diff --git a/.gitignore b/.gitignore index 27a1f83..969f0cf 100644 --- a/.gitignore +++ b/.gitignore @@ -11,6 +11,9 @@ NOS/metadata NOS/nos NU/metadata NU/nu +Oog/metadata +Oog/oog +Oog/xml2txt RO/metadata RO/ro RO/xml2txt diff --git a/Makefile b/Makefile index db85c05..f2159fa 100644 --- a/Makefile +++ b/Makefile @@ -5,6 +5,7 @@ all: make -C NieuwsNL make -C NOS make -C NU + make -C Oog make -C RO make -C RTVNoord make -C Sargasso diff --git a/Oog/Makefile b/Oog/Makefile new file mode 100644 index 0000000..1b26dc9 --- /dev/null +++ b/Oog/Makefile @@ -0,0 +1,13 @@ +all: \ + xml2txt \ + metadata \ + oog + +xml2txt: cmd/xml2txt/*.go + go build -o $@ $^ + +metadata: cmd/metadata/*.go + go build -o $@ $^ + +oog: cmd/oog/*.go + go build -o $@ $^ diff --git a/Oog/cmd/metadata/metadata.go b/Oog/cmd/metadata/metadata.go new file mode 100644 index 0000000..0c4d11f --- /dev/null +++ b/Oog/cmd/metadata/metadata.go @@ -0,0 +1,131 @@ +package main + +import ( + e "codeberg.org/pebbe/errors" + + "bufio" + "encoding/xml" + "fmt" + "html" + "os" + "strings" + "time" +) + +type Item struct { + XMLName xml.Name `xml:"item"` + UnixTime int64 `xml:"unixTime"` +} + +var ( + x = e.ExitErr + escape = html.EscapeString + data = make(map[string][]string) + location *time.Location +) + +func main() { + var err error + location, err = time.LoadLocation("Europe/Amsterdam") + x(err) + + files, err := os.ReadDir(".") + x(err) + for _, file := range files { + filename := file.Name() + if strings.HasSuffix(filename, ".txt") { + doText("", filename) + } else if strings.HasSuffix(filename, ".xml") { + doXml("", filename) + } + } + files, err = os.ReadDir("..") + x(err) + for _, file := range files { + filename := file.Name() + if strings.HasSuffix(filename, ".txt") { + doText("../", filename) + } else if strings.HasSuffix(filename, ".xml") { + doXml("../", filename) + } + } + + files, err = os.ReadDir("xml") + x(err) + for _, file := range files { + filename := file.Name() + if !strings.HasSuffix(filename, ".xml") { + continue + } + aa := strings.Split(filename, ".") + base := strings.Join(aa[1:len(aa)-2], ".") + b, err := os.ReadFile("xml/" + filename) + x(err) + s := string(b) + i := strings.Index(s, "\n \n")) + for _, m := range data[base] { + x(fp.WriteString(" " + m + "\n")) + } + x(fp.WriteString(" \n ")) + x(fp.WriteString(stripMeta(s[i:]))) + x(fp.Close()) + x(os.Rename("xml/"+filename+".tmp", "xml/"+filename)) + } +} + +func doText(dirname, filename string) { + base := filename[:len(filename)-4] + if _, ok := data[base]; !ok { + data[base] = make([]string, 0) + } + fp, err := os.Open(dirname + filename) + x(err) + defer func() { x(fp.Close()) }() + scanner := bufio.NewScanner(fp) + for scanner.Scan() { + line := scanner.Text() + if !strings.HasPrefix(line, "##META") { + continue + } + aa := strings.Fields(line) + if len(aa) > 4 { + data[base] = append(data[base], + fmt.Sprintf(``, + aa[1], + escape(aa[2]), + escape(strings.Join(aa[4:], " ")))) + } + } + x(scanner.Err()) +} + +func doXml(dirname, filename string) { + base := filename[:len(filename)-4] + if _, ok := data[base]; !ok { + data[base] = make([]string, 0) + } + b, err := os.ReadFile(dirname + filename) + x(err) + var item Item + x(xml.Unmarshal(b, &item)) + t := time.Unix(item.UnixTime, 