Files
weatherfeeder/internal/sinks/postgres/doc.go
2026-06-16 19:57:48 -05:00

293 lines
15 KiB
Go

// Package postgres documents weatherfeeder's PostgreSQL sink contract for
// downstream SQL consumers.
//
// This package wires weatherfeeder canonical events into normalized relational
// tables. Downstream consumers can reconstruct the same canonical JSON objects
// that were written by joining parent/child tables as described below.
//
// Canonical input schemas:
// - weather.observation.v1 -> model.WeatherObservation
// - weather.forecast.v1 -> model.WeatherForecastRun
// - weather.forecast_discussion.v1 -> model.WeatherForecastDiscussion
// - weather.weather_story.v1 -> model.WeatherStoryRun
// - weather.alert.v1 -> model.WeatherAlertRun
// - weather.outlook.v2 -> model.WeatherOutlookRun
//
// Parent/child relationships:
// - observations.event_id -> observation_present_weather.event_id
// - forecasts.event_id -> forecast_periods.run_event_id
// - forecast_discussions.event_id -> forecast_discussion_key_messages.run_event_id
// - weather_story_runs.event_id -> weather_stories.run_event_id
// - alert_runs.event_id -> alerts.run_event_id
// - alerts.(run_event_id, alert_index) -> alert_references.(run_event_id, alert_index)
// - outlook_runs.event_id -> outlooks.run_event_id
// - outlook_runs.event_id -> outlook_discussions.run_event_id
//
// Dedupe and retention behavior:
// - Parent primary keys (event_id): observations, forecasts, alert_runs, outlook_runs.
// - Child primary keys use positional indexes to preserve payload order.
// - Prune columns:
// - observations.observed_at
// - observation_present_weather.observed_at
// - forecasts.issued_at
// - forecast_periods.issued_at
// - forecast_discussions.issued_at
// - forecast_discussion_key_messages.issued_at
// - weather_story_runs.as_of
// - weather_stories.as_of
// - alert_runs.as_of
// - alerts.as_of
// - alert_references.as_of
// - outlook_runs.as_of
// - outlooks.as_of
// - outlook_discussions.as_of
//
// Envelope field mapping (shared parent columns)
//
// These columns exist on parent tables such as observations, forecasts,
// forecast_discussions, weather_story_runs, alert_runs, and outlook_runs:
// - event_id TEXT -> event.id
// - event_kind TEXT -> event.kind
// - event_source TEXT -> event.source
// - event_schema TEXT -> event.schema
// - event_emitted_at TIMESTAMPTZ -> event.emitted_at
// - event_effective_at TIMESTAMPTZ NULL -> event.effective_at
//
// Table contract
//
// 1. observations (PK: event_id)
//
// - event_id TEXT -> event.id
// - event_kind TEXT -> event.kind
// - event_source TEXT -> event.source
// - event_schema TEXT -> event.schema
// - event_emitted_at TIMESTAMPTZ -> event.emitted_at
// - event_effective_at TIMESTAMPTZ NULL -> event.effective_at
// - station_id TEXT NULL -> payload.stationId
// - station_name TEXT NULL -> payload.stationName
// - observed_at TIMESTAMPTZ -> payload.timestamp
// - condition_code INTEGER -> payload.conditionCode
// - is_day BOOLEAN NULL -> payload.isDay
// - text_description TEXT NULL -> payload.textDescription
// - temperature_c DOUBLE PRECISION NULL -> payload.temperatureC
// - dewpoint_c DOUBLE PRECISION NULL -> payload.dewpointC
// - wind_direction_degrees DOUBLE PRECISION NULL -> payload.windDirectionDegrees
// - wind_speed_kmh DOUBLE PRECISION NULL -> payload.windSpeedKmh
// - wind_gust_kmh DOUBLE PRECISION NULL -> payload.windGustKmh
// - barometric_pressure_pa DOUBLE PRECISION NULL -> payload.barometricPressurePa
// - visibility_meters DOUBLE PRECISION NULL -> payload.visibilityMeters
// - relative_humidity_percent DOUBLE PRECISION NULL -> payload.relativeHumidityPercent
// - apparent_temperature_c DOUBLE PRECISION NULL -> payload.apparentTemperatureC
//
// 2. observation_present_weather (PK: event_id, weather_index)
//
// - event_id TEXT -> observations.event_id / payload.presentWeather[i]
// - weather_index INTEGER -> i (array position in payload.presentWeather)
// - observed_at TIMESTAMPTZ -> payload.timestamp
// - raw_text TEXT NULL -> JSON-encoded payload.presentWeather[i].raw
//
// Note: raw_text stores compact JSON text. Consumers that need the original
// object should parse raw_text as JSON.
