feat(sector-radar): 建立独立指标与排名领域模型

This commit is contained in:
yuxuanhui
2026-08-29 16:52:45 +08:00
parent 5e4fe7220b
commit 3789008ea6
11 changed files with 1248 additions and 7 deletions
@@ -0,0 +1 @@
"""Independent post-close sector capital radar bounded context."""
@@ -0,0 +1 @@
"""Storage-independent sector radar models and calculation rules."""
@@ -0,0 +1,97 @@
"""Point-in-time stock fact aggregation for sector radar metrics."""
from __future__ import annotations
from collections.abc import Iterable
from decimal import Decimal
from .models import (
MembershipStatus,
SectorDailyAggregate,
SectorMembershipSnapshot,
StockDailyFact,
StockFactStatus,
)
def aggregate_sector_snapshot(
snapshot: SectorMembershipSnapshot,
stock_facts: Iterable[StockDailyFact],
) -> SectorDailyAggregate:
"""Aggregate only the members recorded in one dated membership snapshot.
Unknown membership returns an unavailable aggregate and deliberately
ignores any supplied stock facts. For known membership, suspended and
lifecycle-invalid members are excluded from the expected moneyflow
denominator; missing facts remain expected and reduce coverage.
Args:
snapshot: Dated sector identity and point-in-time member codes.
stock_facts: Normalized facts that may contain records outside the sector.
Returns:
A yuan-denominated aggregate with explicit membership and moneyflow coverage.
Raises:
ValueError: If member facts have a date mismatch or duplicate stock code.
"""
if snapshot.status is MembershipStatus.UNKNOWN:
return SectorDailyAggregate(
trade_date=snapshot.trade_date,
sector_type=snapshot.sector_type,
sector_code=snapshot.sector_code,
sector_name=snapshot.sector_name,
member_count=0,
valid_sample_count=0,
net_amount_yuan=None,
turnover_yuan=None,
membership_coverage=Decimal(0),
moneyflow_coverage=Decimal(0),
)
members = set(snapshot.member_codes)
facts_by_code: dict[str, StockDailyFact] = {}
for fact in stock_facts:
if fact.ts_code not in members:
continue
if fact.trade_date != snapshot.trade_date:
raise ValueError("member stock facts must match the snapshot trade_date")
if fact.ts_code in facts_by_code:
raise ValueError("member stock facts must have unique ts_code values")
facts_by_code[fact.ts_code] = fact
net_amount_total = Decimal(0)
turnover_total = Decimal(0)
valid_count = 0
expected_count = 0
for member_code in snapshot.member_codes:
fact = facts_by_code.get(member_code)
if fact is None or fact.status is StockFactStatus.MISSING:
expected_count += 1
elif fact.status is StockFactStatus.AVAILABLE:
expected_count += 1
net_amount = fact.net_amount_yuan
turnover = fact.turnover_yuan
if net_amount is None or turnover is None:
raise ValueError("available stock facts require both amounts")
net_amount_total += net_amount
turnover_total += turnover
valid_count += 1
moneyflow_coverage = (
Decimal(valid_count) / Decimal(expected_count) if expected_count else Decimal(1)
)
return SectorDailyAggregate(
trade_date=snapshot.trade_date,
sector_type=snapshot.sector_type,
sector_code=snapshot.sector_code,
sector_name=snapshot.sector_name,
member_count=len(snapshot.member_codes),
valid_sample_count=valid_count,
net_amount_yuan=net_amount_total if valid_count else None,
turnover_yuan=turnover_total if valid_count else None,
membership_coverage=Decimal(1),
moneyflow_coverage=moneyflow_coverage,
)
@@ -0,0 +1,204 @@
"""Transparent, versioned metric strategies for the independent radar."""
from __future__ import annotations
from collections.abc import Iterable
from datetime import date
from decimal import Decimal
from typing import Protocol
from .models import (
MetricKind,
MetricObservation,
MetricQuality,
MetricUnit,
SectorDailyAggregate,
)
class MetricStrategy(Protocol):
"""Calculate one named metric from a sector's point-in-time daily history."""
metric_kind: MetricKind
metric_version: str
unit: MetricUnit
def evaluate(
self,
history: Iterable[SectorDailyAggregate],
target_trade_date: date,
) -> MetricObservation:
"""Return the target date observation without inventing missing inputs."""
