837b2efde71507b09033c3ea6f2502e2306f8eef,pymorphy2/analyzer.py,MorphAnalyzer,_parse_as_word_with_known_suffix,#MorphAnalyzer#Any#Any#,219

Before Change

            if total_cnt > 1:
                // parses are sorted inside paradigms, but they are unsorted overall
                result.sort(reverse=True)
                result = [
                    (fixed_word, tag, normal_form, para_id, idx, cnt/total_cnt * ESTIMATE_DECAY)
                    for (cnt, fixed_word, tag, normal_form, para_id, idx) in result
                ]

After Change

        total_counts = [1] * len(self._paradigm_prefixes) // smoothing; XXX: isn"t max_cnt better?

        for prefix_id, prefix in self._paradigm_prefixes:

            if not word.startswith(prefix):
                continue

            suffixes_dawg = self._dictionary.prediction_suffixes_dawgs[prefix_id]

            for i in self._prediction_splits:
                end = word[-i:]  // XXX: this should be counted once, not for each prefix

                para_data = suffixes_dawg.similar_items(end, self._ee)

                for fixed_suffix, parses in para_data:
                    for cnt, para_id, idx in parses:

                        tag = self._build_tag_info(para_id, idx)

                        if not tag.is_productive():
                            continue

                        total_counts[prefix_id] += cnt

                        fixed_word = word[:-i] + fixed_suffix
                        normal_form = self._build_normal_form(para_id, idx, fixed_word)

                        parse = (cnt, fixed_word, tag, normal_form, para_id, idx, prefix_id)
                        reduced_parse = parse[1:4]
                        if reduced_parse in _seen_parses:
                            continue

                        result.append(parse)

                if total_counts[prefix_id] > 1:
                    break

        result = [
            (fixed_word, tag, normal_form, para_id, idx, cnt/total_counts[prefix_id] * ESTIMATE_DECAY)
            for (cnt, fixed_word, tag, normal_form, para_id, idx, prefix_id) in result
        ]

Instances