Files
nlnieuws/cmd/data2json/data2json.go
2026-05-23 13:48:56 +02:00

245 lines
5.5 KiB
Go

package main
import (
e "codeberg.org/pebbe/errors"
"bufio"
"encoding/json"
"fmt"
"os"
"strconv"
"strings"
"time"
)
type Data struct {
Year int `json:"year"`
Week int `json:"week"`
First string `json:"first"`
Last string `json:"last"`
Period int `json:"period"`
Start string `json:"start"`
Max int `json:"max"`
Sources map[string]int `json:"sources"`
Algemeen *Parts `json:"Algemeen"`
Groningen *Parts `json:"Groningen"`
Amsterdam *Parts `json:"Amsterdam"`
Literatuur *Parts `json:"Literatuur"`
Vlaanderen *Parts `json:"Vlaanderen"`
}
type Parts struct {
NieuweNamen [][5]any `json:"nieuwe namen"`
NieuweWoorden [][5]any `json:"nieuwe woorden"`
Personen [][5]any `json:"personen"`
AndereNamen [][5]any `json:"andere namen"`
Locaties [][5]any `json:"locaties"`
Organisaties [][5]any `json:"organisaties"`
}
var (
sources = map[string]string{
"Algemeen": "algemeen",
"Amsterdam": "amsterdam",
"Groningen": "groningen",
"Literatuur": "literatuur",
"Vlaanderen": "vlaanderen",
}
parts = map[string]struct {
file string
suffix string
}{
"nieuwe namen": {"nieuwe-namen", ".t20"},
"nieuwe woorden": {"nieuwe-woorden-extra", ".t20"},
"personen": {"personen", ""},
"andere namen": {"overige-namen", ""},
"locaties": {"locaties", ""},
"organisaties": {"organisaties", ""},
}
maanden = strings.Fields("x januari februari maart april mei juni juli augustus september oktober november december")
dagen = strings.Fields("zondag maandag dinsdag woensdag donderdag vrijdag zaterdag")
year int
week int
size int
x = e.ExitErr
)
func main() {
aa := strings.Split(os.Args[1], "-")
if len(aa) != 2 {
x(fmt.Errorf("ongeldig argument, moet in formaat yyyy-dd zijn"))
}
var err error
year, err = strconv.Atoi(aa[0])
x(err)
week, err = strconv.Atoi(aa[1])
x(err)
size, err = strconv.Atoi(os.Args[2])
x(err)
if year < 1000 || year > 9999 {
x(fmt.Errorf("ongeldig year: %d", year))
}
if week < 1 || week > 53 {
x(fmt.Errorf("ongeldige week: %d", week))
}
start, first, last, names := dates()
max, sources := makeCounts(names)
data := &Data{
Year: year,
Week: week,
First: first,
Last: last,
Period: size,
Start: start,
Max: max,
Sources: sources,
Algemeen: makeParts("Algemeen"),
Groningen: makeParts("Groningen"),
Amsterdam: makeParts("Amsterdam"),
Literatuur: makeParts("Literatuur"),
Vlaanderen: makeParts("Vlaanderen"),
}
b, err := json.Marshal(data)
x(err)
fmt.Println(string(b))
}
func makeParts(source string) *Parts {
return &Parts{
NieuweNamen: makeValues(source, "nieuwe namen"),
NieuweWoorden: makeValues(source, "nieuwe woorden"),
Personen: makeValues(source, "personen"),
AndereNamen: makeValues(source, "andere namen"),
Locaties: makeValues(source, "locaties"),
Organisaties: makeValues(source, "organisaties"),
}
}
func makeValues(source, part string) [][5]any {
v := make([][5]any, 0)
filename := fmt.Sprintf("/net/corpora/nlnieuws/data/%s-%s-%d-%02d-%d%s",
sources[source],
parts[part].file,
year,
week,
size,
parts[part].suffix)
fp, err := os.Open(filename)
x(err)
scanner := bufio.NewScanner(fp)
lineno := 0
for scanner.Scan() {
lineno++
line := scanner.Text()
aa := strings.Split(line, "\t")
count, err := strconv.Atoi(strings.TrimSpace(aa[0]))
x(err)
word := aa[1]
var tags, lemma, postag string
if len(aa) > 2 {
tags = aa[2]
}
if len(aa) > 3 {
lemma = aa[3]
}
if len(aa) > 4 {
postag = aa[4]
}
v = append(v, [5]any{count, word, tags, lemma, postag})
if lineno == 20 {
break
}
}
x(scanner.Err())
return v
}
func makeCounts(names []string) (int, map[string]int) {
max := 0
counts := make(map[string]int)
x(os.Chdir("/net/corpora/nlnieuws"))
files, err := os.ReadDir(".")
x(err)
for _, file := range files {
if !file.IsDir() {
continue
}
filename := file.Name()
if filename[0] < 'A' || filename[0] > 'Z' {
continue
}
count := 0
for _, name := range names {
files2, err := os.ReadDir(filename + "/" + name)
if err != nil {
continue
}
for _, f := range files2 {
if n := f.Name(); strings.HasSuffix(n, ".xml") {
count++
} else if strings.HasSuffix(n, ".skip") {
count--
}
}
}
counts[filename] = count
if count > max {
max = count
}
}
return max, counts
}
func dates() (start, first, last string, names []string) {
// 1 januari
t := time.Date(year, 1, 1, 12, 0, 0, 0, time.UTC)
// zoek eerste donderdag
day := int(t.Weekday()) // 0 = zondag
donderdag := 4 - day
if donderdag < 0 {
donderdag = donderdag + 7
}
// schuif naar maandag voor eerste donderdag: donderdag - 3
// in de gewenste week: 7 * (week - 1)
t = t.AddDate(0, 0, donderdag-3+7*(week-1))
t2 := t.AddDate(0, 0, 6)
tStart := t.AddDate(0, 0, (1-size)*7)
names = make([]string, 0)
t3 := tStart
for range size {
y, w := t3.ISOWeek()
names = append(names, fmt.Sprintf("%d/%02d", y, w))
t3 = t3.AddDate(0, 0, 7)
}
t3 = tStart
for range 7 * size {
names = append(names, fmt.Sprintf("%d/%02d/%02d", t3.Year(), t3.Month(), t3.Day()))
t3 = t3.AddDate(0, 0, 1)
}
return makeDate(tStart), makeDate(t), makeDate(t2), names
}
func makeDate(d time.Time) string {
return fmt.Sprintf("%s %d %s %d", dagen[d.Weekday()][:2], d.Day(), maanden[int(d.Month())][:3], d.Year())
}