#!/usr/bin/env python
# coding=utf8

from __future__ import unicode_literals
from datetime import timedelta
import collections
import functools
import os
import re
import string
from io import StringIO

import pytest
from hypothesis import given, settings, HealthCheck, assume, example
import hypothesis.strategies as st

import srt

REGISTER_SETTINGS = lambda name, **kwargs: settings.register_profile(
    name, suppress_health_check=[HealthCheck.too_slow], deadline=None, **kwargs
)

REGISTER_SETTINGS("base")
REGISTER_SETTINGS("release", max_examples=1000)

settings.load_profile(os.getenv("HYPOTHESIS_PROFILE", "base"))

HOURS_IN_DAY = 24
TIMEDELTA_MAX_DAYS = 999999999
CONTENTLESS_SUB = functools.partial(
    srt.Subtitle, index=1, start=timedelta(seconds=1), end=timedelta(seconds=2)
)


def is_strictly_legal_content(content):
    """
    Filter out things that would violate strict mode. Illegal content
    includes:

    - A content section that starts or ends with a newline
    - A content section that contains blank lines
    """

    if content.strip("\r\n") != content:
        return False
    elif not content.strip():
        return False
    elif "\n\n" in content:
        return False
    else:
        return True


def subs_eq(got, expected, any_order=False):
    """
    Compare Subtitle objects using vars() so that differences are easy to
    identify.
    """
    got_vars = [frozenset(vars(sub).items()) for sub in got]
    expected_vars = [frozenset(vars(sub).items()) for sub in expected]
    if any_order:
        assert collections.Counter(got_vars) == collections.Counter(expected_vars)
    else:
        assert got_vars == expected_vars


def timedeltas(min_value=0, max_value=TIMEDELTA_MAX_DAYS):
    """
    A Hypothesis strategy to generate timedeltas.

    Right now {min,max}_value are shoved into multiple fields in timedelta(),
    which is not very customisable, but it's good enough for our current test
    purposes. If you need more precise control, you may need to add more
    parameters to this function to be able to customise more freely.
    """
    time_unit_strategy = st.integers(min_value=min_value, max_value=max_value)
    timestamp_strategy = st.builds(
        timedelta,
        hours=time_unit_strategy,
        minutes=time_unit_strategy,
        seconds=time_unit_strategy,
    )
    return timestamp_strategy


def equivalent_timestamps(min_value=0, max_value=TIMEDELTA_MAX_DAYS):
    def string_timestamp(hours, minutes, seconds, msecs, paddings):
        hours, minutes, seconds, msecs = map(
            lambda v_and_p: "0" * v_and_p[1] + str(v_and_p[0]),
            zip((hours, minutes, seconds, msecs), paddings),
        )
        return "{}:{}:{},{}".format(hours, minutes, seconds, msecs)

    def ts_field_value():
        return st.integers(min_value=min_value, max_value=max_value)

    def zero_padding():
        return st.integers(min_value=0, max_value=2)

    @st.composite
    def maybe_off_by_one_fields(draw):
        field = draw(ts_field_value())
        field_maybe_plus_one = draw(st.integers(min_value=field, max_value=field + 1))
        return field_maybe_plus_one, field

    def get_equiv_timestamps(h, m, s, ms2, ts1paddings, ts2paddings):
        h2, h1 = h
        m2, m1 = m
        s2, s1 = s
        ms1 = (
            (h2 - h1) * 60 * 60 * 1000 + (m2 - m1) * 60 * 1000 + (s2 - s1) * 1000 + ms2
        )
        return (
            string_timestamp(h2, m2, s2, ms2, ts2paddings),
            string_timestamp(h1, m1, s1, ms1, ts1paddings),
        )

    return st.builds(
        get_equiv_timestamps,
        maybe_off_by_one_fields(),
        maybe_off_by_one_fields(),
        maybe_off_by_one_fields(),
        ts_field_value(),
        st.tuples(*[zero_padding() for _ in range(4)]),
        st.tuples(*[zero_padding() for _ in range(4)]),
    )


def subtitles(strict=True):
    """A Hypothesis strategy to generate Subtitle objects."""
    # max_value settings are just to avoid overflowing TIMEDELTA_MAX_DAYS by
    # using arbitrary low enough numbers.
    #
    # We also skip subs with start time >= end time, so we split them into two
    # groups to avoid overlap.
    start_timestamp_strategy = timedeltas(min_value=0, max_value=500000)
    end_timestamp_strategy = timedeltas(min_value=500001, max_value=999999)

