Files

239 lines
9.2 KiB
Python

"""add meter table and energy_cost_period.meter_id
Introduces the ``meter`` table (one row per physical meter installation epoch)
and a nullable FK column ``energy_cost_period.meter_id`` that attributes each
billing period to a specific physical meter.
**Backfill logic (§3.7 of the M7 design doc)**:
If the database already contains any ``dsmr_reading`` or
``energy_cost_period`` rows, one initial ``meter`` row is created:
label = "Initial meter"
commodity = "electricity"
started_at = earliest dsmr_reading.recorded_at
(or, if none, earliest energy_cost_period.period_start,
or, if still none, the migration timestamp)
ended_at = NULL (still active)
reason = "initial"
All existing ``energy_cost_period`` rows are then back-filled with that
initial meter's id.
**Idempotency**: the backfill is guarded with a check for any existing
``meter`` row whose ``reason = 'initial'`` and ``commodity = 'electricity'``
and ``ended_at IS NULL``, so repeating the upgrade does not create duplicate
meters or overwrite already-filled meter_id values.
**Audit (on-non-degraded periods only)**: after backfilling, the number of
non-degraded ``energy_cost_period`` rows with ``meter_id IS NULL`` must be
zero; if it is not, the migration raises a ``RuntimeError`` and rolls back.
**Data safety**: this migration is additive only — no existing rows are
deleted or overwritten; it only creates a new table, adds a nullable column,
and back-fills that column.
Revision ID: 20260625_13_meter_table
Revises: 20260624_12_dsmr_decouple_telegram_id
Create Date: 2026-06-25 00:00:00.000000
"""
from datetime import datetime, timezone
from typing import Sequence, Union
import sqlalchemy as sa
from alembic import op
revision: str = "20260625_13_meter_table"
down_revision: Union[str, None] = "20260624_12_dsmr_decouple_telegram_id"
branch_labels: Union[str, Sequence[str], None] = None
depends_on: Union[str, Sequence[str], None] = None
def upgrade() -> None:
# ------------------------------------------------------------------ #
# 1. Create the meter table. #
# ------------------------------------------------------------------ #
op.create_table(
"meter",
sa.Column("id", sa.Integer(), autoincrement=True, nullable=False),
sa.Column("label", sa.String(length=255), nullable=False),
sa.Column("commodity", sa.String(length=32), nullable=False),
sa.Column("started_at", sa.DateTime(timezone=True), nullable=False),
sa.Column("ended_at", sa.DateTime(timezone=True), nullable=True),
sa.Column("reason", sa.String(length=64), nullable=False),
sa.Column("note", sa.String(length=1024), nullable=True),
sa.Column("created_at", sa.DateTime(timezone=True), nullable=False),
sa.PrimaryKeyConstraint("id"),
)
# ------------------------------------------------------------------ #
# 2. Add meter_id column to energy_cost_period (nullable FK). #
# ------------------------------------------------------------------ #
with op.batch_alter_table("energy_cost_period", schema=None) as batch_op:
batch_op.add_column(
sa.Column("meter_id", sa.Integer(), nullable=True)
)
batch_op.create_foreign_key(
"fk_energy_cost_period_meter_id",
"meter",
["meter_id"],
["id"],
ondelete="RESTRICT",
)
# ------------------------------------------------------------------ #
# 3. Backfill initial meter (idempotent). #
# ------------------------------------------------------------------ #
conn = op.get_bind()
# Check if there is already an initial meter (idempotency guard).
existing_initial = conn.execute(
sa.text(
"SELECT id FROM meter "
"WHERE reason = 'initial' AND commodity = 'electricity' AND ended_at IS NULL "
"LIMIT 1"
)
).fetchone()
if existing_initial is not None:
# Already backfilled — nothing to do.
return
# Determine whether there is any historical data to create a meter for.
has_readings = conn.execute(
sa.text("SELECT 1 FROM dsmr_reading LIMIT 1")
).fetchone()
has_periods = conn.execute(
sa.text("SELECT 1 FROM energy_cost_period LIMIT 1")
).fetchone()
if not has_readings and not has_periods:
