from agr_assistent.tts.text import SpeechTextStream, normalize_for_speech, split_long_text


def _feed_all(chunks: list[str]) -> list[str]:
    stream = SpeechTextStream()
    result = []
    for chunk in chunks:
        result.extend(stream.feed(chunk))
    return result + stream.flush()


def test_sentences_are_emitted_as_soon_as_complete() -> None:
    stream = SpeechTextStream()

    assert stream.feed("Привет! Как ") == ["Привет!"]
    assert stream.feed("дела? Хоро") == ["Как дела?"]
    assert stream.flush() == ["Хоро"]


def test_decimal_numbers_and_list_markers_do_not_split() -> None:
    assert _feed_all(["Будет 3.", "5 градуса.\n", "1. Первый пункт"]) == [
        "Будет три целых пять десятых градуса.",
        "Первый пункт",
    ]


def test_code_blocks_are_skipped() -> None:
    chunks = ["Пример:\n``", "`python\nprint('Привет')\n", "```\nГотово."]

    assert _feed_all(chunks) == ["Пример:", "Готово."]


def test_normalize_removes_markdown_links_and_latin() -> None:
    text = "**Важно**: смотри [документацию](https://x.io) и https://y.io, Python тоже 😀"

    assert normalize_for_speech(text) == "Важно: смотри документацию и ссылка, тоже"


def test_text_without_cyrillic_is_not_spoken() -> None:
    assert normalize_for_speech("Hello, world! 👋") == ""
    assert _feed_all(["Hello world.\n"]) == []


def test_long_text_is_split_on_commas_and_spaces() -> None:
    text = ", ".join(["слово"] * 100)

    parts = split_long_text(text, limit=50)

    assert all(len(part) <= 50 for part in parts)
    assert " ".join(parts).replace(" ,", ",") == text