//
// 3. forecasts (PK: event_id)
//
// - event_id TEXT -> event.id
// - event_kind TEXT -> event.kind
// - event_source TEXT -> event.source
// - event_schema TEXT -> event.schema
// - event_emitted_at TIMESTAMPTZ -> event.emitted_at
// - event_effective_at TIMESTAMPTZ NULL -> event.effective_at
// - location_id TEXT NULL -> payload.locationId
// - location_name TEXT NULL -> payload.locationName
// - issued_at TIMESTAMPTZ -> payload.issuedAt
// - updated_at TIMESTAMPTZ NULL -> payload.updatedAt
// - product TEXT -> payload.product
// - latitude DOUBLE PRECISION NULL -> payload.latitude
// - longitude DOUBLE PRECISION NULL -> payload.longitude
// - elevation_meters DOUBLE PRECISION NULL -> payload.elevationMeters
// - period_count INTEGER -> len(payload.periods)
//
// 4. forecast_periods (PK: run_event_id, period_index)
//
// - run_event_id TEXT -> forecasts.event_id / payload.periods[i]
// - period_index INTEGER -> i (array position in payload.periods)
// - issued_at TIMESTAMPTZ -> payload.issuedAt (copied from parent)
// - start_time TIMESTAMPTZ -> payload.periods[i].startTime
// - end_time TIMESTAMPTZ -> payload.periods[i].endTime
// - name TEXT NULL -> payload.periods[i].name
// - is_day BOOLEAN NULL -> payload.periods[i].isDay
// - condition_code INTEGER NULL -> payload.periods[i].conditionCode
// - text_description TEXT NULL -> payload.periods[i].textDescription
// - temperature_c DOUBLE PRECISION NULL -> payload.periods[i].temperatureC
// - temperature_c_min DOUBLE PRECISION NULL -> payload.periods[i].temperatureCMin
// - temperature_c_max DOUBLE PRECISION NULL -> payload.periods[i].temperatureCMax
// - dewpoint_c DOUBLE PRECISION NULL -> payload.periods[i].dewpointC
// - relative_humidity_percent DOUBLE PRECISION NULL -> payload.periods[i].relativeHumidityPercent
// - wind_direction_degrees DOUBLE PRECISION NULL -> payload.periods[i].windDirectionDegrees
// - wind_speed_kmh DOUBLE PRECISION NULL -> payload.periods[i].windSpeedKmh
// - wind_gust_kmh DOUBLE PRECISION NULL -> payload.periods[i].windGustKmh
// - barometric_pressure_pa DOUBLE PRECISION NULL -> payload.periods[i].barometricPressurePa
// - visibility_meters DOUBLE PRECISION NULL -> payload.periods[i].visibilityMeters
// - apparent_temperature_c DOUBLE PRECISION NULL -> payload.periods[i].apparentTemperatureC
// - cloud_cover_percent DOUBLE PRECISION NULL -> payload.periods[i].cloudCoverPercent
// - probability_of_precipitation_percent DOUBLE PRECISION NULL -> payload.periods[i].probabilityOfPrecipitationPercent
// - precipitation_amount_mm DOUBLE PRECISION NULL -> payload.periods[i].precipitationAmountMm
// - snowfall_depth_mm DOUBLE PRECISION NULL -> payload.periods[i].snowfallDepthMm
// - uv_index DOUBLE PRECISION NULL -> payload.periods[i].uvIndex
//
// 5. weather_story_runs (PK: event_id)
//
// - event_id TEXT -> event.id
// - event_kind TEXT -> event.kind
// - event_source TEXT -> event.source
// - event_schema TEXT -> event.schema
// - event_emitted_at TIMESTAMPTZ -> event.emitted_at
// - event_effective_at TIMESTAMPTZ NULL -> event.effective_at
// - office_id TEXT NULL -> payload.officeId
// - as_of TIMESTAMPTZ -> payload.asOf
// - story_count INTEGER -> len(payload.stories)
//
// 6. weather_stories (PK: run_event_id, story_index)
//
// - run_event_id TEXT -> weather_story_runs.event_id / payload.stories[i]
// - story_index INTEGER -> i (array position in payload.stories)
