import numpy as np

from shell_history_search.core import EmbeddingService


class TestEmbeddingService:
    def test_init_default(self):
        service = EmbeddingService()
        assert service.model_name == "all-MiniLM-L6-v2"
        assert service.device == "cpu"

    def test_init_custom_model(self, temp_cache_dir):
        service = EmbeddingService(
            model_name="all-MiniLM-L6-v2",
            cache_dir=temp_cache_dir,
        )
        assert service.model_name == "all-MiniLM-L6-v2"
        assert service.cache_dir == temp_cache_dir

    def test_embedding_dim(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        assert service.embedding_dim == 384

    def test_encode_single(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        embedding = service.encode_single("git commit")

        assert isinstance(embedding, np.ndarray)
        assert embedding.shape == (384,)
        assert embedding.dtype == np.float32

    def test_encode_batch(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        embeddings = service.encode(["git add .", "git commit", "git push"])

        assert isinstance(embeddings, np.ndarray)
        assert embeddings.shape == (3, 384)
        assert embeddings.dtype == np.float32

    def test_encode_empty_list(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        embeddings = service.encode([])

        assert isinstance(embeddings, np.ndarray)
        assert embeddings.shape == (0,)

    def test_encode_returns_normalized(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        embedding = service.encode_single("test command")

        norm = np.linalg.norm(embedding)
        assert 0.99 < norm <= 1.01

    def test_embedding_to_blob(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        embedding = service.encode_single("test")

        blob = EmbeddingService.embedding_to_blob(embedding)
        assert isinstance(blob, bytes)
        assert len(blob) == 384 * 4

    def test_blob_to_embedding(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        embedding = service.encode_single("test")
        blob = EmbeddingService.embedding_to_blob(embedding)

        recovered = EmbeddingService.blob_to_embedding(blob, 384)

        assert np.allclose(embedding, recovered)

    def test_cosine_similarity(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        e1 = service.encode_single("git commit")
        e2 = service.encode_single("git add .")
        e3 = service.encode_single("docker run")

        sim_same = EmbeddingService.cosine_similarity(e1, e2)
        sim_diff = EmbeddingService.cosine_similarity(e1, e3)

        assert -1 <= sim_same <= 1
        assert -1 <= sim_diff <= 1
        assert sim_same > sim_diff

    def test_cosine_similarity_perfect_match(self, temp_cache_dir):
        service = EmbeddingService(cache_dir=temp_cache_dir)
        e1 = service.encode_single("same command")

        sim = EmbeddingService.cosine_similarity(e1, e1)
        assert 0.9999 < sim <= 1.0001