...
def _target_aggregate(
history: Iterable[SectorDailyAggregate], target_trade_date: date
) -> SectorDailyAggregate:
matches = tuple(row for row in history if row.trade_date == target_trade_date)
if len(matches) != 1:
raise ValueError("history must contain exactly one target-date aggregate")
return matches[0]
def _quality(row: SectorDailyAggregate) -> MetricQuality:
if row.valid_sample_count < 5 or row.membership_coverage < 1 or row.moneyflow_coverage < 1:
return MetricQuality.AVAILABLE_LIMITED_SAMPLE
return MetricQuality.AVAILABLE
def _observation(
row: SectorDailyAggregate,
*,
metric_kind: MetricKind,
metric_version: str,
unit: MetricUnit,
value: Decimal | None,
quality: MetricQuality | None = None,
) -> MetricObservation:
return MetricObservation(
trade_date=row.trade_date,
sector_type=row.sector_type,
sector_code=row.sector_code,
sector_name=row.sector_name,
metric_kind=metric_kind,
metric_version=metric_version,
implementation_kind="independent",
unit=unit,
value=value,
quality=(
MetricQuality.UNAVAILABLE
if value is None
else quality
if quality is not None
else _quality(row)
),
member_count=row.member_count,
valid_sample_count=row.valid_sample_count,
membership_coverage=row.membership_coverage,
moneyflow_coverage=row.moneyflow_coverage,
)
class AmountNetStrategy:
"""Aggregate main net amount and expose it in hundred-million yuan."""
metric_kind = MetricKind.AMOUNT
metric_version = "zhixing_amount_net_bn_v1"
unit = MetricUnit.CNY_100M
def evaluate(
self,
history: Iterable[SectorDailyAggregate],
target_trade_date: date,
) -> MetricObservation:
"""Return the target net amount; missing moneyflow remains unavailable."""
row = _target_aggregate(history, target_trade_date)
value = None if row.net_amount_yuan is None else row.net_amount_yuan / Decimal("100000000")
return _observation(
row,
metric_kind=self.metric_kind,
metric_version=self.metric_version,
unit=self.unit,
value=value,
)
class RatioTurnoverStrategy:
"""Divide aggregated main net amount by aggregated daily turnover."""
metric_kind = MetricKind.RATIO
metric_version = "zhixing_ratio_turnover_v1"
unit = MetricUnit.RATIO
def evaluate(
self,
history: Iterable[SectorDailyAggregate],
target_trade_date: date,
) -> MetricObservation:
"""Return a ratio only when numerator and positive denominator exist."""
row = _target_aggregate(history, target_trade_date)
value = None
if (
row.net_amount_yuan is not None
and row.turnover_yuan is not None
and row.turnover_yuan > 0
):
value = row.net_amount_yuan / row.turnover_yuan
return _observation(
row,
metric_kind=self.metric_kind,
metric_version=self.metric_version,
unit=self.unit,
value=value,
)
class SwingEqualThreeToTenStrategy:
"""Average transparent 3-to-10-day aggregate turnover ratios equally.
This strategy is deliberately named as a Zhixing implementation. It does
not reproduce or imply OneChartLab's unpublished window weights or score.
"""
metric_kind = MetricKind.SWING
metric_version = "zhixing_swing_equal_3_10_v1"
unit = MetricUnit.RATIO
def evaluate(
self,
history: Iterable[SectorDailyAggregate],
target_trade_date: date,
) -> MetricObservation:
"""Calculate eight complete trading-day windows ending at the target."""
rows = tuple(sorted(history, key=lambda row: row.trade_date))
target = _target_aggregate(rows, target_trade_date)
eligible = tuple(row for row in rows if row.trade_date <= target_trade_date)
if any(
(row.sector_type, row.sector_code) != (target.sector_type, target.sector_code)
for row in eligible
):
raise ValueError("history must contain exactly one sector identity")
if len({row.trade_date for row in eligible}) != len(eligible):
raise ValueError("history must not contain duplicate trade dates")
value: Decimal | None = None
quality: MetricQuality | None = None
if len(eligible) >= 10:
latest = eligible[-10:]
window_ratios: list[Decimal] = []
for window_size in range(3, 11):
window = latest[-window_size:]
net_amount = Decimal(0)
turnover = Decimal(0)
for row in window:
if row.net_amount_yuan is None or row.turnover_yuan is None:
break
net_amount += row.net_amount_yuan
turnover += row.turnover_yuan
else:
if turnover <= 0:
break
window_ratios.append(net_amount / turnover)
continue
break
if len(window_ratios) == 8:
value = sum(window_ratios, start=Decimal(0)) / Decimal(8)
quality = (
MetricQuality.AVAILABLE_LIMITED_SAMPLE
if any(_quality(row) is not MetricQuality.AVAILABLE for row in latest)
else MetricQuality.AVAILABLE
)
return _observation(
target,
metric_kind=self.metric_kind,
metric_version=self.metric_version,
unit=self.unit,
value=value,
quality=quality,
)
@@ -0,0 +1,315 @@
"""Stable domain values for independently produced sector radar metrics."""
from __future__ import annotations
from dataclasses import dataclass
from datetime import date, datetime
from decimal import Decimal
from enum import StrEnum
from typing import Literal
class SectorType(StrEnum):
"""Independent ranking pools supported by the first radar release."""
CONCEPT = "concept"
INDUSTRY = "industry"
class MembershipStatus(StrEnum):
"""Availability of a point-in-time sector membership snapshot."""
