"""군집화 기반 표절 판별 단위테스트 (계획서 2단계 고도화).

핵심 시나리오: "본문(lemma)·키워드는 그대로 두고 인물만 바꾼 표절" 을
element_swap_plagiarism 으로 분해해서 잡는다.
"""
from __future__ import annotations

from app.api.schemas import ExtractedElements
from app.engine.clustering import (
    ClusterIndex,
    signature_similarity,
    _element_signature,
)


def _elem(characters=None, motifs=None, keywords=None, genre=None) -> ExtractedElements:
    return ExtractedElements(
        characters=characters or [],
        motifs=motifs or [],
        keywords=keywords or [],
        genre=genre,
    )


def test_signature_similarity_identical():
    a = _element_signature(_elem(characters=["홍길동"], keywords=["활빈당", "탐관오리"], genre="역사"))
    assert signature_similarity(a, a) == 1.0


def test_clusters_group_similar_docs():
    # 두 홍길동 변형은 한 군집, 어린왕자는 별도 군집
    ids = ["hong-a", "hong-b", "prince"]
    elems = [
        _elem(characters=["홍길동"], keywords=["활빈당", "탐관오리", "재물"], genre="역사"),
        _elem(characters=["홍길동"], keywords=["활빈당", "탐관오리", "재물"], genre="역사"),
        _elem(characters=["어린왕자", "여우"], keywords=["사막", "장미", "별"], genre="동화"),
    ]
    idx = ClusterIndex(ids, elems, link_threshold=0.35)
    assert idx.cluster_of("hong-a") == idx.cluster_of("hong-b")
    assert idx.cluster_of("prince") != idx.cluster_of("hong-a")


def test_route_to_nearest_cluster():
    ids = ["hong", "prince"]
    elems = [
        _elem(characters=["홍길동"], keywords=["활빈당", "탐관오리", "재물"], genre="역사"),
        _elem(characters=["어린왕자"], keywords=["사막", "장미", "별"], genre="동화"),
    ]
    idx = ClusterIndex(ids, elems, link_threshold=0.9)  # 분리 유지
    q = _elem(characters=["홍길동"], keywords=["활빈당", "재물"], genre="역사")
    routed = idx.route(q)
    assert routed is not None
    assert "hong" in routed.members


def test_element_swap_plagiarism_detected():
    """본문 lemma·키워드 유지 + 인물만 교체 → element_swap_plagiarism."""
    ids = ["original"]
    elems = [_elem(characters=["홍길동"], motifs=["복수"], keywords=["활빈당", "탐관오리", "재물", "의적"])]
    lemmas = [["만들다", "빼앗다", "재물", "탐관오리", "활빈당", "나누다"]]
    idx = ClusterIndex(ids, elems, doc_lemmas=lemmas, link_threshold=0.35)

    # 인물(홍길동→김민수)·모티프만 바꾸고 본문 lemma/키워드는 그대로
    q_elem = _elem(characters=["김민수"], motifs=["정의"], keywords=["활빈당", "탐관오리", "재물", "의적"])
    q_lemmas = ["만들다", "빼앗다", "재물", "탐관오리", "활빈당", "나누다"]

    sig = idx.partial_signal("original", q_elem, q_lemmas)
    assert sig is not None
    assert sig.verdict == "element_swap_plagiarism", sig.per_element
    assert "characters" in sig.changed_elements
    assert "lemmas" in sig.retained_elements


def test_near_duplicate_verdict():
    ids = ["original"]
    elems = [_elem(characters=["홍길동"], keywords=["활빈당", "탐관오리", "재물"])]
    lemmas = [["만들다", "빼앗다", "재물"]]
    idx = ClusterIndex(ids, elems, doc_lemmas=lemmas)

    sig = idx.partial_signal(
        "original",
        _elem(characters=["홍길동"], keywords=["활빈당", "탐관오리", "재물"]),
        ["만들다", "빼앗다", "재물"],
    )
    assert sig is not None
    assert sig.verdict == "near_duplicate"


def test_unrelated_gets_none_or_weak():
    ids = ["original"]
    elems = [_elem(characters=["홍길동"], keywords=["활빈당", "탐관오리"])]
    lemmas = [["만들다", "빼앗다"]]
    idx = ClusterIndex(ids, elems, doc_lemmas=lemmas)

    sig = idx.partial_signal(
        "original",
        _elem(characters=["우주비행사"], keywords=["로켓", "행성"]),
        ["날다", "탐사하다"],
    )
    assert sig is not None
    assert sig.verdict in ("none", "weak")
    assert sig.verdict != "element_swap_plagiarism"


def test_empty_index():
    idx = ClusterIndex([], [])
    assert idx.num_clusters == 0
    assert idx.route(_elem(characters=["x"])) is None
    assert idx.candidate_ids(_elem(characters=["x"])) == set()