Datasets:

Larrivhan
/

NeCast-simulations

from __future__ import annotations

import dataclasses

import numpy as np
import pytest

from simulator.config import (
    DEMOGRAPHY_PROB_FIELDS,
    FLOAT32_MIN_POSITIVE,
    INT16_MAX,
    SOURCE_PROB_FIELDS,
    UINT32_MAX,
    Config,
    apply_preset,
    validate_config,
)
from simulator.noise import flat_index_dtype


def valid_config() -> Config:
    return apply_preset(Config(preset="mini"))


@pytest.mark.parametrize("preset", ["mini", "core_v01", "full_v01"])
def test_builtin_presets_are_valid(preset: str) -> None:
    validate_config(apply_preset(Config(preset=preset)))


@pytest.mark.parametrize("preset", ["mini", "core_v01", "full_v01"])
def test_builtin_presets_target_64_bins(preset: str) -> None:
    cfg = apply_preset(Config(preset=preset))

    assert cfg.time_bins == 64
    assert cfg.demography_resolution_log10 > 0.0


@pytest.mark.parametrize(
    ("changes", "message"),
    [
        ({"n_samples": 0}, "n_samples"),
        ({"n_samples": True}, "n_samples"),
        ({"shard_size": 0}, "shard_size"),
        ({"compute_chunk_size": 0}, "compute_chunk_size"),
        ({"seq_len": 0}, "seq_len"),
        ({"time_bins": 0}, "time_bins"),
        ({"demography_resolution_log10": 0.0}, "demography_resolution_log10"),
        ({"demography_resolution_log10": -1.0}, "demography_resolution_log10"),
        ({"n_haplotypes": 1}, "at least 2"),
        ({"n_haplotypes": 3}, "even"),
        ({"n_haplotypes": INT16_MAX + 1}, "int16"),
        ({"min_time": 0.0}, "min_time"),
        ({"min_time": 500.0, "max_time": 500.0}, "max_time"),
        ({"min_time": float("nan")}, "min_time"),
        ({"baseline_mu": 0.0}, "baseline_mu"),
        ({"baseline_rec": -1.0e-8}, "baseline_rec"),
        ({"baseline_mu": FLOAT32_MIN_POSITIVE / 2.0}, "float32"),
        ({"baseline_mu": 5.1e-7}, "synthetic maps"),
        ({"baseline_rec": 1.1e-4}, "baseline_rec"),
        ({"p_clean_constant": -0.01}, "p_clean_constant"),
        ({"genotype_error_min": -0.01}, "genotype_error_min"),
        ({"genotype_error_max": 1.01}, "genotype_error_max"),
        ({"genotype_error_min": 0.2, "genotype_error_max": 0.1}, "must not exceed"),
        ({"missing_rate_min": -0.01}, "missing_rate_min"),
        ({"missing_rate_max": 1.01}, "missing_rate_max"),
        ({"missing_rate_min": 0.2, "missing_rate_max": 0.1}, "must not exceed"),
        ({"phase_switch_rate_per_het_min": -0.1}, "phase_switch_rate_per_het_min"),
        (
            {"phase_switch_rate_per_het_min": 0.2, "phase_switch_rate_per_het_max": 0.1},
            "must not exceed",
        ),
        ({"hotspot_missing_prob": -0.1}, "hotspot_missing_prob"),
        ({"hotspot_missing_prob": 1.1}, "hotspot_missing_prob"),
        ({"observed_map_noise": -0.1}, "observed_map_noise"),
        ({"observed_map_noise": float("inf")}, "observed_map_noise"),
        ({"empirical_ne_template_log10_shift_sd": -0.1}, "log10_shift"),
        ({"empirical_ne_template_contrast_sd": -0.1}, "contrast"),
        ({"empirical_ne_template_time_warp_sd": -0.1}, "time_warp"),
        ({"empirical_ne_template_bin_noise_sd": -0.1}, "bin_noise"),
        ({"map_segments_min": 0}, "map_segments_min"),
        ({"map_segments_min": 10, "map_segments_max": 9}, "map_segments_min"),
        ({"max_retries": 0}, "max_retries"),
        ({"workers": 0}, "workers"),
        ({"workers": -2}, "workers"),
        ({"compression_level": -1}, "compression_level"),
        ({"compression_level": 10}, "compression_level"),
        ({"compression_level": True}, "compression_level"),
        ({"seed": 1.5}, "seed"),
        ({"dataset_name": ""}, "dataset_name"),
        ({"preset": "unknown"}, "preset"),
        ({"out_dir": ""}, "out_dir"),
        ({"out_dir": "."}, "dedicated dataset directory"),
        ({"demography_mixture_version": ""}, "demography_mixture_version"),
        ({"empirical_ne_template_dir": ""}, "empirical_ne_template_dir"),
        ({"stdpopsim_species": ""}, "stdpopsim_species"),
        ({"stdpopsim_contig_length": -1}, "stdpopsim_contig_length"),
        ({"recomb_map_unit": "centimorgans"}, "recomb_map_unit"),
        ({"mut_map_unit": "unknown"}, "mut_map_unit"),
    ],
)
def test_invalid_scalar_and_range_settings_are_rejected(changes: dict, message: str) -> None:
    with pytest.raises(ValueError, match=message):
        validate_config(dataclasses.replace(valid_config(), **changes))