AVAILABLE = "available"
UNKNOWN = "membership_unknown"
class StockFactStatus(StrEnum):
"""Why one member does or does not contribute to a daily aggregate."""
AVAILABLE = "available"
SUSPENDED = "suspended"
MISSING = "missing"
LIFECYCLE_INVALID = "lifecycle_invalid"
LOW_LIQUIDITY = "low_liquidity"
class PublicationStatus(StrEnum):
"""Immutable build states retained for audit and last-good selection."""
RUNNING = "running"
SUCCESS = "success"
PARTIAL = "partial"
FAILED = "failed"
class MetricKind(StrEnum):
"""User-facing metric families without borrowing private score names."""
AMOUNT = "amount"
RATIO = "ratio"
SWING = "swing"
class MetricQuality(StrEnum):
"""Whether a metric is usable and whether its sample needs a warning."""
AVAILABLE = "available"
AVAILABLE_LIMITED_SAMPLE = "available_limited_sample"
UNAVAILABLE = "unavailable"
class MetricUnit(StrEnum):
"""Units exposed by independent metric strategies."""
CNY_100M = "CNY_100M"
RATIO = "ratio"
class RankSide(StrEnum):
"""Ordinary percentile views exposed by the ranking read model."""
TOP = "top"
BOTTOM = "bottom"
ALL = "all"
def _validate_finite_decimal(value: Decimal | None, field_name: str) -> None:
"""Reject non-finite domain values while preserving missing values."""
if value is not None and not value.is_finite():
raise ValueError(f"{field_name} must be finite or None")
def _validate_coverage(value: Decimal, field_name: str) -> None:
"""Require a finite fraction in the inclusive zero-to-one range."""
_validate_finite_decimal(value, field_name)
if value < 0 or value > 1:
raise ValueError(f"{field_name} must be between 0 and 1")
@dataclass(frozen=True, slots=True)
class SectorMembershipSnapshot:
"""One sector's membership as observed for exactly one trade date.
``UNKNOWN`` is an explicit fact: callers must not substitute a current
member list when the historical snapshot is unavailable.
"""
trade_date: date
sector_type: SectorType
sector_code: str
sector_name: str
member_codes: tuple[str, ...]
status: MembershipStatus
source_version: str
def __post_init__(self) -> None:
"""Validate identity, deterministic membership, and unknown semantics."""
if not self.sector_code.strip():
raise ValueError("sector_code must not be empty")
if not self.sector_name.strip():
raise ValueError("sector_name must not be empty")
if not self.source_version.strip():
raise ValueError("source_version must not be empty")
if any(not code.strip() for code in self.member_codes):
raise ValueError("member_codes must not contain empty values")
if len(self.member_codes) != len(set(self.member_codes)):
raise ValueError("member_codes must be unique")
if self.status is MembershipStatus.UNKNOWN and self.member_codes:
raise ValueError("unknown membership must not expose member_codes")
@dataclass(frozen=True, slots=True)
class StockDailyFact:
"""Normalized daily turnover and moneyflow for one member.
Amounts are expressed in yuan. Available facts require both source
values, including an observed zero. Non-available statuses cannot carry
amounts because doing so would blur missing, suspended, and lifecycle
semantics at the metric boundary.
"""
trade_date: date
ts_code: str
status: StockFactStatus
turnover_yuan: Decimal | None = None
net_amount_yuan: Decimal | None = None
def __post_init__(self) -> None:
"""Reject incomplete available facts and hidden non-finite values."""
if not self.ts_code.strip():
raise ValueError("ts_code must not be empty")
_validate_finite_decimal(self.turnover_yuan, "turnover_yuan")
_validate_finite_decimal(self.net_amount_yuan, "net_amount_yuan")
if self.status is StockFactStatus.AVAILABLE:
if self.turnover_yuan is None or self.net_amount_yuan is None:
raise ValueError("available stock facts require both amounts")
if self.turnover_yuan < 0:
raise ValueError("turnover_yuan must not be negative")
elif self.turnover_yuan is not None or self.net_amount_yuan is not None:
raise ValueError("non-available stock facts must not expose amounts")
@dataclass(frozen=True, slots=True)
class RadarPublication:
"""Traceable identity and lifecycle of one immutable radar build revision."""
publication_id: str
target_trade_date: date
status: PublicationStatus
source_version: str
universe_version: str
metric_versions: tuple[str, ...]
input_hash: str | None
coverage: Decimal
started_at: datetime
finished_at: datetime | None = None
error_summary: str | None = None
def __post_init__(self) -> None:
"""Keep running and terminal lifecycle timestamps internally consistent."""
