Unmarshal([]byte(s), &doc); err != nil { if err = json.Unmarshal([]byte(s), &doc); err != nil {
_ = w(err, url) _ = w(err, url)
fp, err2 := os.Create(filename + ".err") fp, err2 := os.Create(filename + ".err")
x(err2) p(err2)
_, err2 = fmt.Fprintf(fp, "%s: %v\n%s\n", url, err, s) p(fmt.Fprintf(fp, "%s: %v\n%s\n", url, err, s))
x(err2) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
return false return false
} }
@@ -193,8 +201,7 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
cats := make([]string, 0) cats := make([]string, 0)
var buffer bytes.Buffer var buffer bytes.Buffer
for _, i := range doc.Graph { for _, i := range doc.Graph {
_, err = buffer.WriteString(html.UnescapeString(i.ArticleBody)) p(buffer.WriteString(html.UnescapeString(i.ArticleBody)))
x(err)
cats = append(cats, i.ArticleSection...) cats = append(cats, i.ArticleSection...)
} }
text := buffer.String() text := buffer.String()
@@ -224,20 +231,17 @@ func doArticle(filename string, url string, timestamp time.Time, needUpdate bool
text = strings.Join(lines, "\n") + "\n" text = strings.Join(lines, "\n") + "\n"
fp, err := os.Create(filename + ".txt") fp, err := os.Create(filename + ".txt")
x(err) p(err)
if len(cats) == 0 { if len(cats) == 0 {
_, err := fmt.Fprintln(fp, "##META text cat =") p(fmt.Fprintln(fp, "##META text cat ="))
x(err)
} else { } else {
for _, cat := range cats { for _, cat := range cats {
_, err := fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)) p(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
x(err)
} }
} }
_, err = fp.WriteString(text) p(fp.WriteString(text))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".txt", timestamp, timestamp))
x(os.Chtimes(filename+".txt", timestamp, timestamp))
} }
return true return true

3
NU/txt2corpus.sh
View File

@@ -51,8 +51,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
NieuwsNL/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -56,18 +56,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NieuwsNL\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"NieuwsNL\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

