comparison.py

import difflib
import itertools

from constants import (
    ATTR_PUBTYPE_CONFERENCE,
    ATTR_PUBTYPE_JOURNAL,
    FIELD_AUTHOR_NAME,
    FIELD_AUTHORS,
    FIELD_JOURNAL,
    FIELD_JOURNAL_NAME,
    FIELD_PUBTYPE,
    FIELD_TITLE,
    FIELD_VENUE,
    FIELD_YEAR,
    SCORE_CRITICAL,
)


def _get_string_comparison_ratio(orig: str, other: str) -> float:
    ratio = difflib.SequenceMatcher(None, orig.lower(), other.lower()).ratio()
    return ratio


def _get_author_comparison_ratio(orig: list, other: list) -> list[float]:
    scores_authors = []
    if len(other) == 0:
        return [0.0]
    elif other[-1].get(FIELD_AUTHOR_NAME, "") == "et al.":
        orig = orig[: len(other)]
    # for _auth_orig, _auth_match in zip(authors_orig, authors_match):
    for _auth_orig, _auth_match in itertools.zip_longest(
        orig, other, fillvalue={"name": ""}
    ):
        _score_author = _get_string_comparison_ratio(
            _auth_orig.get(FIELD_AUTHOR_NAME, ""),
            _auth_match.get(FIELD_AUTHOR_NAME, ""),
        )
        scores_authors.append(_score_author)
    return scores_authors


def compare_references(original: dict, match: dict) -> dict:
    scores = {}
    for field in (FIELD_TITLE, FIELD_VENUE):
        _field_orig = original.get(field, "")
        _field_match = match.get(field, "")
        _ratio = _get_string_comparison_ratio(_field_orig, _field_match)
        scores[field] = _ratio

    scores[FIELD_YEAR] = 1 / (1 + abs(original[FIELD_YEAR] - match[FIELD_YEAR]))

    authors_orig = original.get(FIELD_AUTHORS, [])
    authors_match = match.get(FIELD_AUTHORS, [])
    scores_authors = _get_author_comparison_ratio(authors_orig, authors_match)
    scores[FIELD_AUTHORS] = sum(scores_authors) / len(scores_authors)

    _match_pub_types = match.get(FIELD_PUBTYPE, [])
    if not _match_pub_types:
        _match_pub_types = []

    if ATTR_PUBTYPE_CONFERENCE in _match_pub_types:
        _conf_orig = original.get(FIELD_VENUE, "")
        _conf_match = match.get(FIELD_JOURNAL, {}).get(FIELD_JOURNAL_NAME, "")
        _ratio = _get_string_comparison_ratio(_conf_orig, _conf_match)
        scores[FIELD_VENUE] = _ratio
    elif ATTR_PUBTYPE_JOURNAL in _match_pub_types:
        journal_orig = original.get(FIELD_JOURNAL, {})
        journal_match = match.get(FIELD_JOURNAL, {})
        _scores_journal = [
            _get_string_comparison_ratio(str(_o), str(journal_match.get(_k, "")))
            for _k, _o in journal_orig.items()
            if _k in journal_match
        ]
        if not _scores_journal:
            scores[FIELD_VENUE] = SCORE_CRITICAL
        else:
            scores[FIELD_VENUE] = sum(_scores_journal) / len(_scores_journal)
    return scores