"""Tests für die Auto-Learning-Pattern-Library."""

from __future__ import annotations

import json

import pytest


@pytest.fixture
def tmp_lib(tmp_path, monkeypatch):
    p = tmp_path / "patterns.json"
    monkeypatch.setenv("AGENT_PATTERN_LIBRARY", str(p))
    import compliance.services.specialist_agents._pattern_library as lib
    lib._invalidate_cache()
    yield lib, p
    lib._invalidate_cache()


def test_record_creates_file(tmp_lib):
    lib, p = tmp_lib
    assert not p.exists()
    lib.record("kontakt_telefon", "Telefonnr.", 0.9, "impressum")
    assert p.exists()
    data = json.loads(p.read_text())
    assert len(data["patterns"]) == 1
    assert data["patterns"][0]["label_used"] == "Telefonnr."
    assert data["patterns"][0]["observed_count"] == 1


def test_record_increments_existing(tmp_lib):
    lib, _ = tmp_lib
    lib.record("kontakt_telefon", "Telefonnr.", 0.9, "impressum")
    lib.record("kontakt_telefon", "Telefonnr.", 0.85, "impressum")
    lib.record("kontakt_telefon", "telefonnr.", 0.8, "impressum")  # case-i
    raws = lib.list_all()
    assert len(raws) == 1
    assert raws[0]["observed_count"] == 3


def test_record_separate_per_field_id(tmp_lib):
    lib, _ = tmp_lib
    lib.record("kontakt_telefon", "Tel", 0.9, "impressum")
    lib.record("kontakt_email", "Tel", 0.9, "impressum")
    assert len(lib.list_all()) == 2


def test_record_empty_inputs_noop(tmp_lib):
    lib, p = tmp_lib
    lib.record("", "Tel", 0.9, "impressum")
    lib.record("kontakt_telefon", "", 0.9, "impressum")
    lib.record("kontakt_telefon", "Tel", 0.9, "")
    assert not p.exists()


def test_load_patterns_returns_compiled_regex(tmp_lib):
    lib, _ = tmp_lib
    lib.record("kontakt_telefon", "Telefonnr.", 0.9, "impressum")
    pats = lib.load_patterns_for("kontakt_telefon", "impressum")
    assert len(pats) == 1
    m = pats[0].search("Hier: Telefonnr. 0761/12345")
    assert m is not None


def test_load_patterns_filters_low_confidence(tmp_lib):
    lib, _ = tmp_lib
    lib.record("kontakt_telefon", "WeakLabel", 0.3, "impressum")
    pats = lib.load_patterns_for(
        "kontakt_telefon", "impressum", min_avg_confidence=0.5,
    )
    assert pats == []
    # observed_count filter
    pats = lib.load_patterns_for(
        "kontakt_telefon", "impressum", min_observed=2,
    )
    assert pats == []


def test_label_to_regex_telefon():
    from compliance.services.specialist_agents._pattern_library import (
        _label_to_regex,
    )
    rx = _label_to_regex("Telefonnr.")
    import re
    assert re.search(rx, "Telefonnr. 0761/12345", re.I)
    assert re.search(rx, "Telefonnr 0761", re.I)


def test_label_to_regex_email():
    from compliance.services.specialist_agents._pattern_library import (
        _label_to_regex,
    )
    rx = _label_to_regex("Mailadresse")
    import re
    assert re.search(rx, "Mailadresse: x@y.de", re.I)


def test_prune_low_confidence_keeps_recent(tmp_lib):
    lib, _ = tmp_lib
    lib.record("kontakt_telefon", "Tel", 0.9, "impressum")
    pruned = lib.prune_low_confidence(min_runs_before_prune=100)
    assert pruned == 0  # Nur einmal observed → noch nicht prunen
    assert len(lib.list_all()) == 1


def test_load_patterns_for_nonexistent_returns_empty(tmp_lib):
    lib, _ = tmp_lib
    assert lib.load_patterns_for("ghost", "impressum") == []