path: root/skate/zippy.go

// This file contains various "reducers", e.g. functions, that take two data
// streams and will apply a computation on matched groups.
package skate

import (
	"fmt"
	"io"
	"log"
	"strings"

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

// groupLogf logs a message alongsize a serialized group.
func groupLogf(g *zipkey.Group, s string, vs ...interface{}) {
	log.Printf(s, vs...)
	b, _ := json.Marshal(g)
	log.Println(string(b))
}

// ZippyExact takes a release and refs reader (key, doc) and assigns a fixed
// match result. XXX: allow empty keys
func ZippyExact(releases, refs io.Reader, matchResult MatchResult, w io.Writer) error {
	var (
		enc     = json.NewEncoder(w)
		keyer   = makeKeyFunc("\t", 1)
		i       = 0
		grouper = func(g *zipkey.Group) error {
			i++
			if i%10000 == 0 {
				log.Printf("processed %v groups", i)
			}
			var (
				target *Release
				ref    *Ref
				err    error
			)
			if len(g.G0) == 0 || len(g.G1) == 0 {
				return nil
			}
			if target, err = stringToRelease(Cut(g.G0[0], 2)); err != nil {
				groupLogf(g, "[skip] failed to parse release: %v", err)
				return nil
			}
			for _, line := range g.G1 {
				if ref, err = stringToRef(Cut(line, 2)); err != nil {
					groupLogf(g, "[skip] failed to parse ref: %v", err)
					continue
				}
				var bref BiblioRef
				bref.SourceReleaseIdent = ref.ReleaseIdent
				bref.SourceWorkIdent = ref.WorkIdent
				bref.SourceReleaseStage = ref.ReleaseStage
				bref.SourceYear = fmt.Sprintf("%d", ref.ReleaseYear)
				bref.RefIndex = ref.Index + 1 // we want 1-index (also helps with omitempty)
				bref.RefKey = ref.Key
				bref.TargetReleaseIdent = target.Ident
				bref.TargetWorkIdent = target.WorkID
				bref.MatchProvenance = ref.RefSource
				bref.MatchStatus = matchResult.Status.Short()
				bref.MatchReason = matchResult.Reason.Short()
				if err := enc.Encode(bref); err != nil {
					return err
				}
			}
			return nil
		}
	)
	zipper := zipkey.New(releases, refs, keyer, grouper)
	return zipper.Run()
}

// ZippyExactWiki takes a release and wiki reader (key, doc) and assigns a
// fixed match result.
func ZippyExactWiki(releases, wiki io.Reader, mr MatchResult, w io.Writer) error {
	var (
		enc     = json.NewEncoder(w)
		keyer   = makeKeyFunc("\t", 1)
		grouper = func(g *zipkey.Group) error {
			var (
				target *Release
				wiki   *MinimalCitations
				err    error
			)
			if len(g.G0) == 0 || len(g.G1) == 0 {
				return nil
			}
			if target, err = stringToRelease(Cut(g.G0[0], 2)); err != nil {
				return err
			}
			for _, line := range g.G1 {
				if wiki, err = stringToWiki(Cut(line, 2)); err != nil {
					return err
				}
				var bref BiblioRef
				bref.Key = fmt.Sprintf("%s_%s", slugifyString(wiki.PageTitle), target.Ident) // XXX: what should we use?
				bref.SourceWikipediaArticle = wiki.PageTitle
				bref.TargetReleaseIdent = target.Ident
				bref.TargetWorkIdent = target.WorkID
				bref.MatchProvenance = "wikipedia"
				bref.MatchStatus = mr.Status.Short()
				bref.MatchReason = mr.Reason.Short()
				if err := enc.Encode(bref); err != nil {
					return err
				}
			}
			return nil
		}
	)
	zipper := zipkey.New(releases, wiki, keyer, grouper)
	return zipper.Run()
}