119
NieuwsNL/cmd/nieuwsnl/nieuwsnl.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"bytes" "bytes"
"encoding/xml" "encoding/xml"
@@ -30,8 +30,8 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
w = util.WarnErr w = e.WarnErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -48,26 +48,36 @@ func fileDate(filename string) string {
s := string(b) s := string(b)
i1 := strings.Index(s, "<unixTime>") + 10 i1 := strings.Index(s, "<unixTime>") + 10
i2 := strings.Index(s, "</unixTime>") i2 := strings.Index(s, "</unixTime>")
if i2 < i1 {
return ""
}
return s[i1:i2] return s[i1:i2]
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
req, err := http.NewRequest("GET", "https://nieuws.nl/sitemap/news.xml", nil) req, err := http.NewRequest("GET", "https://nieuws.nl/sitemap/news.xml", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -75,7 +85,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) 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()) dirname := fmt.Sprintf("/net/corpora/nlnieuws/NieuwsNL/%d/%02d/%02d", t.Year(), int(t.Month()), t.Day())
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
continue continue
@@ -85,22 +95,24 @@ func main() {
ts := fmt.Sprintf("%d", t.Unix()) ts := fmt.Sprintf("%d", t.Unix())
needUpdate := fileDate(filename+".xml") != ts needUpdate := fileDate(filename+".xml") != ts
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
if !doArticle(filename, item.Link, item.Title, t, needUpdate) { p(fp.WriteString("</item>\n"))
x(os.Remove(filename + ".xml")) 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) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
doc, err := gokogiri.ParseHtml(body) doc, err := gokogiri.ParseHtml(body)
x(err) p(err)
var buf bytes.Buffer var buf bytes.Buffer
fouten := make([]string, 0) fouten := make([]string, 0)
@@ -138,10 +150,9 @@ func doArticle(filename string, url string, title string, timestamp time.Time, n
var cat string var cat string
aa, err := root.Search(`//a[contains(@class, "articleHeader__info__category")]`) aa, err := root.Search(`//a[contains(@class, "articleHeader__info__category")]`)
x(err) p(err)
if len(aa) == 0 { if len(aa) == 0 {
_, err = fmt.Fprintln(&buf, "##META text cat =") p(fmt.Fprintln(&buf, "##META text cat ="))
x(err)
_ = w(fmt.Errorf("no cat: %s", url)) _ = w(fmt.Errorf("no cat: %s", url))
// geen fout, maar waarschuwing als er meer fouten zijn // geen fout, maar waarschuwing als er meer fouten zijn
fouten = append(fouten, fmt.Sprintf("no text: %s\n", url)) 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 { } else {
for _, a := range aa { for _, a := range aa {
cat = strings.ReplaceAll(a.Content(), "\n", " ") cat = strings.ReplaceAll(a.Content(), "\n", " ")
_, err = fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(cat)) p(fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(cat)))
x(err)
} }
} }
_, err = buf.WriteString(addEnd(fixSpace(title))) p(buf.WriteString(addEnd(fixSpace(title))))
x(err)
// oud: //div[@id="article-blocks"]//p // oud: //div[@id="article-blocks"]//p
pp, err := root.Search(`//div[@id="article-blocks"]//div[contains(@class, "paragraph-content")]`) pp, err := root.Search(`//div[@id="article-blocks"]//div[contains(@class, "paragraph-content")]`)
x(err) p(err)
if len(pp) == 0 { if len(pp) == 0 {
_ = w(fmt.Errorf("empty: %s", url)) _ = w(fmt.Errorf("empty: %s", url))
// dit is echt fout // dit is echt fout
fouten = append(fouten, fmt.Sprintf("empty: %s\n", url)) fouten = append(fouten, fmt.Sprintf("empty: %s\n", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
for _, fout := range fouten { for _, fout := range fouten {
_, err = fp.WriteString(fout) p(fp.WriteString(fout))
x(err)
} }
x(fp.Close()) p(fp.Close())
x(os.Chtimes(filename+".err", timestamp, timestamp)) p(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false // echt fout return false // echt fout
} }
for _, p := range pp { for _, p1 := range pp {
_, err = buf.WriteString(addEnd(fixSpace(p.Content()))) p(buf.WriteString(addEnd(fixSpace(p1.Content()))))
x(err)
} }
fp, err := os.Create(filename + ".txt") fp, err := os.Create(filename + ".txt")
x(err) p(err)
_, err = fp.Write(buf.Bytes()) p(fp.Write(buf.Bytes()))
x(err) p(fp.Close())
x(fp.Close())
x(os.Chtimes(filename+".txt", timestamp, timestamp)) p(os.Chtimes(filename+".txt", timestamp, timestamp))
return true return true
} }

3
NieuwsNL/txt2corpus.sh
View File

@@ -53,8 +53,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../../metadata 2> err ../../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
RO/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -66,18 +66,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Reporters Online\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Reporters Online\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

56
RO/cmd/ro/ro.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -26,7 +26,7 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -36,22 +36,29 @@ func exists(filename string) bool {
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
req, err := http.NewRequest("GET", "https://reportersonline.nl/feed/", nil) req, err := http.NewRequest("GET", "https://reportersonline.nl/feed/", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -59,7 +66,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/RO/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/RO/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -71,19 +78,24 @@ func main() {
} }
filename := dirname + "/" + url.PathEscape(basename) filename := dirname + "/" + url.PathEscape(basename)
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
p(fp.WriteString("</item>\n"))
p(fp.Close())
p(os.Chtimes(filename+".xml", t, t))
ok = true
}()
} }
} }

14
RO/cmd/xml2txt/xml2txt.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -19,7 +19,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`) reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
) )
@@ -57,11 +57,9 @@ func main() {
var item Item var item Item
x(xml.Unmarshal(b, &item)) x(xml.Unmarshal(b, &item))
for _, cat := range item.Cats { for _, cat := range item.Cats {
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)) x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
x(err)
} }
_, err = fp.WriteString(addEnd(fixSpace(item.Title))) x(fp.WriteString(addEnd(fixSpace(item.Title))))
x(err)
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`)) doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
x(err) x(err)
root := doc.Root() root := doc.Root()
@@ -73,10 +71,8 @@ func main() {
pp, err := root.Search(`//body//p[not(.//a[contains(@href,"reportersonline.nl/support")])]`) pp, err := root.Search(`//body//p[not(.//a[contains(@href,"reportersonline.nl/support")])]`)
x(err) x(err)
for _, p := range pp { for _, p := range pp {
_, err = fp.WriteString(addEnd(fixSpace(p.Content()))) x(fp.WriteString(addEnd(fixSpace(p.Content()))))
x(err)
} }
x(err)
x(fp.Close()) x(fp.Close())
} }
} }