0).In(location) + data[base] = append(data[base], + fmt.Sprintf(``, + t.Year(), + int(t.Month()), + t.Day())) +} + +func stripMeta(s string) string { + i1 := strings.Index(s, "") + if i1 < 0 { + return s + } + i2 := i1 + strings.Index(s[i1:], "") + 11 + return s[:i1] + strings.TrimLeft(s[i2:], " \t\r\n") +} diff --git a/Oog/cmd/oog/oog.go b/Oog/cmd/oog/oog.go new file mode 100644 index 0000000..dc07f7e --- /dev/null +++ b/Oog/cmd/oog/oog.go @@ -0,0 +1,126 @@ +package main + +import ( + e "codeberg.org/pebbe/errors" + + "encoding/xml" + "fmt" + "io" + "net/http" + "net/url" + "os" + "path/filepath" + "strings" + "time" +) + +type Rss struct { + XMLName xml.Name `xml:"rss"` + Items []ItemT `xml:"channel>item"` +} + +type ItemT struct { + PubDate string `xml:"pubDate"` + UnixTime int64 `xml:"unixTime"` + Link string `xml:"link"` + Guid string `xml:"guid"` + Data []byte `xml:",innerxml"` +} + +var ( + p = e.PanicErr + agent = "AhrefsBot/7.0" +) + +func exists(filename string) bool { + _, err := os.Stat(filename) + return err == nil +} + +func main() { + defer func() { + if e.Panicked { + _ = recover() + os.Exit(1) + } + }() + + myLock := "/net/corpora/nlnieuws/Oog/lock" + mkLock(myLock) + defer func() { + _ = os.Remove(myLock) + }() + + req, err := http.NewRequest("GET", "https://www.oogtv.nl/feed/", nil) + p(err) + req.Header.Set("User-Agent", agent) + + client := &http.Client{} + resp, err := client.Do(req) + p(err) + body, err := io.ReadAll(resp.Body) + p(err) + p(resp.Body.Close()) + + var rss Rss + p(xml.Unmarshal(body, &rss)) + + if len(rss.Items) == 0 { + p(fmt.Errorf("len(rss.Items) == 0")) + } + + for _, item := range rss.Items { + t, err := time.Parse(time.RFC1123Z, item.PubDate) + if err != nil { + t, err = time.Parse(time.RFC1123, item.PubDate) + } + p(err) + year, week := t.ISOWeek() + dirname := fmt.Sprintf("/net/corpora/nlnieuws/Oog/%d/%02d", year, week) + if exists(dirname + "/lock") { + continue + } + basename := strings.TrimPrefix(item.Guid, "https://www.oogtv.nl/?p=") + if i := strings.LastIndex(basename, "/"); i > 0 { + basename = basename[:i] + } + filename := dirname + "/" + url.PathEscape(basename) + + p(os.MkdirAll(dirname, 0777)) + func() { + var ok bool + defer func() { + if e.Panicked { + fmt.Fprintln(os.Stderr, "----", filename) + fmt.Fprintln(os.Stderr, "----", item.Link) + } + if !ok { + _ = os.Remove(filename + ".xml") + } + }() + fp, err := os.Create(filename + ".xml") + p(err) + p(fp.WriteString("\n\n")) + p(fmt.Fprintf(fp, "%d", t.Unix())) + p(fp.Write(item.Data)) + p(fp.WriteString("\n")) + p(fp.Close()) + p(os.Chtimes(filename+".xml", t, t)) + ok = true + }() + } + +} + +func mkLock(filename string) { + pid := os.Getpid() + link := fmt.Sprintf("%s.%d", filepath.Base(filename), pid) + p(os.Symlink(link, filename)) + + name, err := os.Readlink(filename) + p(err) + + if name != link { + p(fmt.Errorf("wrong lock name %q, should be %q", name, link)) + } +} diff --git a/Oog/cmd/xml2txt/xml2txt.go b/Oog/cmd/xml2txt/xml2txt.go new file mode 100644 index 0000000..4be9dc4 --- /dev/null +++ b/Oog/cmd/xml2txt/xml2txt.go @@ -0,0 +1,97 @@ +package main + +import ( + e "codeberg.org/pebbe/errors" + "github.com/jbowtie/gokogiri" + + "encoding/xml" + "fmt" + "os" + "regexp" + "strings" + "time" +) + +type Item struct { + Title string `xml:"title"` + Text string `xml:"encoded"` + Cats []string `xml:"category"` +} + +var ( + x = e.ExitErr + + reYearWeek = regexp.MustCompile(`^2[0-9][0-9][0-9]-[0-5][0-9]$`) +) + +func main() { + + var ds string + switch len(os.Args) { + case 1: + year, week := time.Now().AddDate(0, 0, -7).ISOWeek() + ds = fmt.Sprintf("%d-%02d", year, week) + case 2: + if !reYearWeek.MatchString(os.Args[1]) { + x(fmt.Errorf("arg must be yyyy-ww")) + } + ds = os.Args[1] + default: + x(fmt.Errorf("too many arguments")) + } + dp := ds[:4] + "/" + ds[5:] + + x(os.Chdir("/net/corpora/nlnieuws/Oog/" + dp)) + x(os.MkdirAll("out", 0777)) + files, err := os.ReadDir(".") + x(err) + for _, file := range files { + filename := file.Name() + if !strings.HasSuffix(filename, ".xml") { + continue + } + b, err := os.ReadFile(filename) + x(err) + fp, err := os.Create("out/" + filename[:len(filename)-4] + ".txt") + x(err) + var item Item + x(xml.Unmarshal(b, &item)) + for _, cat := range item.Cats { + x(fmt.Fprintf(fp, "##META text tag = %s\n", fixSpace(cat))) + } + x(fp.WriteString(addEnd(fixSpace(item.Title)))) + doc, err := gokogiri.ParseHtml([]byte(`` + item.Text + ``)) + x(err) + root := doc.Root() + pp, err := root.Search(`//body/p`) + x(err) + for _, p := range pp { + x(fp.WriteString(addEnd(fixSpace(p.Content())))) + } + x(fp.Close()) + } +} + +func addEnd(s string) string { + s = strings.TrimSpace(s) + n := len(s) + if n == 0 { + return "" + } + if n > 0 { + if strings.ContainsAny(s[n-1:], ".!?") { + return s + "\n" + } + } + if n > 1 { + s2 := s[n-2:] + if s2 == `."` || s2 == `!"` || s2 == `?"` || s2 == `.'` || s2 == `!'` || s2 == `?'` { + return s + "\n" + } + } + return s + ".\n" +} + +func fixSpace(s string) string { + return strings.Join(strings.Fields(s), " ") +} diff --git a/Oog/txt2corpus.sh b/Oog/txt2corpus.sh new file mode 100755 index 0000000..083e049 --- /dev/null +++ b/Oog/txt2corpus.sh @@ -0,0 +1,67 @@ +#!/bin/bash + +set -e + +unset CDPATH +PATH=/net/corpora/nlnieuws/bin:/net/aps/bin:$PATH +export TZ=Europe/Amsterdam +. /net/aps/etc/alpino-activate.sh > /dev/null + +if [ "$1" = "" ] +then + ds=`ISOWeek -7` +else + case "$1" in + 2[0-9][0-9][0-9]-[0-5][0-9]) + ds=$1 + ;; + *) + echo INVALID + exit 1 + ;; + esac +fi + +dp=${ds//-//} + +corpus=/net/corpora/nlnieuws/Oog/corpus/$ds + +cd /net/corpora/nlnieuws/Oog/$dp + +ln -s lock.$$ lock +if [ "`readlink lock`" != lock.$$ ] +then + echo Getting lock failed + exit 1 +fi + +rm -fr out +mkdir out + +../../xml2txt $ds + +rm -f $corpus.lines +for i in out/*.txt +do + b=`basename $i .txt` + perl -p -e 's/^\s*//; s/^##META.*\n//' $i | tokenize.sh \ + | perl -e '$n = 0; while(<>) { $n++; print("oog.'$b'.$n|$_"); }' \ + >> $corpus.lines +done + +cd out +mkdir xml +Alpino -flag treebank xml debug=1 end_hook=xml user_max=900000 -parse < $corpus.lines 2> $corpus.log + +../../../metadata + +cd xml +alto -o $corpus.data.dz *.xml 2> /dev/null + +# telling per bericht, niet per zin +/net/corpora/nlnieuws/namen.sh -x T -s $corpus.data.dz > $corpus.tag.txt + +cd ../.. +rm -fr out + +rm -f lock diff --git a/collect.sh b/collect.sh index 8b15bdf..3f916b0 100755 --- a/collect.sh +++ b/collect.sh @@ -40,7 +40,7 @@ cd /net/corpora/nlnieuws/data declare -A parts parts[alles]='.' parts[algemeen]='NOS|NU|NieuwsNL|RO|Sargasso' -parts[groningen]='GG|RTVNoord|Sikkom' +parts[groningen]='GG|Oog|RTVNoord|Sikkom' parts[AT5]='AT5' parts[GG]='GG' parts[NOS]='NOS'