    # \r is not legal inside Subtitle.content, it should have already been
    # normalised to \n.
    content_strategy = st.text(min_size=1).filter(lambda x: "\r" not in x)
    proprietary_strategy = st.text().filter(
        lambda x: all(eol not in x for eol in "\r\n")
    )

    if strict:
        content_strategy = content_strategy.filter(is_strictly_legal_content)

    subtitle_strategy = st.builds(
        srt.Subtitle,
        index=st.integers(min_value=0),
        start=start_timestamp_strategy,
        end=end_timestamp_strategy,
        proprietary=proprietary_strategy,
        content=content_strategy,
    )

    return subtitle_strategy


@given(st.lists(subtitles()))
def test_compose_and_parse_from_file(input_subs):
    srt_file = StringIO(srt.compose(input_subs, reindex=False))
    reparsed_subs = srt.parse(srt_file)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()))
def test_compose_and_parse_from_file_bom(input_subs):
    srt_file = StringIO("\ufeff" + srt.compose(input_subs, reindex=False))
    reparsed_subs = srt.parse(srt_file)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()))
def test_compose_and_parse_strict(input_subs):
    composed = srt.compose(input_subs, reindex=False)
    reparsed_subs = srt.parse(composed)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()))
def test_can_compose_without_ending_blank_line(input_subs):
    """
    Many sub editors don't add a blank line to the end, and many editors accept
    it. We should just accept this too in input.
    """
    composed = srt.compose(input_subs, reindex=False)
    composed_without_ending_blank = composed[:-1]
    reparsed_subs = srt.parse(composed_without_ending_blank)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()))
def test_can_compose_without_eol_at_all(input_subs):
    composed = srt.compose(input_subs, reindex=False)
    composed_without_ending_blank = composed.rstrip("\r\n")
    reparsed_subs = srt.parse(composed_without_ending_blank)
    subs_eq(reparsed_subs, input_subs)


@given(st.text().filter(is_strictly_legal_content))
def test_compose_and_parse_strict_mode(content):
    # sub.content should not have OS-specific line separators, only \n
    assume("\r" not in content)

    content = "\n" + content + "\n\n" + content + "\n"
    sub = CONTENTLESS_SUB(content=content)

    parsed_strict = list(srt.parse(sub.to_srt()))[0]
    parsed_unstrict = list(srt.parse(sub.to_srt(strict=False)))[0]

    # Strict mode should remove blank lines in content, leading, and trailing
    # newlines.
    assert not parsed_strict.content.startswith("\n")
    assert not parsed_strict.content.endswith("\n")
    assert "\n\n" not in parsed_strict.content

    # When strict mode is false, no processing should be applied to the
    # content (other than \r\n becoming \n).
    assert parsed_unstrict.content == sub.content.replace("\r\n", "\n")


@given(st.integers(min_value=1, max_value=TIMEDELTA_MAX_DAYS))
def test_timedelta_to_srt_timestamp_can_go_over_24_hours(days):
    srt_timestamp = srt.timedelta_to_srt_timestamp(timedelta(days=days))
    srt_timestamp_hours = int(srt_timestamp.split(":")[0])
    assert srt_timestamp_hours == days * HOURS_IN_DAY


@given(subtitles())
def test_subtitle_equality(sub_1):
    sub_2 = srt.Subtitle(**vars(sub_1))
    assert sub_1 == sub_2


@given(subtitles())
def test_subtitle_inequality(sub_1):
    sub_2 = srt.Subtitle(**vars(sub_1))
    sub_2.index += 1
    assert sub_1 != sub_2


@given(subtitles())
def test_subtitle_from_scratch_equality(subtitle):
    srt_block = subtitle.to_srt()

    # Get two totally new sets of objects so as not to affect the hash
    # comparison
    sub_1 = list(srt.parse(srt_block))[0]
    sub_2 = list(srt.parse(srt_block))[0]

    subs_eq([sub_1], [sub_2])
    # In case subs_eq and eq disagree for some reason
    assert sub_1 == sub_2
    assert hash(sub_1) == hash(sub_2)


@given(st.lists(subtitles()))
def test_parsing_spaced_arrow(subs):
    spaced_block = srt.compose(subs, reindex=False, strict=False).replace("-->", "- >")
    reparsed_subtitles = srt.parse(spaced_block)
    subs_eq(reparsed_subtitles, subs)