def test_source_mixture_requires_a_positive_weight() -> None:
    cfg = dataclasses.replace(valid_config(), **{name: 0.0 for name in SOURCE_PROB_FIELDS})
    with pytest.raises(ValueError, match="source probability"):
        validate_config(cfg)


def test_disabled_stdpopsim_weight_is_not_an_enabled_source() -> None:
    source_weights = {name: 0.0 for name in SOURCE_PROB_FIELDS}
    source_weights["p_stdpopsim_anchor"] = 1.0
    cfg = dataclasses.replace(valid_config(), **source_weights, enable_stdpopsim=False)
    with pytest.raises(ValueError, match="enabled source probability"):
        validate_config(cfg)

    validate_config(dataclasses.replace(cfg, enable_stdpopsim=True))


def test_demography_mixture_requires_a_positive_weight() -> None:
    cfg = dataclasses.replace(valid_config(), **{name: 0.0 for name in DEMOGRAPHY_PROB_FIELDS})
    with pytest.raises(ValueError, match="demography probability"):
        validate_config(cfg)


def test_sequence_positions_must_fit_uint32_storage() -> None:
    with pytest.raises(ValueError, match="uint32 position"):
        validate_config(dataclasses.replace(valid_config(), seq_len=UINT32_MAX + 1))


@pytest.mark.parametrize(
    ("seq_len", "n_haplotypes"),
    [(60_000_000, 64), (80_000_000, 128), (80_000_000, 256), (40_000_000, 256)],
)
def test_a_long_sequence_with_many_haplotypes_is_accepted(
    seq_len: int, n_haplotypes: int
) -> None:
    """The sparse missing index addresses variants, not bases.
    Every configuration here was rejected by a precondition that multiplied
    ``seq_len`` by ``n_haplotypes``, which is the size of the matrix only if
    every base is segregating.  Variants are three orders of magnitude rarer,
    and the width that is actually used is chosen per row from the matrix that
    was built.
    """

    validate_config(
        dataclasses.replace(
            valid_config(), seq_len=seq_len, n_haplotypes=n_haplotypes
        )
    )


def test_the_flat_index_width_follows_the_matrix_it_addresses() -> None:
    limit = np.iinfo(np.uint32).max
    assert flat_index_dtype(0) == np.uint32
    assert flat_index_dtype(limit) == np.uint32
    assert flat_index_dtype(limit + 1) == np.uint64
    assert flat_index_dtype(10 * limit) == np.uint64