if not self.publication_id.strip():
raise ValueError("publication_id must not be empty")
if not self.source_version.strip() or not self.universe_version.strip():
raise ValueError("publication source versions must not be empty")
if not self.metric_versions or any(not value.strip() for value in self.metric_versions):
raise ValueError("metric_versions must contain named strategies")
if len(self.metric_versions) != len(set(self.metric_versions)):
raise ValueError("metric_versions must be unique")
_validate_coverage(self.coverage, "coverage")
if self.started_at.tzinfo is None:
raise ValueError("started_at must be timezone-aware")
is_running = self.status is PublicationStatus.RUNNING
if is_running != (self.finished_at is None):
raise ValueError("finished_at must be absent only while publication is running")
if self.finished_at is not None:
if self.finished_at.tzinfo is None:
raise ValueError("finished_at must be timezone-aware")
if self.finished_at < self.started_at:
raise ValueError("finished_at must not precede started_at")
if self.status is PublicationStatus.SUCCESS and self.input_hash is None:
raise ValueError("successful publication requires input_hash")
if self.input_hash is not None and (
len(self.input_hash) != 64
or any(character not in "0123456789abcdef" for character in self.input_hash)
):
raise ValueError("input_hash must be a lowercase SHA-256 hex digest")
@dataclass(frozen=True, slots=True)
class SectorDailyAggregate:
"""One sector's point-in-time daily facts after source normalization.
Amounts use yuan so strategies cannot accidentally mix Tushare's
``moneyflow_dc.net_amount`` (ten-thousand yuan) with ``daily.amount``
(thousand yuan). ``None`` means missing source data; zero remains an
observed value.
"""
trade_date: date
sector_type: SectorType
sector_code: str
sector_name: str
member_count: int
valid_sample_count: int
net_amount_yuan: Decimal | None
turnover_yuan: Decimal | None
membership_coverage: Decimal
moneyflow_coverage: Decimal
def __post_init__(self) -> None:
"""Validate counts, coverage, and finite normalized values."""
if not self.sector_code.strip():
raise ValueError("sector_code must not be empty")
if not self.sector_name.strip():
raise ValueError("sector_name must not be empty")
if self.member_count < 0:
raise ValueError("member_count must not be negative")
if not 0 <= self.valid_sample_count <= self.member_count:
raise ValueError("valid_sample_count must be within member_count")
_validate_finite_decimal(self.net_amount_yuan, "net_amount_yuan")
_validate_finite_decimal(self.turnover_yuan, "turnover_yuan")
_validate_coverage(self.membership_coverage, "membership_coverage")
_validate_coverage(self.moneyflow_coverage, "moneyflow_coverage")
@dataclass(frozen=True, slots=True)
class MetricObservation:
"""One versioned independent metric value ready for cross-sectional ranking."""
trade_date: date
sector_type: SectorType
sector_code: str
sector_name: str
metric_kind: MetricKind
metric_version: str
implementation_kind: Literal["independent"]
unit: MetricUnit
value: Decimal | None
quality: MetricQuality
member_count: int
valid_sample_count: int
membership_coverage: Decimal
moneyflow_coverage: Decimal
def __post_init__(self) -> None:
"""Keep unavailable and finite-value states internally consistent."""
_validate_finite_decimal(self.value, "value")
if self.value is None and self.quality is not MetricQuality.UNAVAILABLE:
raise ValueError("a missing metric value must be unavailable")
if self.value is not None and self.quality is MetricQuality.UNAVAILABLE:
raise ValueError("an unavailable metric must not expose a value")
@dataclass(frozen=True, slots=True)
class RankChange:
"""One previous-publication rank delta using past minus current rank."""
days: int
value: int | None
def __post_init__(self) -> None:
"""Limit the public comparison window to one through five days."""
if not 1 <= self.days <= 5:
raise ValueError("rank change days must be between 1 and 5")
@dataclass(frozen=True, slots=True)
class RankedMetric:
"""A metric observation with its position inside one independent pool."""
observation: MetricObservation
rank_position: int | None
rank_percentile: Decimal | None
rank_changes: tuple[RankChange, ...] = ()
def __post_init__(self) -> None:
"""Require rank position and percentile to be present or absent together."""
if (self.rank_position is None) != (self.rank_percentile is None):
raise ValueError("rank_position and rank_percentile must be paired")
if self.rank_position is not None and self.rank_position < 1:
raise ValueError("rank_position must be positive")
_validate_finite_decimal(self.rank_percentile, "rank_percentile")
if self.rank_percentile is not None and not 0 < self.rank_percentile <= 100:
raise ValueError("rank_percentile must be within (0, 100]")
days = [change.days for change in self.rank_changes]
if len(days) != len(set(days)):
raise ValueError("rank change days must be unique")
def rank_change(self, days: int) -> int | None:
"""Return one configured rank delta, or ``None`` when history is absent."""
if not 1 <= days <= 5:
raise ValueError("rank change days must be between 1 and 5")
return next(
(change.value for change in self.rank_changes if change.days == days),
None,
)
@@ -0,0 +1,216 @@
"""Deterministic cross-sectional ranking for independent sector pools."""