@given(st.lists(subtitles()))
def test_parsing_spaced_ender_arrow(subs):
    # Seen in BSG subtitles
    spaced_block = srt.compose(subs, reindex=False, strict=False).replace("-->", "-- >")
    reparsed_subtitles = srt.parse(spaced_block)
    subs_eq(reparsed_subtitles, subs)


@given(st.lists(subtitles()))
def test_parsing_no_ws_arrow(subs):
    spaced_block = srt.compose(subs, reindex=False, strict=False).replace(
        " --> ", "-->"
    )
    reparsed_subtitles = srt.parse(spaced_block)
    subs_eq(reparsed_subtitles, subs)


@given(st.text(string.whitespace), st.lists(subtitles()))
def test_parsing_leading_whitespace(ws, subs):
    prews_block = ws + srt.compose(subs, reindex=False, strict=False)
    reparsed_subtitles = srt.parse(prews_block)
    subs_eq(reparsed_subtitles, subs)


@given(st.lists(subtitles()))
def test_parsing_negative_index(subs):
    for sub in subs:
        sub.index *= -1
    prews_block = srt.compose(subs, reindex=False, strict=False)
    reparsed_subtitles = srt.parse(prews_block)
    subs_eq(reparsed_subtitles, subs)


@given(st.lists(subtitles()))
def test_parsing_content_with_blank_lines(subs):
    for subtitle in subs:
        # We stuff a blank line in the middle so as to trigger the "special"
        # content parsing for erroneous SRT files that have blank lines.
        subtitle.content = subtitle.content + "\n\n" + subtitle.content

    reparsed_subtitles = srt.parse(srt.compose(subs, reindex=False, strict=False))
    subs_eq(reparsed_subtitles, subs)


@given(st.lists(subtitles()))
def test_parsing_no_content(subs):
    for subtitle in subs:
        subtitle.content = ""

    reparsed_subtitles = srt.parse(srt.compose(subs, reindex=False, strict=False))
    subs_eq(reparsed_subtitles, subs)


@given(st.lists(subtitles()), st.lists(subtitles()), st.text(alphabet="\n\r\t "))
def test_subs_missing_content_removed(content_subs, contentless_subs, contentless_text):
    for sub in contentless_subs:
        sub.content = contentless_text

    subs = contentless_subs + content_subs
    composed_subs = list(srt.sort_and_reindex(subs, in_place=True))

    # We should have composed the same subs as there are in content_subs, as
    # all contentless_subs should have been stripped.
    subs_eq(composed_subs, content_subs, any_order=True)

    # The subtitles should be reindexed starting at start_index, excluding
    # contentless subs
    default_start_index = 1
    assert [sub.index for sub in composed_subs] == list(
        range(default_start_index, default_start_index + len(composed_subs))
    )


@given(
    st.lists(subtitles()),
    st.lists(subtitles()),
    timedeltas(min_value=-999, max_value=-1),
)
def test_subs_starts_before_zero_removed(positive_subs, negative_subs, negative_td):
    for sub in negative_subs:
        sub.start = negative_td
        sub.end = negative_td  # Just to avoid tripping any start >= end errors

    subs = positive_subs + negative_subs
    composed_subs = list(srt.sort_and_reindex(subs, in_place=True))

    # There should be no negative subs
    subs_eq(composed_subs, positive_subs, any_order=True)


@given(st.lists(subtitles(), min_size=1), st.integers(min_value=0))
def test_sort_and_reindex(input_subs, start_index):
    for sub in input_subs:
        # Pin all subs to same end time and index so that start time is
        # compared only, must be guaranteed to be < sub.start, see how
        # start_timestamp_strategy is done
        sub.end = timedelta(500001)
        sub.index = 1

    reindexed_subs = list(
        srt.sort_and_reindex(input_subs, start_index=start_index, in_place=True)
    )

    # The subtitles should be reindexed starting at start_index
    assert [sub.index for sub in reindexed_subs] == list(
        range(start_index, start_index + len(input_subs))
    )

    # The subtitles should be sorted by start time
    expected_sorting = sorted(input_subs, key=lambda sub: sub.start)
    assert reindexed_subs == expected_sorting


@given(st.lists(subtitles()))
def test_sort_and_reindex_no_skip(input_subs):
    # end time > start time should not trigger a skip if skip=False
    for sub in input_subs:
        old_start = sub.start
        sub.start = sub.end
        sub.end = old_start

    reindexed_subs = list(srt.sort_and_reindex(input_subs, skip=False))

