"""Tests for `app.core.embeddings.HashEmbedder`."""

from __future__ import annotations

import math

import pytest

from app.core.embeddings import HashEmbedder


@pytest.mark.asyncio
async def test_hash_embedder_returns_correct_dimension():
    emb = HashEmbedder(dimension=128)
    vecs = await emb.embed(["hello world"])
    assert len(vecs) == 1
    assert len(vecs[0]) == 128


@pytest.mark.asyncio
async def test_hash_embedder_empty_text_returns_zero_vector():
    emb = HashEmbedder(dimension=64)
    vecs = await emb.embed([""])
    assert vecs[0] == [0.0] * 64


@pytest.mark.asyncio
async def test_hash_embedder_deterministic():
    emb = HashEmbedder(dimension=64)
    v1 = (await emb.embed(["the quick brown fox"]))[0]
    v2 = (await emb.embed(["the quick brown fox"]))[0]
    assert v1 == v2


@pytest.mark.asyncio
async def test_hash_embedder_normalized():
    emb = HashEmbedder(dimension=64)
    v = (await emb.embed(["some text with multiple words for hashing"]))[0]
    norm = math.sqrt(sum(x * x for x in v))
    assert abs(norm - 1.0) < 1e-6


@pytest.mark.asyncio
async def test_hash_embedder_similar_texts_have_overlap():
    """Texts sharing tokens should have non-zero cosine similarity."""
    emb = HashEmbedder(dimension=256)
    v1 = (await emb.embed(["the dragon attacks the village"]))[0]
    v2 = (await emb.embed(["the dragon breathes fire"]))[0]
    v3 = (await emb.embed(["quantum mechanics equations"]))[0]
    # Cosine similarity (vectors are already normalized)
    sim_12 = sum(a * b for a, b in zip(v1, v2))
    sim_13 = sum(a * b for a, b in zip(v1, v3))
    # Shared-token texts should be more similar than disjoint ones
    assert sim_12 > sim_13


@pytest.mark.asyncio
async def test_hash_embedder_batch():
    emb = HashEmbedder(dimension=32)
    vecs = await emb.embed(["a", "b", "c"])
    assert len(vecs) == 3
    assert all(len(v) == 32 for v in vecs)


def test_hash_embedder_invalid_dimension():
    with pytest.raises(ValueError):
        HashEmbedder(dimension=0)
    with pytest.raises(ValueError):
        HashEmbedder(dimension=-1)