anya/services/similarity.py

from collections import Counter
import logging
from math import sqrt
from collections.abc import Sequence

logger = logging.getLogger(__name__)


def cosine_similarity(a: Sequence[float], b: Sequence[float]) -> float:
    logger.debug("Computing cosine similarity between numerical vectors of length %d and %d", len(a), len(b))
    if len(a) != len(b):
        logger.error("Vector dimension mismatch: vector 'a' length (%d) != vector 'b' length (%d)", len(a), len(b))
        raise ValueError("Vectors must have the same dimension")

    dot = 0.0
    norm_a_sq = 0.0
    norm_b_sq = 0.0

    for x, y in zip(a, b):
        dot += x * y
        norm_a_sq += x * x
        norm_b_sq += y * y

    denominator = sqrt(norm_a_sq * norm_b_sq)
    if denominator == 0.0:
        logger.debug("Zero denominator encountered in cosine_similarity (norm_a_sq=%f, norm_b_sq=%f). Returning 0.0", norm_a_sq, norm_b_sq)
        return 0.0

    similarity = dot / denominator
    logger.debug("Calculated vector cosine similarity: dot=%f, denominator=%f, similarity=%f", dot, denominator, similarity)
    return similarity


def cosine_lists(a: list[str], b: list[str], *, casefold: bool = True) -> float:
    logger.debug("Computing token cosine similarity for list_a=%s and list_b=%s (casefold=%s)", a, b, casefold)

    def tokens(xs: list[str]) -> Counter[str]:
        return Counter(x.casefold() if casefold else x for x in xs)

    ca, cb = tokens(a), tokens(b)
    if not ca or not cb:
        logger.debug("Empty token set detected (count_a=%d, count_b=%d). Cosine similarity is 0.0", len(ca), len(cb))
        return 0.0

    common_tokens = ca.keys() & cb.keys()
    dot = sum(ca[t] * cb[t] for t in common_tokens)
    norm_a = sqrt(sum(v * v for v in ca.values()))
    norm_b = sqrt(sum(v * v for v in cb.values()))

    if norm_a == 0.0 or norm_b == 0.0:
        logger.debug("Zero norm detected (norm_a=%f, norm_b=%f). Cosine similarity is 0.0", norm_a, norm_b)
        return 0.0

    similarity = dot / (norm_a * norm_b)
    logger.debug("Token similarity calculation: common_tokens=%s, dot=%f, norm_a=%f, norm_b=%f -> similarity=%.4f",
                 list(common_tokens), dot, norm_a, norm_b, similarity)
    return similarity