authoritymix

#!/usr/bin/python
# -*- encoding: utf-8 -*-

import itertools, logging
from metadata import JSONMetadataWrapper, CF
from copy import deepcopy
import db, dbutil, nameutil, util

logging.basicConfig(level=logging.INFO)
logger = logging.getLogger('authoritymix')

def distAuthorityTuples(alreadyPresent, revisions, k, name):
    if name is None:
        return
    name = util.norm(name, util.NormLevel.ONLY_LETTERS)
    for rev in revisions:
        meta = JSONMetadataWrapper(rev.meta)
        for author in meta.get('dc.contributor.author', what=None):
            xget = lambda key: util.noneIfEmpty(author.get(key))
            authority = xget('authority')
            if authority in alreadyPresent:
                continue
            curName = xget(k)
            if curName:
                curName = util.norm(curName, util.NormLevel.ONLY_LETTERS)
                yield (nameutil.levenshtein(name, curName), xget('authority'))

def main():
    for mainRev, otherRevs in dbutil.yieldRevGroups():
        meta = JSONMetadataWrapper(deepcopy(mainRev.meta))

        alreadyPresent = set(meta.get('dc.contributor.author', what='authority')) - {None, ''}
        authors = meta.get('dc.contributor.author', what=None)

        for author in authors:
            xget = lambda key: util.noneIfEmpty(author.get(key))
            if xget('authority') is None:
                # Se não estiver definido ID de autoridade para este autor, procura
                # o ID do nome de autor mais próximo presente nas duplicatas
                scoredTuples = sorted(itertools.chain(
                    distAuthorityTuples(alreadyPresent, otherRevs, 'value', xget('value')),
                    distAuthorityTuples(alreadyPresent, otherRevs, '_nomecompleto', xget('_nomecompleto'))
                ))
                confidence = CF.UNCERTAIN
                if len(scoredTuples) > 1 and scoredTuples[0][0] == scoredTuples[1][0]:
                    # Empate de distância
                    confidence = CF.AMBIGUOUS
                if len(scoredTuples) > 0 and scoredTuples[0][1] is not None:
                    author['authority'] = scoredTuples[0][1]
                    author['confidence'] = confidence

        # Cria nova revisão atualizada caso os metadados tenham mudado
        if meta.json != mainRev.meta:
            logger.info('Atualizado metadado do item %r: dc.contributor.author=%r',
                        mainRev.item_id, authors)
            newRev = db.Revision(item_id=mainRev.item_id, source='authoritymix', meta=meta.json)
            db.session.add(newRev)
            # Atualiza revisão principal das revisões que tem esta marcada como duplicata
            for rev in otherRevs:
                assert rev in db.session
                rev.duplicate_of = newRev
            db.session.commit()

    db.session.refresh_materialized_view(db.LastRevision)


if __name__ == '__main__':
    main()