from __future__ import annotations
from collections import defaultdict
from collections.abc import Iterable, Mapping
from dataclasses import replace
from datetime import date
from decimal import Decimal
from .models import (
MetricKind,
MetricObservation,
RankChange,
RankedMetric,
RankSide,
SectorType,
)
PoolKey = tuple[date, SectorType, MetricKind, str]
SectorMetricKey = tuple[SectorType, str, MetricKind, str]
def _pool_key(observation: MetricObservation) -> PoolKey:
return (
observation.trade_date,
observation.sector_type,
observation.metric_kind,
observation.metric_version,
)
def _sector_metric_key(observation: MetricObservation) -> SectorMetricKey:
return (
observation.sector_type,
observation.sector_code,
observation.metric_kind,
observation.metric_version,
)
def _available_sort_key(observation: MetricObservation) -> tuple[Decimal, str]:
if observation.value is None:
raise ValueError("unavailable observations cannot use the ranking sort key")
return (-observation.value, observation.sector_code)
def rank_metric_observations(
observations: Iterable[MetricObservation],
) -> tuple[RankedMetric, ...]:
"""Rank observations by value within date, type, metric, and version.
Concept and industry observations never share a pool. Equal metric values
use ascending sector code as the documented Zhixing tie-breaker. Missing
values remain visible but do not consume a rank.
"""
pools: defaultdict[PoolKey, list[MetricObservation]] = defaultdict(list)
for observation in observations:
pools[_pool_key(observation)].append(observation)
result: list[RankedMetric] = []
for pool_key in sorted(
pools,
key=lambda key: (key[0], key[1].value, key[2].value, key[3]),
):
pool = pools[pool_key]
codes = [observation.sector_code for observation in pool]
if len(codes) != len(set(codes)):
raise ValueError("a ranking pool must not contain duplicate sector codes")
available = sorted(
(observation for observation in pool if observation.value is not None),
key=_available_sort_key,
)
pool_size = len(available)
for rank_position, observation in enumerate(available, start=1):
rank_percentile = (
Decimal(100) * Decimal(pool_size - rank_position + 1) / Decimal(pool_size)
)
result.append(
RankedMetric(
observation=observation,
rank_position=rank_position,
rank_percentile=rank_percentile,
)
)
result.extend(
RankedMetric(
observation=observation,
rank_position=None,
rank_percentile=None,
)
for observation in sorted(
(observation for observation in pool if observation.value is None),
key=lambda observation: observation.sector_code,
)
)
return tuple(result)
def select_percentile_side(
rankings: Iterable[RankedMetric], side: RankSide
) -> tuple[RankedMetric, ...]:
"""Select confirmed inclusive percentile sides without fixed row counts."""
rows = tuple(rankings)
if side is RankSide.ALL:
return rows
threshold_rows = tuple(
row
for row in rows
if row.rank_percentile is not None
and (
row.rank_percentile >= Decimal(90)
if side is RankSide.TOP
else row.rank_percentile <= Decimal(10)
)
)
if side is RankSide.TOP:
return threshold_rows
pools: defaultdict[PoolKey, list[RankedMetric]] = defaultdict(list)
for row in threshold_rows:
pools[_pool_key(row.observation)].append(row)
result: list[RankedMetric] = []
for pool_key in sorted(
pools,
key=lambda key: (key[0], key[1].value, key[2].value, key[3]),
):
result.extend(
sorted(
pools[pool_key],
key=lambda row: (
row.observation.value if row.observation.value is not None else Decimal(0),
row.observation.sector_code,
),
)
)
return tuple(result)
def with_rank_changes(
current_rankings: Iterable[RankedMetric],
history_by_days: Mapping[int, Iterable[RankedMetric]],
) -> tuple[RankedMetric, ...]:
"""Attach 1-to-5-day deltas without turning missing history into zero."""
history_indexes: dict[int, dict[SectorMetricKey, int | None]] = {}
for days, historical_rankings in history_by_days.items():
if not 1 <= days <= 5:
raise ValueError("rank change days must be between 1 and 5")
index: dict[SectorMetricKey, int | None] = {}
for row in historical_rankings:
key = _sector_metric_key(row.observation)
if key in index:
raise ValueError("historical rankings must have unique sector metrics")
index[key] = row.rank_position
history_indexes[days] = index
result: list[RankedMetric] = []
for row in current_rankings:
key = _sector_metric_key(row.observation)
changes: list[RankChange] = []
for days in sorted(history_indexes):
past_rank = history_indexes[days].get(key)
value = (
past_rank - row.rank_position
if past_rank is not None and row.rank_position is not None
else None
)
changes.append(RankChange(days=days, value=value))
result.append(replace(row, rank_changes=tuple(changes)))
return tuple(result)
def select_rank_change_side(
rankings: Iterable[RankedMetric],
*,
days: int,
side: RankSide,
) -> tuple[RankedMetric, ...]:
"""Select the strongest or weakest ceiling-ten-percent rank changes per pool."""