// - as_of TIMESTAMPTZ -> payload.asOf (copied from parent)
// - office_id TEXT NULL -> payload.stories[i].officeId
// - start_time TIMESTAMPTZ -> payload.stories[i].startTime
// - end_time TIMESTAMPTZ -> payload.stories[i].endTime
// - updated_at TIMESTAMPTZ -> payload.stories[i].updatedAt
// - title TEXT NULL -> payload.stories[i].title
// - description TEXT NULL -> payload.stories[i].description
// - alt_text TEXT NULL -> payload.stories[i].altText
// - priority BOOLEAN -> payload.stories[i].priority
// - story_order INTEGER -> payload.stories[i].order
// - download_url TEXT NULL -> payload.stories[i].downloadUrl
//
// 7. alert_runs (PK: event_id)
//
// - event_id TEXT -> event.id
// - event_kind TEXT -> event.kind
// - event_source TEXT -> event.source
// - event_schema TEXT -> event.schema
// - event_emitted_at TIMESTAMPTZ -> event.emitted_at
// - event_effective_at TIMESTAMPTZ NULL -> event.effective_at
// - location_id TEXT NULL -> payload.locationId
// - location_name TEXT NULL -> payload.locationName
// - as_of TIMESTAMPTZ -> payload.asOf
// - latitude DOUBLE PRECISION NULL -> payload.latitude
// - longitude DOUBLE PRECISION NULL -> payload.longitude
// - alert_count INTEGER -> len(payload.alerts)
//
// 8. alerts (PK: run_event_id, alert_index)
//
// - run_event_id TEXT -> alert_runs.event_id / payload.alerts[i]
// - alert_index INTEGER -> i (array position in payload.alerts)
// - as_of TIMESTAMPTZ -> payload.asOf (copied from parent)
// - alert_id TEXT -> payload.alerts[i].id
// - event TEXT NULL -> payload.alerts[i].event
// - headline TEXT NULL -> payload.alerts[i].headline
// - severity TEXT NULL -> payload.alerts[i].severity
// - urgency TEXT NULL -> payload.alerts[i].urgency
// - certainty TEXT NULL -> payload.alerts[i].certainty
// - status TEXT NULL -> payload.alerts[i].status
// - message_type TEXT NULL -> payload.alerts[i].messageType
// - category TEXT NULL -> payload.alerts[i].category
// - response TEXT NULL -> payload.alerts[i].response
// - description TEXT NULL -> payload.alerts[i].description
// - instruction TEXT NULL -> payload.alerts[i].instruction
// - sent TIMESTAMPTZ NULL -> payload.alerts[i].sent
// - effective TIMESTAMPTZ NULL -> payload.alerts[i].effective
// - onset TIMESTAMPTZ NULL -> payload.alerts[i].onset
// - ends TIMESTAMPTZ NULL -> payload.alerts[i].ends
// - expires TIMESTAMPTZ NULL -> payload.alerts[i].expires
// - area_description TEXT NULL -> payload.alerts[i].areaDescription
// - sender_name TEXT NULL -> payload.alerts[i].senderName
// - reference_count INTEGER -> len(payload.alerts[i].references)
//
// 9. alert_references (PK: run_event_id, alert_index, reference_index)
//
// - run_event_id TEXT -> alert_runs.event_id / payload.alerts[i].references[j]
// - alert_index INTEGER -> i (array position in payload.alerts)
// - reference_index INTEGER -> j (array position in payload.alerts[i].references)
// - as_of TIMESTAMPTZ -> payload.asOf (copied from parent)
// - id TEXT NULL -> payload.alerts[i].references[j].id
// - identifier TEXT NULL -> payload.alerts[i].references[j].identifier
// - sender TEXT NULL -> payload.alerts[i].references[j].sender
// - sent TIMESTAMPTZ NULL -> payload.alerts[i].references[j].sent
//
// 10. outlook_runs (PK: event_id)
//
// - event_id TEXT -> event.id
// - event_kind TEXT -> event.kind
// - event_source TEXT -> event.source