    # Nothing should have been skipped
    assert len(reindexed_subs) == len(input_subs)


@given(st.lists(subtitles()))
def test_sort_and_reindex_handles_no_index(input_subs):
    # end time > start time should not trigger a skip if skip=False
    for sub in input_subs:
        old_start = sub.start
        sub.start = sub.end
        sub.end = old_start
        sub.index = None

    reindexed_subs = list(srt.sort_and_reindex(input_subs))

    # Everything should have been skipped
    assert not reindexed_subs


@given(st.lists(subtitles(), min_size=1))
def test_sort_and_reindex_same_start_time_uses_end(input_subs):
    for sub in input_subs:
        # Pin all subs to same start time and index so that end time is
        # compared only
        sub.start = timedelta(1)
        sub.index = 1

    reindexed_subs = list(srt.sort_and_reindex(input_subs, in_place=True))

    # The subtitles should be sorted by end time when start time is the same
    expected_sorting = sorted(input_subs, key=lambda sub: sub.end)
    assert reindexed_subs == expected_sorting


@given(st.lists(subtitles(), min_size=1))
def test_sort_and_reindex_same_start_and_end_time_uses_index(input_subs):
    for sub in input_subs:
        # Pin all subs to same start and end time so that index is compared
        # only
        sub.start = timedelta(1)
        sub.end = timedelta(2)

    reindexed_subs = list(srt.sort_and_reindex(input_subs, in_place=True))

    # The subtitles should be sorted by index when start and end time are the
    # same
    expected_sorting = sorted(input_subs, key=lambda sub: sub.index)
    assert reindexed_subs == expected_sorting


@given(st.lists(subtitles(), min_size=1), st.integers(min_value=0))
def test_sort_and_reindex_not_in_place_matches(input_subs, start_index):
    # Make copies for both sort_and_reindex calls so that they can't affect
    # each other
    not_in_place_subs = [srt.Subtitle(**vars(sub)) for sub in input_subs]
    in_place_subs = [srt.Subtitle(**vars(sub)) for sub in input_subs]

    nip_ids = [id(sub) for sub in not_in_place_subs]
    ip_ids = [id(sub) for sub in in_place_subs]

    not_in_place_output = list(
        srt.sort_and_reindex(not_in_place_subs, start_index=start_index)
    )
    in_place_output = list(
        srt.sort_and_reindex(in_place_subs, start_index=start_index, in_place=True)
    )

    # The results in each case should be the same
    subs_eq(not_in_place_output, in_place_output)

    # Not in place sort_and_reindex should have created new subs
    assert not any(id(sub) in nip_ids for sub in not_in_place_output)

    # In place sort_and_reindex should be reusing the same subs
    assert all(id(sub) in ip_ids for sub in in_place_output)


@given(
    st.lists(subtitles(), min_size=1),
    st.integers(min_value=0),
    st.text(min_size=1),
    timedeltas(),
)
def test_parser_noncontiguous(subs, fake_idx, garbage, fake_timedelta):
    composed = srt.compose(subs)

    # Put some garbage between subs that should trigger our failed parsing
    # detection. Since we do some magic to try and detect blank lines that
    # don't really delimit subtitles, it has to look at least a little like an
    # SRT block.
    srt_timestamp = srt.timedelta_to_srt_timestamp(fake_timedelta)
    composed = composed.replace(
        "\n\n", "\n\n%d\n%s %s" % (fake_idx, srt_timestamp, garbage)
    )

    with pytest.raises(srt.SRTParseError):
        list(srt.parse(composed))


@given(
    st.lists(subtitles(), min_size=1),
    st.integers(min_value=0),
    st.text(min_size=1),
    timedeltas(),
)
def test_parser_noncontiguous_ignore_errors(subs, fake_idx, garbage, fake_timedelta):
    composed = srt.compose(subs)
    srt_timestamp = srt.timedelta_to_srt_timestamp(fake_timedelta)
    composed = composed.replace(
        "\n\n", "\n\n%d\n%s %s" % (fake_idx, srt_timestamp, garbage)
    )
    # Should not raise, we have ignore_errors
    list(srt.parse(composed, ignore_errors=True))


def _parseable_as_int(text):
    try:
        int(text)
    except ValueError:
        return False
    return True


def _parseable_as_float(text):
    try:
        float(text)
    except ValueError:
        return False
    return True