if not 1 <= days <= 5:
raise ValueError("rank change days must be between 1 and 5")
pools: defaultdict[PoolKey, list[RankedMetric]] = defaultdict(list)
for row in rankings:
pools[_pool_key(row.observation)].append(row)
result: list[RankedMetric] = []
for pool_key in sorted(
pools,
key=lambda key: (key[0], key[1].value, key[2].value, key[3]),
):
pool = pools[pool_key]
pool_size = sum(row.rank_position is not None for row in pool)
take_count = max(1, (pool_size + 9) // 10) if pool_size else 0
candidates = tuple(
(change, row) for row in pool if (change := row.rank_change(days)) is not None
)
if side is RankSide.BOTTOM:
ordered = sorted(
candidates,
key=lambda item: (item[0], item[1].observation.sector_code),
)
else:
ordered = sorted(
candidates,
key=lambda item: (-item[0], item[1].observation.sector_code),
)
selected = ordered if side is RankSide.ALL else ordered[:take_count]
result.extend(row for _, row in selected)
return tuple(result)
@@ -0,0 +1,146 @@
from datetime import UTC, date, datetime
from decimal import Decimal
import pytest
from zhixing_server.modules.sector_radar.domain.facts import aggregate_sector_snapshot
from zhixing_server.modules.sector_radar.domain.models import (
MembershipStatus,
PublicationStatus,
RadarPublication,
SectorMembershipSnapshot,
SectorType,
StockDailyFact,
StockFactStatus,
)
TARGET_DATE = date(2026, 8, 28)
def test_point_in_time_aggregation_distinguishes_suspension_missing_and_zero() -> None:
snapshot = SectorMembershipSnapshot(
trade_date=TARGET_DATE,
sector_type=SectorType.CONCEPT,
sector_code="BK0001.DC",
sector_name="示例概念",
member_codes=("000001.SZ", "000002.SZ", "000003.SZ", "000004.SZ"),
status=MembershipStatus.AVAILABLE,
source_version="dc-member-20260828-a",
)
facts = (
StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000001.SZ",
status=StockFactStatus.AVAILABLE,
turnover_yuan=Decimal("1000"),
net_amount_yuan=Decimal("100"),
),
StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000002.SZ",
status=StockFactStatus.AVAILABLE,
turnover_yuan=Decimal("2000"),
net_amount_yuan=Decimal("0"),
),
StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000003.SZ",
status=StockFactStatus.SUSPENDED,
),
StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000004.SZ",
status=StockFactStatus.MISSING,
),
)
aggregate = aggregate_sector_snapshot(snapshot, facts)
assert aggregate.member_count == 4
assert aggregate.valid_sample_count == 2
assert aggregate.net_amount_yuan == Decimal("100")
assert aggregate.turnover_yuan == Decimal("3000")
assert aggregate.membership_coverage == Decimal("1")
assert aggregate.moneyflow_coverage == Decimal("2") / Decimal("3")
def test_unknown_membership_never_falls_back_to_available_stock_facts() -> None:
snapshot = SectorMembershipSnapshot(
trade_date=TARGET_DATE,
sector_type=SectorType.INDUSTRY,
sector_code="BK1001.DC",
sector_name="示例行业",
member_codes=(),
status=MembershipStatus.UNKNOWN,
source_version="dc-member-missing",
)
fact = StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000001.SZ",
status=StockFactStatus.AVAILABLE,
turnover_yuan=Decimal("1000"),
net_amount_yuan=Decimal("100"),
)
aggregate = aggregate_sector_snapshot(snapshot, (fact,))
assert aggregate.member_count == 0
assert aggregate.net_amount_yuan is None
assert aggregate.turnover_yuan is None
assert aggregate.membership_coverage == Decimal("0")
def test_stock_fact_rejects_non_finite_values_and_invalid_status_payloads() -> None:
with pytest.raises(ValueError, match="finite"):
StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000001.SZ",
status=StockFactStatus.AVAILABLE,
turnover_yuan=Decimal("Infinity"),
net_amount_yuan=Decimal("1"),
)