3
RO/txt2corpus.sh
View File

@@ -53,8 +53,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
Sargasso/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -66,18 +66,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sargasso\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sargasso\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

56
Sargasso/cmd/sargasso/sargasso.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -26,7 +26,7 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -36,22 +36,29 @@ func exists(filename string) bool {
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
req, err := http.NewRequest("GET", "https://sargasso.nl/feed/", nil) req, err := http.NewRequest("GET", "https://sargasso.nl/feed/", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -59,7 +66,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/Sargasso/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/Sargasso/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -71,19 +78,24 @@ func main() {
} }
filename := dirname + "/" + url.PathEscape(basename) filename := dirname + "/" + url.PathEscape(basename)
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
p(fp.WriteString("</)item>\n"))
p(fp.Close())
p(os.Chtimes(filename+".xml", t, t))
ok = true
}()
} }
} }

14
Sargasso/cmd/xml2txt/xml2txt.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -19,7 +19,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`) reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
) )
@@ -57,21 +57,17 @@ func main() {
var item Item var item Item
x(xml.Unmarshal(b, &item)) x(xml.Unmarshal(b, &item))
for _, cat := range item.Cats { for _, cat := range item.Cats {
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)) x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
x(err)
} }
_, err = fp.WriteString(addEnd(fixSpace(item.Title))) x(fp.WriteString(addEnd(fixSpace(item.Title))))
x(err)
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`)) doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
x(err) x(err)
root := doc.Root() root := doc.Root()
pp, err := root.Search(`//body//p`) pp, err := root.Search(`//body//p`)
x(err) x(err)
for _, p := range pp { for _, p := range pp {
_, err = fp.WriteString(addEnd(fixSpace(p.Content()))) x(fp.WriteString(addEnd(fixSpace(p.Content()))))
x(err)
} }
x(err)
x(fp.Close()) x(fp.Close())
} }
} }

3
Sargasso/txt2corpus.sh
View File

@@ -53,8 +53,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
Sikkom/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -16,7 +16,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
) )
@@ -51,18 +51,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sikkom\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Sikkom\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

132
Sikkom/cmd/sikkom/sikkom.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -29,8 +29,8 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
w = util.WarnErr w = e.WarnErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -47,21 +47,31 @@ func fileDate(filename string) string {
s := string(b) s := string(b)
i1 := strings.Index(s, "<unixTime>") + 10 i1 := strings.Index(s, "<unixTime>") + 10
i2 := strings.Index(s, "</unixTime>") i2 := strings.Index(s, "</unixTime>")
if i2 < i1 {
return ""
}
return s[i1:i2] return s[i1:i2]
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
resp, err := http.Get("https://www.sikkom.nl/api/feed/rss") resp, err := http.Get("https://www.sikkom.nl/api/feed/rss")
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -69,7 +79,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/Sikkom/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/Sikkom/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -80,22 +90,24 @@ func main() {
ts := fmt.Sprintf("%d", t.Unix()) ts := fmt.Sprintf("%d", t.Unix())
needUpdate := fileDate(filename+".xml") != ts needUpdate := fileDate(filename+".xml") != ts
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
if !doArticle(filename, item.Link, item.Title, t, needUpdate) { p(fp.WriteString("</item>\n"))
x(os.Remove(filename + ".xml")) 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) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
s := string(body) 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)) _ = w(fmt.Errorf("script jsonld not found: %s", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
_, err = fmt.Fprintf(fp, "script jsonld not found: %s\n", url) p(fmt.Fprintf(fp, "script jsonld not found: %s\n", url))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err := os.Create(filename + ".json") fp, err := os.Create(filename + ".json")
x(err) p(err)
_, err = fp.WriteString(s) p(fp.WriteString(s))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".json", timestamp, timestamp))
x(os.Chtimes(filename+".json", timestamp, timestamp))
doc, err := gokogiri.ParseHtml(body) doc, err := gokogiri.ParseHtml(body)
x(err) p(err)
root := doc.Root() root := doc.Root()
pp, err := root.Search(`//div[contains(@class,"article-page__body")]//p`) pp, err := root.Search(`//div[contains(@class,"article-page__body")]//p`)
x(err) p(err)
if len(pp) == 0 { if len(pp) == 0 {
_ = w(fmt.Errorf("empty: %s", url)) _ = w(fmt.Errorf("empty: %s", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
_, err = fmt.Fprintf(fp, "empty: %s\n", url) p(fmt.Fprintf(fp, "empty: %s\n", url))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(err)
x(fp.Close()) p(fp.Close())
x(os.Chtimes(filename+".html", timestamp, timestamp)) p(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err = os.Create(filename + ".txt") fp, err = os.Create(filename + ".txt")
x(err) p(err)
_, err = fp.WriteString(addEnd(fixSpace(title))) p(fp.WriteString(addEnd(fixSpace(title))))
x(err)
for _, p := range pp { for _, p1 := range pp {
_, err = fp.WriteString(addEnd(fixSpace(p.Content()))) p(fp.WriteString(addEnd(fixSpace(p1.Content()))))
x(err)
} }
x(fp.Close()) p(fp.Close())
return true return true
} }