def test_stdpopsim_contig_positions_must_fit_uint32_storage() -> None:
    cfg = dataclasses.replace(valid_config(), stdpopsim_contig_length=UINT32_MAX + 1)
    with pytest.raises(ValueError, match="uint32 position"):
        validate_config(cfg)


def test_a_long_stdpopsim_contig_is_accepted() -> None:
    validate_config(
        dataclasses.replace(
            valid_config(),
            n_haplotypes=256,
            stdpopsim_contig_length=UINT32_MAX // 2 + 1,
        )
    )


def test_probability_fields_are_weights_and_need_not_sum_to_one() -> None:
    source_weights = {name: 0.0 for name in SOURCE_PROB_FIELDS}
    source_weights["p_custom_hetero"] = 4.0
    demography_weights = {name: 0.0 for name in DEMOGRAPHY_PROB_FIELDS}
    demography_weights["p_constant"] = 2.0
    validate_config(dataclasses.replace(valid_config(), **source_weights, **demography_weights))


@pytest.mark.parametrize(
    "source_mix",
    [
        "",
        ", ,",
        "smcpp_1kg:",
        ":1",
        "smcpp_1kg:not-a-number",
        "smcpp_1kg:-0.1",
        "smcpp_1kg:nan",
        "smcpp_1kg:0,phlash_unified:0",
        "smcpp_1kg:1,smcpp_1kg:0",
        "smcpp_1kg:1e308,phlash_unified:1e308",
        "smcpp_typo:1",
    ],
)
def test_invalid_empirical_ne_source_mix_is_rejected(source_mix: str) -> None:
    cfg = dataclasses.replace(valid_config(), empirical_ne_template_source_mix=source_mix)
    with pytest.raises(ValueError, match="empirical_ne_template_source_mix"):
        validate_config(cfg)


@pytest.mark.parametrize(
    "name",
    [
        "force",
        "compression",
        "include_truth",
        "enable_stdpopsim",
        "mut_map_scale_to_baseline",
    ],
)
def test_boolean_settings_reject_integer_substitutes(name: str) -> None:
    with pytest.raises(ValueError, match=name):
        validate_config(dataclasses.replace(valid_config(), **{name: 1}))

	from __future__ import annotations

	import dataclasses

	import numpy as np
	import pytest

	from simulator.config import (
	DEMOGRAPHY_PROB_FIELDS,
	FLOAT32_MIN_POSITIVE,
	INT16_MAX,
	SOURCE_PROB_FIELDS,
	UINT32_MAX,
	Config,
	apply_preset,
	validate_config,
	)
	from simulator.noise import flat_index_dtype


	def valid_config() -> Config:
	return apply_preset(Config(preset="mini"))


	@pytest.mark.parametrize("preset", ["mini", "core_v01", "full_v01"])
	def test_builtin_presets_are_valid(preset: str) -> None:
	validate_config(apply_preset(Config(preset=preset)))


	@pytest.mark.parametrize("preset", ["mini", "core_v01", "full_v01"])
	def test_builtin_presets_target_64_bins(preset: str) -> None:
	cfg = apply_preset(Config(preset=preset))