with pytest.raises(ValueError, match="must not expose amounts"):
StockDailyFact(
trade_date=TARGET_DATE,
ts_code="000001.SZ",
status=StockFactStatus.SUSPENDED,
turnover_yuan=Decimal("0"),
)
def test_publication_requires_terminal_completion_and_replay_identity() -> None:
started_at = datetime(2026, 8, 28, 17, 30, tzinfo=UTC)
publication = RadarPublication(
publication_id="radar-20260828-a",
target_trade_date=TARGET_DATE,
status=PublicationStatus.SUCCESS,
source_version="tushare-pro-v1",
universe_version="eastmoney-dc-20260828-a",
metric_versions=(
"zhixing_amount_net_bn_v1",
"zhixing_ratio_turnover_v1",
"zhixing_swing_equal_3_10_v1",
),
input_hash="a" * 64,
coverage=Decimal("0.995"),
started_at=started_at,
finished_at=datetime(2026, 8, 28, 17, 35, tzinfo=UTC),
)
assert publication.status is PublicationStatus.SUCCESS
with pytest.raises(ValueError, match="finished_at"):
RadarPublication(
publication_id="radar-20260828-running",
target_trade_date=TARGET_DATE,
status=PublicationStatus.RUNNING,
source_version="tushare-pro-v1",
universe_version="eastmoney-dc-20260828-a",
metric_versions=("zhixing_amount_net_bn_v1",),
input_hash=None,
coverage=Decimal("0"),
started_at=started_at,
finished_at=started_at,
)
@@ -0,0 +1,112 @@
from datetime import date
from decimal import Decimal
from zhixing_server.modules.sector_radar.domain.metrics import (
AmountNetStrategy,
RatioTurnoverStrategy,
SwingEqualThreeToTenStrategy,
)
from zhixing_server.modules.sector_radar.domain.models import (
MetricQuality,
SectorDailyAggregate,
SectorType,
)
TARGET_DATE = date(2026, 8, 28)
def make_aggregate(
*,
net_amount_yuan: Decimal | None = Decimal("125000000"),
turnover_yuan: Decimal | None = Decimal("5000000000"),
) -> SectorDailyAggregate:
return SectorDailyAggregate(
trade_date=TARGET_DATE,
sector_type=SectorType.CONCEPT,
sector_code="BK0001.DC",
sector_name="示例概念",
member_count=10,
valid_sample_count=10,
net_amount_yuan=net_amount_yuan,
turnover_yuan=turnover_yuan,
membership_coverage=Decimal("1"),
moneyflow_coverage=Decimal("1"),
)
def test_amount_and_ratio_strategies_expose_independent_versioned_values() -> None:
aggregate = make_aggregate()
amount = AmountNetStrategy().evaluate((aggregate,), TARGET_DATE)
ratio = RatioTurnoverStrategy().evaluate((aggregate,), TARGET_DATE)
assert amount.value == Decimal("1.25")
assert amount.metric_version == "zhixing_amount_net_bn_v1"
assert amount.implementation_kind == "independent"
assert amount.unit == "CNY_100M"
assert amount.quality is MetricQuality.AVAILABLE
assert ratio.value == Decimal("0.025")
assert ratio.metric_version == "zhixing_ratio_turnover_v1"
assert ratio.implementation_kind == "independent"
assert ratio.unit == "ratio"
def test_missing_moneyflow_is_unavailable_but_zero_remains_a_real_value() -> None:
missing = AmountNetStrategy().evaluate((make_aggregate(net_amount_yuan=None),), TARGET_DATE)
zero = AmountNetStrategy().evaluate(
(make_aggregate(net_amount_yuan=Decimal("0")),), TARGET_DATE
)
assert missing.value is None
assert missing.quality is MetricQuality.UNAVAILABLE
assert zero.value == Decimal("0")
assert zero.quality is MetricQuality.AVAILABLE
def test_swing_strategy_uses_each_days_point_in_time_aggregate() -> None:
history = tuple(
SectorDailyAggregate(
trade_date=date(2026, 8, 18 + offset),
sector_type=SectorType.CONCEPT,
sector_code="BK0001.DC",
sector_name="示例概念",
member_count=6 + offset,
valid_sample_count=6 + offset,
net_amount_yuan=Decimal(str(offset + 1)),
turnover_yuan=Decimal("100"),
membership_coverage=Decimal("1"),
moneyflow_coverage=Decimal("1"),
)
for offset in range(10)
)
result = SwingEqualThreeToTenStrategy().evaluate(history, date(2026, 8, 27))