@given(st.lists(subtitles()), st.text(min_size=1))
def test_parser_noncontiguous_leading(subs, garbage):
    # Issue #50 permits leading whitespace, see test_parsing_leading_whitespace
    assume(not garbage.isspace())

    # Issue #56 permits negative indexes, see test_parsing_negative_index. It
    # also shouldn't just be a number, because then we'd confuse it with our
    # index...
    assume(garbage.strip()[0] != ".")
    assume(garbage.strip()[0] != "-")
    assume(not _parseable_as_int(garbage.strip()))
    assume(not _parseable_as_float(garbage.strip()))

    # Put some garbage at the beginning that should trigger our noncontiguity
    # checks
    composed = garbage + srt.compose(subs)

    with pytest.raises(srt.SRTParseError):
        list(srt.parse(composed))


@given(
    st.lists(subtitles(), min_size=1),
    st.integers(min_value=0),
    st.text(min_size=1),
    timedeltas(),
)
def test_parser_didnt_match_to_end_raises(subs, fake_idx, garbage, fake_timedelta):
    srt_blocks = [sub.to_srt() for sub in subs]
    srt_timestamp = srt.timedelta_to_srt_timestamp(fake_timedelta)
    garbage = "\n\n%d\n%s %s" % (fake_idx, srt_timestamp, garbage)
    srt_blocks.append(garbage)
    composed = "".join(srt_blocks)

    with pytest.raises(srt.SRTParseError) as thrown_exc:
        list(srt.parse(composed))

    # Since we will consume as many \n as needed until we meet the lookahead
    # assertion, leading newlines in `garbage` will be stripped.
    garbage_stripped = garbage.lstrip("\n")

    assert garbage_stripped == thrown_exc.value.unmatched_content
    assert len(composed) - len(garbage_stripped) == thrown_exc.value.expected_start
    assert len(composed) == thrown_exc.value.actual_start


@given(st.lists(subtitles()))
def test_parser_can_parse_with_dot_msec_delimiter(subs):
    original_srt_blocks = [sub.to_srt() for sub in subs]
    dot_srt_blocks = []

    for srt_block in original_srt_blocks:
        srt_lines = srt_block.split("\n")
        # We should only do the first two, as it might also be in the
        # proprietary metadata, causing this test to fail.
        dot_timestamp = srt_lines[1].replace(",", ".", 2)
        srt_lines[1] = dot_timestamp
        dot_srt_blocks.append("\n".join(srt_lines))

    composed_with_dots = "".join(dot_srt_blocks)
    reparsed_subs = srt.parse(composed_with_dots)
    subs_eq(reparsed_subs, subs)


@given(st.lists(subtitles()))
def test_parser_can_parse_with_fullwidth_delimiter(subs):
    original_srt_blocks = [sub.to_srt() for sub in subs]
    dot_srt_blocks = []

    for srt_block in original_srt_blocks:
        srt_lines = srt_block.split("\n")
        dot_timestamp = srt_lines[1].replace(",", "，", 1).replace(":", "：", 1)
        srt_lines[1] = dot_timestamp
        dot_srt_blocks.append("\n".join(srt_lines))

    composed_with_fullwidth = "".join(dot_srt_blocks)
    reparsed_subs = srt.parse(composed_with_fullwidth)
    subs_eq(reparsed_subs, subs)


@given(st.lists(subtitles()))
def test_parser_can_parse_with_no_msec(subs):
    original_srt_blocks = [sub.to_srt() for sub in subs]
    srt_blocks = []

    for srt_block in original_srt_blocks:
        srt_lines = srt_block.split("\n")
        # We should only do the first two, as it might also be in the
        # proprietary metadata, causing this test to fail.
        srt_lines[1] = re.sub(",[0-9]+", "", srt_lines[1], 2)
        srt_blocks.append("\n".join(srt_lines))

    composed = "".join(srt_blocks)
    reparsed_subs = srt.parse(composed)
    subs_eq(reparsed_subs, subs)


@given(subtitles())
def test_repr_doesnt_crash(sub):
    # Not much we can do here, but we should make sure __repr__ doesn't crash
    # or anything and it does at least vaguely look like what we want
    assert "Subtitle" in repr(sub)
    assert str(sub.index) in repr(sub)


@given(subtitles(), subtitles())
def test_parser_accepts_final_no_newline_no_content(sub1, sub2):
    # Limit size so we know how much to remove
    sub2.content = ""
    subs = [sub1, sub2]