3
Sikkom/txt2corpus.sh
View File

@@ -51,8 +51,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
Tzum/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -66,18 +66,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Tzum\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"Tzum\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

56
Tzum/cmd/tzum/tzum.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -26,7 +26,7 @@ type ItemT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
agent = "AhrefsBot/7.0" agent = "AhrefsBot/7.0"
) )
@@ -36,22 +36,29 @@ func exists(filename string) bool {
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
req, err := http.NewRequest("GET", "https://www.tzum.info/feed/", nil) req, err := http.NewRequest("GET", "https://www.tzum.info/feed/", nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -59,7 +66,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.PubDate) t, err = time.Parse(time.RFC1123, item.PubDate)
} }
x(err) p(err)
year, week := t.ISOWeek() year, week := t.ISOWeek()
dirname := fmt.Sprintf("/net/corpora/nlnieuws/Tzum/%d/%02d", year, week) dirname := fmt.Sprintf("/net/corpora/nlnieuws/Tzum/%d/%02d", year, week)
if exists(dirname + "/lock") { if exists(dirname + "/lock") {
@@ -71,19 +78,24 @@ func main() {
} }
filename := dirname + "/" + url.PathEscape(basename) filename := dirname + "/" + url.PathEscape(basename)
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
p(fp.WriteString("</item>\n"))
p(fp.Close())
p(os.Chtimes(filename+".xml", t, t))
ok = true
}()
} }
} }

14
Tzum/cmd/xml2txt/xml2txt.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"encoding/xml" "encoding/xml"
"fmt" "fmt"
@@ -19,7 +19,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`) reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`)
) )
@@ -57,11 +57,9 @@ func main() {
var item Item var item Item
x(xml.Unmarshal(b, &item)) x(xml.Unmarshal(b, &item))
for _, cat := range item.Cats { for _, cat := range item.Cats {
_, err = fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)) x(fmt.Fprintf(fp, "##META text cat = %s\n", fixSpace(cat)))
x(err)
} }
_, err = fp.WriteString(addEnd(fixSpace(item.Title))) x(fp.WriteString(addEnd(fixSpace(item.Title))))
x(err)
doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`)) doc, err := gokogiri.ParseHtml([]byte(`<html><body>` + item.Text + `</body></html>`))
x(err) x(err)
root := doc.Root() root := doc.Root()
@@ -70,11 +68,9 @@ func main() {
for _, p := range pp { for _, p := range pp {
s := p.Content() s := p.Content()
if !strings.Contains(s, "verscheen eerst op Tzum.") { if !strings.Contains(s, "verscheen eerst op Tzum.") {
_, err = fp.WriteString(addEnd(fixSpace(p.Content()))) x(fp.WriteString(addEnd(fixSpace(p.Content()))))
x(err)
} }
} }
x(err)
x(fp.Close()) x(fp.Close())
} }
} }