# The worked 3..10-day window ratios average to exactly 0.0725.
assert result.value == Decimal("0.0725")
assert result.metric_version == "zhixing_swing_equal_3_10_v1"
assert result.member_count == 15
def test_swing_strategy_carries_forward_limited_historical_sample_quality() -> None:
history = tuple(
SectorDailyAggregate(
trade_date=date(2026, 8, 18 + offset),
sector_type=SectorType.INDUSTRY,
sector_code="BK1001.DC",
sector_name="示例行业",
member_count=10,
valid_sample_count=4 if offset == 0 else 10,
net_amount_yuan=Decimal("10"),
turnover_yuan=Decimal("100"),
membership_coverage=Decimal("1"),
moneyflow_coverage=Decimal("1"),
)
for offset in range(10)
)
result = SwingEqualThreeToTenStrategy().evaluate(history, date(2026, 8, 27))
assert result.value == Decimal("0.1")
assert result.quality is MetricQuality.AVAILABLE_LIMITED_SAMPLE
@@ -0,0 +1,147 @@
from datetime import date
from decimal import Decimal
from zhixing_server.modules.sector_radar.domain.models import (
MetricKind,
MetricObservation,
MetricQuality,
MetricUnit,
RankSide,
SectorType,
)
from zhixing_server.modules.sector_radar.domain.ranking import (
rank_metric_observations,
select_percentile_side,
select_rank_change_side,
with_rank_changes,
)
TARGET_DATE = date(2026, 8, 28)
def make_observation(
sector_code: str,
sector_type: SectorType,
value: str | None,
*,
trade_date: date = TARGET_DATE,
) -> MetricObservation:
metric_value = Decimal(value) if value is not None else None
return MetricObservation(
trade_date=trade_date,
sector_type=sector_type,
sector_code=sector_code,
sector_name=sector_code,
metric_kind=MetricKind.AMOUNT,
metric_version="zhixing_amount_net_bn_v1",
implementation_kind="independent",
unit=MetricUnit.CNY_100M,
value=metric_value,
quality=(
MetricQuality.AVAILABLE if metric_value is not None else MetricQuality.UNAVAILABLE
),
member_count=10,
valid_sample_count=10 if metric_value is not None else 0,
membership_coverage=Decimal("1"),
moneyflow_coverage=Decimal("1"),
)
def test_ranking_separates_types_and_uses_code_as_stable_tie_breaker() -> None:
observations = (
make_observation("BK2002.DC", SectorType.INDUSTRY, "20"),
make_observation("BK1002.DC", SectorType.CONCEPT, "30"),
make_observation("BK2001.DC", SectorType.INDUSTRY, "20"),
make_observation("BK1001.DC", SectorType.CONCEPT, "10"),
)
ranked = rank_metric_observations(tuple(reversed(observations)))
by_code = {row.observation.sector_code: row for row in ranked}
assert by_code["BK1002.DC"].rank_position == 1
assert by_code["BK1002.DC"].rank_percentile == Decimal("100")
assert by_code["BK1001.DC"].rank_position == 2
assert by_code["BK1001.DC"].rank_percentile == Decimal("50")
assert by_code["BK2001.DC"].rank_position == 1
assert by_code["BK2002.DC"].rank_position == 2
def test_ranking_handles_empty_and_single_element_pools() -> None:
assert rank_metric_observations(()) == ()
[single] = rank_metric_observations((make_observation("BK0001.DC", SectorType.CONCEPT, "0"),))
assert single.rank_position == 1
assert single.rank_percentile == Decimal("100")
def test_percentile_sides_use_confirmed_inclusive_thresholds() -> None:
ranked = rank_metric_observations(
make_observation(f"BK{position:04d}.DC", SectorType.CONCEPT, str(11 - position))
for position in range(1, 11)
)
top = select_percentile_side(ranked, RankSide.TOP)
bottom = select_percentile_side(ranked, RankSide.BOTTOM)
assert [row.observation.sector_code for row in top] == ["BK0001.DC", "BK0002.DC"]
assert [row.observation.sector_code for row in bottom] == ["BK0010.DC"]
def test_rank_change_is_past_rank_minus_current_and_preserves_missing_history() -> None:
current = rank_metric_observations(
(
make_observation("BK0001.DC", SectorType.CONCEPT, "30"),
make_observation("BK0002.DC", SectorType.CONCEPT, "20"),
)
)
previous = rank_metric_observations(
(
make_observation(
"BK0001.DC",
SectorType.CONCEPT,
"10",
trade_date=date(2026, 8, 27),
),
make_observation(
"BK0002.DC",
SectorType.CONCEPT,
"40",
trade_date=date(2026, 8, 27),
),
)
)
changed = with_rank_changes(current, {1: previous, 5: ()})
by_code = {row.observation.sector_code: row for row in changed}
assert by_code["BK0001.DC"].rank_change(1) == 1
assert by_code["BK0002.DC"].rank_change(1) == -1
assert by_code["BK0001.DC"].rank_change(5) is None
def test_rank_change_sides_take_ceiling_ten_percent_per_pool() -> None:
current = rank_metric_observations(
make_observation(f"BK{position:04d}.DC", SectorType.CONCEPT, str(12 - position))
for position in range(1, 12)
)
previous = rank_metric_observations(
make_observation(
f"BK{position:04d}.DC",
SectorType.CONCEPT,
str(position),
trade_date=date(2026, 8, 27),
)
for position in range(1, 12)
)
changed = with_rank_changes(current, {1: previous})
top = select_rank_change_side(changed, days=1, side=RankSide.TOP)
bottom = select_rank_change_side(changed, days=1, side=RankSide.BOTTOM)
assert [row.observation.sector_code for row in top] == ["BK0001.DC", "BK0002.DC"]
assert [row.observation.sector_code for row in bottom] == [
"BK0011.DC",
"BK0010.DC",
]