    # Remove the last newlines so that there are none. Cannot use rstrip since
    # there might be other stuff that gets matched in proprietary
    stripped_srt_blocks = srt.compose(subs, reindex=False)[:-2]

    reparsed_subs = srt.parse(stripped_srt_blocks)
    subs_eq(reparsed_subs, subs)


@given(st.lists(subtitles()))
def test_parser_accepts_newline_no_content(subs):
    for sub in subs:
        # Limit size so we know how many lines to remove
        sub.content = ""

    # Remove the last \n so that there is only one
    stripped_srt_blocks = "".join(sub.to_srt()[:-1] for sub in subs)

    reparsed_subs = srt.parse(stripped_srt_blocks)
    subs_eq(reparsed_subs, subs)


@given(st.lists(subtitles()))
def test_compose_and_parse_strict_crlf(input_subs):
    composed_raw = srt.compose(input_subs, reindex=False)
    composed = composed_raw.replace("\n", "\r\n")
    reparsed_subs = list(srt.parse(composed))

    for sub in reparsed_subs:
        sub.content = sub.content.replace("\r\n", "\n")

    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()), st.one_of(st.just("\n"), st.just("\r\n")))
def test_compose_and_parse_strict_custom_eol(input_subs, eol):
    composed = srt.compose(input_subs, reindex=False, eol=eol)
    reparsed_subs = srt.parse(composed)
    subs_eq(reparsed_subs, input_subs)


@given(equivalent_timestamps())
def test_equal_timestamps_despite_different_fields_parsed_as_equal(timestamps):
    ts1, ts2 = timestamps
    assert srt.srt_timestamp_to_timedelta(ts1) == srt.srt_timestamp_to_timedelta(ts2)


@given(timedeltas())
def test_bad_timestamp_format_raises(ts):
    ts = srt.timedelta_to_srt_timestamp(ts)
    ts = ts.replace(":", "t", 1)
    with pytest.raises(srt.TimestampParseError):
        srt.srt_timestamp_to_timedelta(ts)


@given(st.lists(subtitles()), st.lists(st.sampled_from(string.whitespace)))
def test_can_parse_index_trailing_ws(input_subs, whitespace):
    out = ""

    for sub in input_subs:
        lines = sub.to_srt().split("\n")
        lines[0] = lines[0] + "".join(whitespace)
        out += "\n".join(lines)

    reparsed_subs = srt.parse(out)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()))
def test_can_parse_index_with_dot(input_subs):
    # Seen in Battlestar Galactica subs
    out = ""

    for sub in input_subs:
        lines = sub.to_srt().split("\n")
        lines[0] = lines[0] + "." + lines[0]
        out += "\n".join(lines)

    reparsed_subs = srt.parse(out)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles()), st.lists(st.just("0")))
def test_can_parse_index_leading_zeroes(input_subs, zeroes):
    out = ""

    for sub in input_subs:
        lines = sub.to_srt().split("\n")
        lines[0] = "".join(zeroes) + lines[0]
        out += "\n".join(lines)

    reparsed_subs = srt.parse(out)
    subs_eq(reparsed_subs, input_subs)


@given(st.lists(subtitles(), min_size=1))
def test_parse_file_with_missing_index(input_subs):  # cf. issue #51
    for sub in input_subs:
        try:
            int(sub.content.strip().split("\n")[-1])
        except ValueError:
            pass
        else:
            # If the final line with actual content is a number, we'll parse it
            # as the index, so ignore that
            assume(False)

    out_no_index = ""
    out_zero_index = ""

    for sub in input_subs:
        block = sub.to_srt()
        block = block[block.index("\n") + 1 :]
        out_no_index += block

    input_subs_copy = [srt.Subtitle(**vars(sub)) for sub in input_subs]
    for sub in input_subs_copy:
        # sub.index == None will get rendered in to_srt as 0
        sub.index = 0
        out_zero_index += sub.to_srt()

    subs_no_index = list(srt.parse(out_no_index))
    subs_zero_index = list(srt.parse(out_zero_index))

    # One should have index == None, one should have index == 0...
    assert subs_no_index
    assert subs_zero_index
    assert all(sub.index == None for sub in subs_no_index)
    assert all(sub.index == 0 for sub in subs_zero_index)
    assert subs_no_index != subs_zero_index

    # ...but they should render the same...
    assert srt.compose(subs_no_index, reindex=False) == srt.compose(
        subs_zero_index, reindex=False
    )

    # ...and sort the same.
    assert srt.compose(subs_no_index) == srt.compose(subs_zero_index)