// skate-wikipedia-doi extracts DOI from wikipedia reference dataset.
package main

import (
	"flag"
	"fmt"
	"log"
	"os"
	"regexp"
	"runtime"
	"strings"

	"git.archive.org/martin/cgraph/skate"
	"git.archive.org/martin/cgraph/skate/parallel"
	"github.com/segmentio/encoding/json"
)

var (
	numWorkers   = flag.Int("w", runtime.NumCPU(), "number of workers")
	batchSize    = flag.Int("b", 100000, "batch size")
	bytesNewline = []byte("\n")
	wsReplacer   = strings.NewReplacer("\t", "", "\n", "", " ", "")
	patDOI       = regexp.MustCompile(`(10[.][0-9]{1,8}/[^ ]*[\w])`)
)

func main() {
	flag.Parse()
	pp := parallel.NewProcessor(os.Stdin, os.Stdout, func(p []byte) ([]byte, error) {
		var w skate.MinimalCitations
		if err := json.Unmarshal(p, &w); err != nil {
			return nil, err
		}
		ids := w.ParseIDList()
		if ids.DOI == "" {
			return nil, nil
		}
		match := patDOI.FindStringSubmatch(ids.DOI)
		if len(match) == 0 {
			return nil, nil
		}
		var (
			doi       = skate.SanitizeDOI(wsReplacer.Replace(match[0]))
			pageTitle = strings.TrimSpace(w.PageTitle)
			s         = fmt.Sprintf("%s\t%s\t%s", doi, pageTitle, string(p))
		)
		return []byte(s), nil

	})
	pp.NumWorkers = *numWorkers
	pp.BatchSize = *batchSize
	if err := pp.Run(); err != nil {
		log.Fatal(err)
	}
}