3
Tzum/txt2corpus.sh
View File

@@ -53,8 +53,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

19
VRT/cmd/metadata/metadata.go
View File

@@ -1,7 +1,7 @@
package main package main
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"bufio" "bufio"
"encoding/xml" "encoding/xml"
@@ -18,7 +18,7 @@ type Item struct {
} }
var ( var (
x = util.CheckErr x = e.ExitErr
escape = html.EscapeString escape = html.EscapeString
data = make(map[string][]string) data = make(map[string][]string)
location *time.Location location *time.Location
@@ -56,18 +56,13 @@ func main() {
i += strings.Index(s[i:], "<") i += strings.Index(s[i:], "<")
fp, err := os.Create("xml/" + filename + ".tmp") fp, err := os.Create("xml/" + filename + ".tmp")
x(err) x(err)
_, err = fp.WriteString(s[:i]) x(fp.WriteString(s[:i]))
x(err) x(fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"VRT\"/>\n"))
_, err = fp.WriteString("<metadata>\n <meta type=\"text\" name=\"source\" value=\"VRT\"/>\n")
x(err)
for _, m := range data[base] { for _, m := range data[base] {
_, err = fp.WriteString(" " + m + "\n") x(fp.WriteString(" " + m + "\n"))
x(err)
} }
_, err = fp.WriteString(" </metadata>\n ") x(fp.WriteString(" </metadata>\n "))
x(err) x(fp.WriteString(stripMeta(s[i:])))
_, err = fp.WriteString(stripMeta(s[i:]))
x(err)
x(fp.Close()) x(fp.Close())
x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) x(os.Rename("xml/"+filename+".tmp", "xml/"+filename))
} }