	assert cfg.time_bins == 64
	assert cfg.demography_resolution_log10 > 0.0


	@pytest.mark.parametrize(
	("changes", "message"),
	[
	({"n_samples": 0}, "n_samples"),
	({"n_samples": True}, "n_samples"),
	({"shard_size": 0}, "shard_size"),
	({"compute_chunk_size": 0}, "compute_chunk_size"),
	({"seq_len": 0}, "seq_len"),
	({"time_bins": 0}, "time_bins"),
	({"demography_resolution_log10": 0.0}, "demography_resolution_log10"),
	({"demography_resolution_log10": -1.0}, "demography_resolution_log10"),
	({"n_haplotypes": 1}, "at least 2"),
	({"n_haplotypes": 3}, "even"),
	({"n_haplotypes": INT16_MAX + 1}, "int16"),
	({"min_time": 0.0}, "min_time"),
	({"min_time": 500.0, "max_time": 500.0}, "max_time"),
	({"min_time": float("nan")}, "min_time"),
	({"baseline_mu": 0.0}, "baseline_mu"),
	({"baseline_rec": -1.0e-8}, "baseline_rec"),
	({"baseline_mu": FLOAT32_MIN_POSITIVE / 2.0}, "float32"),
	({"baseline_mu": 5.1e-7}, "synthetic maps"),
	({"baseline_rec": 1.1e-4}, "baseline_rec"),
	({"p_clean_constant": -0.01}, "p_clean_constant"),
	({"genotype_error_min": -0.01}, "genotype_error_min"),
	({"genotype_error_max": 1.01}, "genotype_error_max"),
	({"genotype_error_min": 0.2, "genotype_error_max": 0.1}, "must not exceed"),
	({"missing_rate_min": -0.01}, "missing_rate_min"),
	({"missing_rate_max": 1.01}, "missing_rate_max"),
	({"missing_rate_min": 0.2, "missing_rate_max": 0.1}, "must not exceed"),
	({"phase_switch_rate_per_het_min": -0.1}, "phase_switch_rate_per_het_min"),
	(
	{"phase_switch_rate_per_het_min": 0.2, "phase_switch_rate_per_het_max": 0.1},
	"must not exceed",
	),
	({"hotspot_missing_prob": -0.1}, "hotspot_missing_prob"),
	({"hotspot_missing_prob": 1.1}, "hotspot_missing_prob"),
	({"observed_map_noise": -0.1}, "observed_map_noise"),
	({"observed_map_noise": float("inf")}, "observed_map_noise"),
	({"empirical_ne_template_log10_shift_sd": -0.1}, "log10_shift"),
	({"empirical_ne_template_contrast_sd": -0.1}, "contrast"),
	({"empirical_ne_template_time_warp_sd": -0.1}, "time_warp"),
	({"empirical_ne_template_bin_noise_sd": -0.1}, "bin_noise"),
	({"map_segments_min": 0}, "map_segments_min"),
	({"map_segments_min": 10, "map_segments_max": 9}, "map_segments_min"),
	({"max_retries": 0}, "max_retries"),
	({"workers": 0}, "workers"),
	({"workers": -2}, "workers"),
	({"compression_level": -1}, "compression_level"),
	({"compression_level": 10}, "compression_level"),
	({"compression_level": True}, "compression_level"),
	({"seed": 1.5}, "seed"),
	({"dataset_name": ""}, "dataset_name"),
	({"preset": "unknown"}, "preset"),
	({"out_dir": ""}, "out_dir"),
	({"out_dir": "."}, "dedicated dataset directory"),
	({"demography_mixture_version": ""}, "demography_mixture_version"),
	({"empirical_ne_template_dir": ""}, "empirical_ne_template_dir"),
	({"stdpopsim_species": ""}, "stdpopsim_species"),
	({"stdpopsim_contig_length": -1}, "stdpopsim_contig_length"),
	({"recomb_map_unit": "centimorgans"}, "recomb_map_unit"),
	({"mut_map_unit": "unknown"}, "mut_map_unit"),
	],
	)
	def test_invalid_scalar_and_range_settings_are_rejected(changes: dict, message: str) -> None:
	with pytest.raises(ValueError, match=message):
	validate_config(dataclasses.replace(valid_config(), **changes))


	def test_source_mixture_requires_a_positive_weight() -> None:
	cfg = dataclasses.replace(valid_config(), **{name: 0.0 for name in SOURCE_PROB_FIELDS})
	with pytest.raises(ValueError, match="source probability"):
	validate_config(cfg)


	def test_disabled_stdpopsim_weight_is_not_an_enabled_source() -> None:
	source_weights = {name: 0.0 for name in SOURCE_PROB_FIELDS}
	source_weights["p_stdpopsim_anchor"] = 1.0
	cfg = dataclasses.replace(valid_config(), **source_weights, enable_stdpopsim=False)
	with pytest.raises(ValueError, match="enabled source probability"):
	validate_config(cfg)

	validate_config(dataclasses.replace(cfg, enable_stdpopsim=True))


	def test_demography_mixture_requires_a_positive_weight() -> None:
	cfg = dataclasses.replace(valid_config(), **{name: 0.0 for name in DEMOGRAPHY_PROB_FIELDS})
	with pytest.raises(ValueError, match="demography probability"):
	validate_config(cfg)


	def test_sequence_positions_must_fit_uint32_storage() -> None:
	with pytest.raises(ValueError, match="uint32 position"):
	validate_config(dataclasses.replace(valid_config(), seq_len=UINT32_MAX + 1))