# Empty database: no historical data, so no initial meter is needed.
# meter_id will remain NULL on any future rows until T02 service layer
# starts populating it.
return
# Determine started_at: earliest dsmr_reading.recorded_at, falling back to
# earliest energy_cost_period.period_start, and finally to now().
earliest_reading_row = conn.execute(
sa.text("SELECT MIN(recorded_at) AS ts FROM dsmr_reading")
).fetchone()
earliest_period_row = conn.execute(
sa.text("SELECT MIN(period_start) AS ts FROM energy_cost_period")
).fetchone()
started_at_value: datetime | None = None
if earliest_reading_row and earliest_reading_row[0] is not None:
# SQLite returns ISO strings for datetime columns; parse to datetime.
raw = earliest_reading_row[0]
started_at_value = _parse_sqlite_datetime(raw)
if started_at_value is None and earliest_period_row and earliest_period_row[0] is not None:
raw = earliest_period_row[0]
started_at_value = _parse_sqlite_datetime(raw)
if started_at_value is None:
started_at_value = datetime.now(tz=timezone.utc)
now_utc = datetime.now(tz=timezone.utc)
# Insert the initial meter row.
conn.execute(
sa.text(
"INSERT INTO meter (label, commodity, started_at, ended_at, reason, note, created_at) "
"VALUES (:label, :commodity, :started_at, NULL, :reason, NULL, :created_at)"
),
{
"label": "Initial meter",
"commodity": "electricity",
"started_at": _iso(started_at_value),
"reason": "initial",
"created_at": _iso(now_utc),
},
)
# Retrieve the newly created meter id.
meter_row = conn.execute(
sa.text(
"SELECT id FROM meter "
"WHERE reason = 'initial' AND commodity = 'electricity' AND ended_at IS NULL "
"LIMIT 1"
)
).fetchone()
assert meter_row is not None, "Initial meter row not found after insert"
meter_id: int = meter_row[0]
# Back-fill all existing energy_cost_period rows that have meter_id IS NULL.
conn.execute(
sa.text(
"UPDATE energy_cost_period SET meter_id = :mid WHERE meter_id IS NULL"
),
{"mid": meter_id},
)
# ------------------------------------------------------------------ #
# 4. Audit: verify all non-degraded periods have a meter_id. #
# ------------------------------------------------------------------ #
unmatched_row = conn.execute(
sa.text(
"SELECT COUNT(*) FROM energy_cost_period "
"WHERE meter_id IS NULL AND degraded = 0"
)
).fetchone()
unmatched_count: int = unmatched_row[0] if unmatched_row else 0
if unmatched_count != 0:
raise RuntimeError(
f"Meter backfill audit failed: {unmatched_count} non-degraded "
"energy_cost_period row(s) still have meter_id IS NULL after backfill. "
"Migration aborted to protect data integrity."
)
def downgrade() -> None:
# Remove the FK column from energy_cost_period first (references meter).
with op.batch_alter_table("energy_cost_period", schema=None) as batch_op:
batch_op.drop_constraint("fk_energy_cost_period_meter_id", type_="foreignkey")
batch_op.drop_column("meter_id")
# Drop the meter table.
op.drop_table("meter")
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def _parse_sqlite_datetime(value: str | datetime) -> datetime:
"""Parse a SQLite datetime value into an aware UTC datetime.
SQLite stores datetimes as ISO 8601 strings. SQLAlchemy may return them
as plain strings or as naive datetimes (no tzinfo) depending on the driver
and column declaration. This helper normalises both forms to an aware UTC
``datetime``.
"""
if isinstance(value, datetime):
if value.tzinfo is None:
return value.replace(tzinfo=timezone.utc)
return value
# String form — strip trailing Z or +00:00 variants, then attach UTC.
s = str(value).strip()
for suffix in ("+00:00", "Z", " UTC"):
if s.endswith(suffix):
s = s[: -len(suffix)]
# SQLite uses space as the T separator in some formats.
s = s.replace(" ", "T")
try:
dt = datetime.fromisoformat(s)
except ValueError:
# Fallback: strip subseconds if present to handle unusual formats.
dt = datetime.strptime(s[:19], "%Y-%m-%dT%H:%M:%S")
return dt.replace(tzinfo=timezone.utc)
def _iso(dt: datetime) -> str:
"""Serialise a datetime to an ISO 8601 string for SQLite storage."""
if dt.tzinfo is not None:
dt = dt.astimezone(timezone.utc).replace(tzinfo=None)
return dt.strftime("%Y-%m-%dT%H:%M:%S")