176
VRT/cmd/vrt/vrt.go
View File

@@ -1,8 +1,8 @@
package main package main
import ( import (
e "codeberg.org/pebbe/errors"
"github.com/jbowtie/gokogiri" "github.com/jbowtie/gokogiri"
"github.com/pebbe/util"
"bytes" "bytes"
"encoding/xml" "encoding/xml"
@@ -43,8 +43,8 @@ type LinkT struct {
} }
var ( var (
x = util.CheckErr p = e.PanicErr
w = util.WarnErr w = e.WarnErr
// agent = "AhrefsBot/7.0" // 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" 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) s := string(b)
i1 := strings.Index(s, "<unixTime>") + 10 i1 := strings.Index(s, "<unixTime>") + 10
i2 := strings.Index(s, "</unixTime>") i2 := strings.Index(s, "</unixTime>")
if i2 < i1 {
return ""
}
return s[i1:i2] return s[i1:i2]
} }
func main() { func main() {
defer func() {
if e.Panicked() {
_ = recover()
os.Exit(1)
}
}()
resp, err := http.Get("https://www.vrt.be/vrtnws/nl.rss.headlines.xml") resp, err := http.Get("https://www.vrt.be/vrtnws/nl.rss.headlines.xml")
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
var rss Rss var rss Rss
x(xml.Unmarshal(body, &rss)) p(xml.Unmarshal(body, &rss))
if len(rss.Items) == 0 { if len(rss.Items) == 0 {
x(fmt.Errorf("len(rss.Items) == 0")) p(fmt.Errorf("len(rss.Items) == 0"))
} }
for _, item := range rss.Items { for _, item := range rss.Items {
@@ -84,7 +94,7 @@ func main() {
if err != nil { if err != nil {
t, err = time.Parse(time.RFC1123, item.Published) t, err = time.Parse(time.RFC1123, item.Published)
} }
x(err) p(err)
t2, err := time.Parse(time.RFC3339Nano, item.Updated) t2, err := time.Parse(time.RFC3339Nano, item.Updated)
if err != nil { if err != nil {
t2, _ = time.Parse(time.RFC1123, item.Updated) t2, _ = time.Parse(time.RFC1123, item.Updated)
@@ -102,29 +112,31 @@ func main() {
ts := fmt.Sprintf("%d", t.Unix()) ts := fmt.Sprintf("%d", t.Unix())
needUpdate := fileDate(filename+".xml") != ts needUpdate := fileDate(filename+".xml") != ts
x(os.MkdirAll(dirname, 0777)) p(os.MkdirAll(dirname, 0777))
fp, err := os.Create(filename + ".xml") func() {
x(err) var ok bool
_, err = fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n") defer func() {
x(err) if !ok {
_, err = fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()) _ = os.Remove(filename + ".xml")
x(err) }
_, err = fp.Write(item.Data) }()
x(err) fp, err := os.Create(filename + ".xml")
_, err = fp.WriteString("</item>\n") p(err)
x(err) p(fp.WriteString("<?xml version=\"1.0\" encoding=\"UTF-8\"?>\n<item>\n"))
x(fp.Close()) p(fmt.Fprintf(fp, "<unixTime>%d</unixTime>", t.Unix()))
x(os.Chtimes(filename+".xml", t, t)) p(fp.Write(item.Data))
p(fp.WriteString("</item>\n"))
p(fp.Close())
p(os.Chtimes(filename+".xml", t, t))
var link string var link string
for _, l := range item.Link { for _, l := range item.Link {
if l.Type == "text/html" { if l.Type == "text/html" {
link = l.Href link = l.Href
}
} }
} ok = doArticle(filename, link, item.Title.Text, item.Nstag, item.Nslabeltag, t, needUpdate)
if !doArticle(filename, link, item.Title.Text, item.Nstag, item.Nslabeltag, t, needUpdate) { }()
x(os.Remove(filename + ".xml"))
}
} }
} }
@@ -142,15 +154,15 @@ func doArticle(filename string, url string, title string, tags []string, labels
time.Sleep(2 * time.Second) time.Sleep(2 * time.Second)
req, err := http.NewRequest("GET", url, nil) req, err := http.NewRequest("GET", url, nil)
x(err) p(err)
req.Header.Set("User-Agent", agent) req.Header.Set("User-Agent", agent)
client := &http.Client{} client := &http.Client{}
resp, err := client.Do(req) resp, err := client.Do(req)
x(err) p(err)
body, err := io.ReadAll(resp.Body) body, err := io.ReadAll(resp.Body)
x(err) p(err)
x(resp.Body.Close()) p(resp.Body.Close())
/* /*
s := string(body) 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)) _ = w(fmt.Errorf("script jsonld not found: %s", url))
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
_, err = fmt.Fprintf(fp, "script jsonld not found: %s\n", url) p(fmt.Fprintf(fp, "script jsonld not found: %s\n", url))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err := os.Create(filename + ".json") fp, err := os.Create(filename + ".json")
x(err) p(err)
_, err = fp.WriteString(s) p(fp.WriteString(s))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".json", timestamp, timestamp))
x(os.Chtimes(filename+".json", timestamp, timestamp))
*/ */
var buf bytes.Buffer var buf bytes.Buffer
doc, err := gokogiri.ParseHtml(body) doc, err := gokogiri.ParseHtml(body)
x(err) p(err)
root := doc.Root() root := doc.Root()
lnn, err := root.Search(`//head/link[@rel="canonical"]/@href`) lnn, err := root.Search(`//head/link[@rel="canonical"]/@href`)
x(err) p(err)
for _, ln := range lnn { for _, ln := range lnn {
if strings.Contains(ln.String(), "/liveblog/") { if strings.Contains(ln.String(), "/liveblog/") {
fp, err := os.Create(filename + ".skip") fp, err := os.Create(filename + ".skip")
x(err) p(err)
_, err = fp.WriteString("liveblog\n") p(fp.WriteString("liveblog\n"))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".skip", timestamp, timestamp))
x(os.Chtimes(filename+".skip", timestamp, timestamp))
return true return true
} }
} }
if len(tags) == 0 { if len(tags) == 0 {
_, err = fmt.Fprintln(&buf, "##META text cat =") p(fmt.Fprintln(&buf, "##META text cat ="))
x(err)
} else { } else {
for _, tag := range tags { for _, tag := range tags {
_, err = fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(tag)) p(fmt.Fprintf(&buf, "##META text cat = %s\n", fixSpace(tag)))
x(err)
} }
} }
if len(labels) == 0 { if len(labels) == 0 {
_, err = fmt.Fprintln(&buf, "##META text label =") p(fmt.Fprintln(&buf, "##META text label ="))
x(err)
} else { } else {
for _, label := range labels { for _, label := range labels {
_, err = fmt.Fprintf(&buf, "##META text label = %s\n", fixSpace(label)) p(fmt.Fprintf(&buf, "##META text label = %s\n", fixSpace(label)))
x(err)
} }
} }
_, err = buf.WriteString(addEnd(fixSpace(title))) _, err = buf.WriteString(addEnd(fixSpace(title)))
x(err) p(err)
fouten := make([]string, 0) fouten := make([]string, 0)
found := false found := false
pp, err := root.Search(`//div[@data-sentry-component="ArticleHeading"]//*[contains(@class,"prose-article-body-r")]`) pp, err := root.Search(`//div[@data-sentry-component="ArticleHeading"]//*[contains(@class,"prose-article-body-r")]`)
x(err) p(err)
for _, p := range pp { for _, p1 := range pp {
_, err = fmt.Fprint(&buf, addEnd(fixSpace(p.Content()))) p(fmt.Fprint(&buf, addEnd(fixSpace(p1.Content()))))
x(err)
found = true found = true
} }
if !found { 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="ArticleText"]//*[contains(@class,"prose-article-body-r")]` +
` | ` + ` | ` +
`//div[@data-sentry-component="ArticleTitle"]//h2`) `//div[@data-sentry-component="ArticleTitle"]//h2`)
x(err) p(err)
for _, p := range pp { for _, p1 := range pp {
_, err = fmt.Fprint(&buf, addEnd(fixSpace(p.Content()))) p(fmt.Fprint(&buf, addEnd(fixSpace(p1.Content()))))
x(err)
found = true found = true
} }
if !found { if !found {
@@ -270,31 +272,27 @@ func doArticle(filename string, url string, title string, tags []string, labels
if len(fouten) > 0 { if len(fouten) > 0 {
fp, err := os.Create(filename + ".err") fp, err := os.Create(filename + ".err")
x(err) p(err)
for _, fout := range fouten { for _, fout := range fouten {
_, err = fp.WriteString(fout) p(fp.WriteString(fout))
x(err)
} }
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".err", timestamp, timestamp))
x(os.Chtimes(filename+".err", timestamp, timestamp))
fp, err = os.Create(filename + ".html") fp, err = os.Create(filename + ".html")
x(err) p(err)
_, err = fp.Write(body) p(fp.Write(body))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".html", timestamp, timestamp))
x(os.Chtimes(filename+".html", timestamp, timestamp))
return false return false
} }
fp, err := os.Create(filename + ".txt") fp, err := os.Create(filename + ".txt")
x(err) p(err)
_, err = fp.Write(buf.Bytes()) p(fp.Write(buf.Bytes()))
x(err) p(fp.Close())
x(fp.Close()) p(os.Chtimes(filename+".txt", timestamp, timestamp))
x(os.Chtimes(filename+".txt", timestamp, timestamp))
return true return true
} }