	@pytest.mark.parametrize(
	("seq_len", "n_haplotypes"),
	[(60_000_000, 64), (80_000_000, 128), (80_000_000, 256), (40_000_000, 256)],
	)
	def test_a_long_sequence_with_many_haplotypes_is_accepted(
	seq_len: int, n_haplotypes: int
	) -> None:
	"""The sparse missing index addresses variants, not bases.

	Every configuration here was rejected by a precondition that multiplied
	``seq_len`` by ``n_haplotypes``, which is the size of the matrix only if
	every base is segregating. Variants are three orders of magnitude rarer,
	and the width that is actually used is chosen per row from the matrix that
	was built.
	"""

	validate_config(
	dataclasses.replace(
	valid_config(), seq_len=seq_len, n_haplotypes=n_haplotypes
	)
	)


	def test_the_flat_index_width_follows_the_matrix_it_addresses() -> None:
	limit = np.iinfo(np.uint32).max
	assert flat_index_dtype(0) == np.uint32
	assert flat_index_dtype(limit) == np.uint32
	assert flat_index_dtype(limit + 1) == np.uint64
	assert flat_index_dtype(10 * limit) == np.uint64


	def test_stdpopsim_contig_positions_must_fit_uint32_storage() -> None:
	cfg = dataclasses.replace(valid_config(), stdpopsim_contig_length=UINT32_MAX + 1)
	with pytest.raises(ValueError, match="uint32 position"):
	validate_config(cfg)


	def test_a_long_stdpopsim_contig_is_accepted() -> None:
	validate_config(
	dataclasses.replace(
	valid_config(),
	n_haplotypes=256,
	stdpopsim_contig_length=UINT32_MAX // 2 + 1,
	)
	)


	def test_probability_fields_are_weights_and_need_not_sum_to_one() -> None:
	source_weights = {name: 0.0 for name in SOURCE_PROB_FIELDS}
	source_weights["p_custom_hetero"] = 4.0
	demography_weights = {name: 0.0 for name in DEMOGRAPHY_PROB_FIELDS}
	demography_weights["p_constant"] = 2.0
	validate_config(dataclasses.replace(valid_config(), source_weights, demography_weights))


	@pytest.mark.parametrize(
	"source_mix",
	[
	"",
	", ,",
	"smcpp_1kg:",
	":1",
	"smcpp_1kg:not-a-number",
	"smcpp_1kg:-0.1",
	"smcpp_1kg:nan",
	"smcpp_1kg:0,phlash_unified:0",
	"smcpp_1kg:1,smcpp_1kg:0",
	"smcpp_1kg:1e308,phlash_unified:1e308",
	"smcpp_typo:1",
	],
	)
	def test_invalid_empirical_ne_source_mix_is_rejected(source_mix: str) -> None:
	cfg = dataclasses.replace(valid_config(), empirical_ne_template_source_mix=source_mix)
	with pytest.raises(ValueError, match="empirical_ne_template_source_mix"):
	validate_config(cfg)


	@pytest.mark.parametrize(
	"name",
	[
	"force",
	"compression",
	"include_truth",
	"enable_stdpopsim",
	"mut_map_scale_to_baseline",
	],
	)
	def test_boolean_settings_reject_integer_substitutes(name: str) -> None:
	with pytest.raises(ValueError, match=name):
	validate_config(dataclasses.replace(valid_config(), **{name: 1}))