tests/test_game.py

"""Headless tests for the Game wrapper, using the dummy policy (no GPU/Gradio).

Run:  python -m pytest tests/ -q
"""

import sys
from pathlib import Path

sys.path.insert(0, str(Path(__file__).resolve().parents[1]))

from ttw import narrate
from ttw.dummy import make_random_policy
from ttw.game import Game


def _game():
    return Game(make_random_policy(seed=3), deck_seed=1)


def test_baseline_recorded_before_any_step():
    g = _game()
    assert not g.price_frame().empty  # turn 0 prices recorded
    assert g.gini_frame().iloc[0]["turn"] == 0


def test_step_advances_and_grows_history():
    g = _game()
    rows_before = len(g.price_frame())
    g.step()
    assert g.world.turn == 1
    assert len(g.price_frame()) > rows_before
    assert g.gini_frame().iloc[-1]["turn"] == 1


def test_price_frame_shape_is_long_per_good():
    g = _game()
    g.step()
    pf = g.price_frame()
    assert set(pf.columns) == {"turn", "good", "price"}
    # one row per good per recorded turn (turn 0 baseline + turn 1)
    assert (pf["turn"] == 1).sum() == 5


def test_town_frame_lists_all_creatures_sorted_by_wealth():
    g = _game()
    g.step()
    tf = g.town_frame()
    assert len(tf) == 5
    nets = list(tf["net worth"])
    assert nets == sorted(nets, reverse=True)  # richest first


def test_tempt_fate_draws_legend_and_tickers():
    g = _game()
    title = g.tempt_fate()
    assert isinstance(title, str) and title
    assert any("inspired by" in line for line in g.ticker)


def test_shock_gold_rush_echoes_to_ticker():
    g = _game()
    before = sum(c.pebbles for c in g.world.alive())
    g.shock("gold_rush", amount=10)
    assert sum(c.pebbles for c in g.world.alive()) == before + 10 * 5
    assert any("windfall" in line.lower() for line in g.ticker)


def test_traces_markdown_handles_dummy_policy_gracefully():
    """Dummy policy has no raw thoughts; traces view must not crash."""
    g = _game()
    g.step()
    assert isinstance(g.traces_markdown(), str)  # no exception, returns a string


def test_legend_markdown_before_and_after_tempt():
    g = _game()
    assert "No legend yet" in g.legend_markdown()
    g.tempt_fate()
    md = g.legend_markdown()
    assert "Inspired by" in md and "✨" in md
    assert g.current_legend is not None


def test_avatar_and_mood_emoji():
    assert narrate.avatar("Fenn") == "🦊"
    assert narrate.avatar("Nobody") == "🐾"  # fallback
    assert narrate.mood_emoji(10) == "😊"
    assert narrate.mood_emoji(4) == "😣"


def test_town_frame_has_avatars_and_mood_faces():
    g = _game()
    g.step()
    tf = g.town_frame()
    assert any("🦊" in str(v) for v in tf["creature"])  # avatars present
    assert all(any(face in str(m) for face in "😊🙂😟😣") for m in tf["mood"])


# Phase 6 operator-console HTML accessors (arity is enforced in app.py).
_FIREWALL_BANNED = ("true", "false", "truth", "legend_key")


def test_phase6_accessors_return_str():
    g = _game()
    assert isinstance(g.heat_bar_html(), str)
    assert isinstance(g.relationship_map_html(), str)
    assert isinstance(g.cutscene_html(), str)


def test_heat_bar_html_well_formed():
    from ttw.world import INVESTIGATION_THRESHOLD

    g = _game()
    html = g.heat_bar_html()
    assert html  # non-empty
    assert str(INVESTIGATION_THRESHOLD) in html  # threshold readout, not hardcoded
    assert "color:#" in html  # inline color set
    assert not any(b in html for b in _FIREWALL_BANNED)  # truth firewall intact


def test_relationship_map_html_covers_cast_and_patron():
    g = _game()
    # Drive a couple of moves so at least one sentiment cell is nonzero.
    g.queue_player_move("bribe", target="Fenn", amount=40)
    g.step()
    g.queue_player_move("short", good="berries", amount=20)
    g.step()
    html = g.relationship_map_html()
    assert "<table" in html
    assert "Patron" in html
    for name in g.world.creatures:
        assert name in html
    assert "#2e2519" in html  # forced dark text on the light cells


def test_cutscene_html_never_blank():
    g = _game()
    before = g.cutscene_html()
    assert before  # placeholder before any events, never blank
    assert "The wood is still." in before
    g.step()
    after = g.cutscene_html()
    assert after
    assert "#fff8e6" not in after  # the dark surface is never lightened
    assert "#161b22" in after  # and is the intended dark playbill background


# Judge-facing polish: lab attribution, the story spine, and the raw-output panel.
def test_creature_lab_color_maps_engine_and_falls_back():
    from ttw.council import creature_lab_color

    assert creature_lab_color("Fenn") == "#76b900"  # NVIDIA green (nemotron)
    assert creature_lab_color("Oona") == "#10a37f"  # OpenAI teal (gptoss)
    assert creature_lab_color("Nobody") == "#cdbf9a"  # neutral fallback


def test_chapter_markdown_is_html_stepper_with_one_active():
    g = _game()
    spine = g.chapter_markdown()
    # All five chapter short-titles are present as pills.
    for title in ("First Whispers", "The Heat Rises", "Reckoning", "The Long Game"):
        assert title in spine
    # Exactly one pill is the active one (the accent underline appears once).
    assert spine.count("border-bottom:3px solid #2f6b3d") == 1


def test_traces_markdown_tags_the_model():
    g = _game()
    # Thoughts are normally filled from policy.state["raw"]; set them directly to
    # test the model tag rendering without a live council.
    g.thoughts = {"Oona": "hold the line", "Pip": "buy acorns"}
    md = g.traces_markdown()
    assert "[gpt-oss-20B]" in md  # Oona's model
    assert "[ttw-trader-0.5B]" in md  # Pip's model


def test_raw_output_markdown_empty_state():
    g = _game()  # dummy policy exposes no .state
    md = g.raw_output_markdown()
    assert isinstance(md, str)
    assert "No raw model output" in md


def test_raw_output_markdown_tags_and_fence_safe():
    g = _game()
    g.policy.state = {"raw": {"Oona": '{"thought":"x"} ``` # inject'}}
    md = g.raw_output_markdown()
    assert "gpt-oss-20B" in md  # tagged by model
    assert "```json" in md  # rendered as a fenced block
    # Exactly the open/close fence remain; a stray fence in model output that
    # would break out and inject markup is neutralized.
    assert md.count("```") == 2