// ZippyVerifyRefs takes a release and refs reader (tsv, with ident, key, doc)
// and will execute gf for each group found.
func ZippyVerifyRefs(releases, refs io.Reader, w io.Writer) error {
	// Define a grouper, working on one set of refs and releases with the same
	// key at a time. Here, we do verification and write out the generated
	// biblioref.
	var (
		enc     = json.NewEncoder(w)
		keyer   = makeKeyFunc("\t", 1)
		grouper = func(g *zipkey.Group) error {
			var (
				re, pivot *Release
				err       error
			)
			if len(g.G0) == 0 || len(g.G1) == 0 {
				return nil
			}
			if pivot, err = stringToRelease(Cut(g.G0[0], 2)); err != nil {
				return err
			}
			for _, line := range g.G1 {
				if re, err = stringToRelease(Cut(line, 2)); err != nil {
					return err
				}
				result := Verify(pivot, re)
				switch result.Status {
				case StatusExact, StatusStrong:
					if result.Reason == ReasonDOI {
						continue
					}
					br := generateBiblioRef(re, pivot, result, "fuzzy")
					if err := enc.Encode(br); err != nil {
						return err
					}
				default:
					// XXX: We want to add unmatched pieces as well; here?  We
					// probably want to do a single final pass to complete the
					// dataset.
				}
			}
			return nil
		}
	)
	zipper := zipkey.New(releases, refs, keyer, grouper)
	return zipper.Run()
}

// ZippyVerifyRefsOpenLibraryTable takes OL editions (as release) and refs (as
// release) and emits biblioref.
func ZippyVerifyRefsOpenLibraryTable(olr, refs io.Reader, w io.Writer) error {
	// Define a grouper, working on one set of refs and releases with the same
	// key at a time.
	var (
		keyer   = makeKeyFunc("\t", 1)
		grouper = func(g *zipkey.Group) error {
			var (
				re, pivot *Release
				err       error
			)
			if len(g.G0) == 0 || len(g.G1) == 0 {
				return nil
			}
			// We take a single edition from OL; XXX: might want to link to work.
			if pivot, err = stringToRelease(Cut(g.G0[0], 2)); err != nil {
				return err
			}
			for _, line := range g.G1 {
				if re, err = stringToRelease(Cut(line, 2)); err != nil {
					return err
				}
				// The refs have a container name, but not a title, but here we
				// compare against titles from open library.
				re.Title = re.ContainerName
				result := Verify(pivot, re)
				fmt.Printf("%s\t%s\t%s\t%s\t%s\t%s\t%s\n",
					result.Status.Short(),
					result.Reason.Short(),
					pivot.Extra.OpenLibrary.WorkID,
					re.Ident,
					CutSep(g.G0[0], "\t", 1),
					pivot.Title,
					re.Title)
			}
			return nil
		}
	)
	zipper := zipkey.New(olr, refs, keyer, grouper)
	return zipper.Run()
}

// Cut returns a specific column (1-indexed, like CutSep) from a tabular
// file, returns empty string if column is invalid.
func Cut(line string, column int) string {
	return CutSep(line, "\t", column)
}

// CutSep allows to specify a separator.
func CutSep(line, sep string, column int) string {
	parts := strings.Split(strings.TrimSpace(line), sep)
	if len(parts) < column {
		return ""
	} else {
		return parts[column-1]
	}
}

// makeKeyFunc creates a function that can be used as keyFunc, selecting a
// column from fields separated by sep; column is 1-indexed.
func makeKeyFunc(sep string, column int) func(string) (string, error) {
	return func(s string) (string, error) {
		if k := CutSep(s, sep, column); k == "" {
			return k, fmt.Errorf("cannot get key from column %d in line (len=%d): %s", column, len(s), s)
		} else {
			return k, nil
		}
	}
}

func stringToRelease(s string) (r *Release, err error) {
	err = json.Unmarshal([]byte(s), &r)
	return
}

func stringToRef(s string) (r *Ref, err error) {
	err = json.Unmarshal([]byte(s), &r)
	return
}

func stringToWiki(s string) (r *MinimalCitations, err error) {
	err = json.Unmarshal([]byte(s), &r)
	return
}