// - event_schema TEXT -> event.schema
// - event_emitted_at TIMESTAMPTZ -> event.emitted_at
// - event_effective_at TIMESTAMPTZ NULL -> event.effective_at
// - location_id TEXT NULL -> payload.locationId
// - location_name TEXT NULL -> payload.locationName
// - latitude DOUBLE PRECISION NULL -> payload.latitude
// - longitude DOUBLE PRECISION NULL -> payload.longitude
// - as_of TIMESTAMPTZ -> payload.asOf
// - issued_at TIMESTAMPTZ NULL -> payload.issuedAt
// - outlook_count INTEGER -> len(payload.outlooks)
// - discussion_count INTEGER -> len(payload.discussions)
//
// 11. outlooks (PK: run_event_id, outlook_index)
//
// - run_event_id TEXT -> outlook_runs.event_id / payload.outlooks[i]
// - outlook_index INTEGER -> i (array position in payload.outlooks)
// - as_of TIMESTAMPTZ -> payload.asOf (copied from parent)
// - outlook_id TEXT -> payload.outlooks[i].id
// - provider TEXT -> payload.outlooks[i].provider
// - product TEXT -> payload.outlooks[i].product
// - day INTEGER -> payload.outlooks[i].day
// - outlook_type TEXT -> payload.outlooks[i].outlookType
// - label TEXT -> payload.outlooks[i].label
// - label_text TEXT NULL -> payload.outlooks[i].labelText
// - severity_rank INTEGER NULL -> payload.outlooks[i].severityRank
// - valid_from TIMESTAMPTZ -> payload.outlooks[i].validFrom
// - valid_to TIMESTAMPTZ -> payload.outlooks[i].validTo
// - issued_at TIMESTAMPTZ -> payload.outlooks[i].issuedAt
// - expires_at TIMESTAMPTZ -> payload.outlooks[i].expiresAt
// - forecaster TEXT NULL -> payload.outlooks[i].forecaster
// - source_url TEXT NULL -> payload.outlooks[i].sourceUrl
// - image_url TEXT NULL -> payload.outlooks[i].imageUrl
// - contains_location BOOLEAN -> payload.outlooks[i].containsLocation
// - geometry_json TEXT -> compact JSON payload.outlooks[i].geometry
//
// 12. outlook_discussions (PK: run_event_id, discussion_index)
//
// - run_event_id TEXT -> outlook_runs.event_id / payload.discussions[i]
// - discussion_index INTEGER -> i (array position in payload.discussions)
// - as_of TIMESTAMPTZ -> payload.asOf (copied from parent)
// - day INTEGER -> payload.discussions[i].day
// - headline TEXT NULL -> payload.discussions[i].headline
// - summary TEXT NULL -> payload.discussions[i].summary
// - discussion TEXT NULL -> payload.discussions[i].discussion
// - updated_at TIMESTAMPTZ NULL -> payload.discussions[i].updatedAt
//
// Reconstructing canonical JSON payloads
//
// - WeatherObservation:
// read one row from observations, then join child rows by event_id ordered by
// weather_index to rebuild presentWeather arrays.
//
// - WeatherForecastRun:
// read one row from forecasts, then join forecast_periods by run_event_id
// ordered by period_index to rebuild periods.
//
// - WeatherStoryRun:
// read one row from weather_story_runs, then join weather_stories by
// run_event_id ordered by story_index to rebuild stories.
//
// - WeatherAlertRun:
// read one row from alert_runs, join alerts by run_event_id ordered by
// alert_index, then join alert_references by (run_event_id, alert_index)
// ordered by reference_index to rebuild references per alert.
//
// - WeatherOutlookRun:
// read one row from outlook_runs, join outlooks by run_event_id ordered by
// outlook_index to rebuild outlooks, then join outlook_discussions by
// run_event_id ordered by discussion_index to rebuild discussions.
package postgres