3
VRT/txt2corpus.sh
View File

@@ -51,8 +51,7 @@ cd out
mkdir xml mkdir xml
Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log
../../../metadata 2> err ../../../metadata
rm err
cd xml cd xml
alto -o $corpus.data.dz *.xml 2> /dev/null alto -o $corpus.data.dz *.xml 2> /dev/null

4
cmd/ISOWeek/ISOWeek.go
View File

@@ -23,7 +23,7 @@ Dit programma geeft wel de juiste uitvoer.
*/ */
import ( import (
"github.com/pebbe/util" e "codeberg.org/pebbe/errors"
"fmt" "fmt"
"os" "os"
@@ -32,7 +32,7 @@ import (
) )
var ( var (
x = util.CheckErr x = e.ExitErr
) )
func main() { func main() {

9
go.mod
View File

@@ -1,8 +1,7 @@
module nlnieuws module nlnieuws
go 1.25.0 go 1.26.1
require ( require github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5
github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5
github.com/pebbe/util v0.10.0 require codeberg.org/pebbe/errors v0.3.0
)

4
go.sum
View File

@@ -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 h1:tQbR4RKFBFi0+Ll69dXejKKUbQVNaOAT2fjlDvSAfx4=
github.com/jbowtie/gokogiri v0.0.0-20250107075044-de0f9d4877a5/go.mod h1:kQE2lxPgVKe0JsBZMFFfMm5zBDCuRhaHFKOBzZeCLiw= 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
View File

@@ -15,6 +15,7 @@ overige opties:
-n int : max aantal resultaten -n int : max aantal resultaten
-p : gebruik pager -p : gebruik pager
-s : tel hits één keer per bericht -s : tel hits één keer per bericht
-v : verbose
" "
exit exit
} }
@@ -23,7 +24,8 @@ SINGLE=0
SELECT=0 SELECT=0
USEPAGER=0 USEPAGER=0
LIMIT=0 LIMIT=0
while getopts 'sin:p' opt VERBOSE=0
while getopts 'sin:pv' opt
do do
case "$opt" in case "$opt" in
i) i)
@@ -38,6 +40,9 @@ do
s) s)
SINGLE=1 SINGLE=1
;; ;;
v)
VERBOSE=1
;;
*) *)
usage usage
;; ;;
@@ -66,7 +71,11 @@ else
fi fi
search () { 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 "$@" for i in "$@"