|
| 1 | +#!/usr/bin/env bash |
| 2 | +# cleanup-implausible-years.sh |
| 3 | +# |
| 4 | +# One-time cleanup: nulls out implausible release/master years from existing |
| 5 | +# data in Neo4j and PostgreSQL. |
| 6 | +# |
| 7 | +# Background: a Discogs *release* carries its date in <released> (a date string), |
| 8 | +# not <year>, so the extractor's year-range rules β which key on the `year` |
| 9 | +# field β never fired for releases. The release year is derived consumer-side |
| 10 | +# in common/data_normalizer.py, which (prior to this fix) only rejected year 0. |
| 11 | +# As a result, releases with antiquity/sentinel dates (e.g. "0400-01-01") were |
| 12 | +# stored as year 400, polluting the Insights "Genre Trends" chart. |
| 13 | +# |
| 14 | +# The code fix (common/data_normalizer.py) stops new bad years at ingest. This |
| 15 | +# script brings *existing* data in line without a full re-ingest. New ingests |
| 16 | +# need no cleanup. |
| 17 | +# |
| 18 | +# Scope: Discogs Release/Master nodes (Neo4j) and the public `releases`/`masters` |
| 19 | +# tables (PostgreSQL). MusicBrainz entities take a different ingest path and are |
| 20 | +# out of scope. |
| 21 | +# |
| 22 | +# Plausible range: [1860, current_year + 1] β must stay in sync with |
| 23 | +# MIN_RELEASE_YEAR in common/data_normalizer.py and extractor/extraction-rules.yaml. |
| 24 | +# |
| 25 | +# Usage: |
| 26 | +# ./scripts/cleanup-implausible-years.sh # dry run β counts only, no changes |
| 27 | +# ./scripts/cleanup-implausible-years.sh --apply # perform the cleanup |
| 28 | +# |
| 29 | +# Environment variables (all optional, defaults shown): |
| 30 | +# NEO4J_CONTAINER docker container name (default: discogsography-neo4j) |
| 31 | +# NEO4J_USER Neo4j username (default: neo4j) |
| 32 | +# NEO4J_PASSWORD Neo4j password (default: discogsography) |
| 33 | +# POSTGRES_CONTAINER docker container name (default: discogsography-postgres) |
| 34 | +# POSTGRES_USER PostgreSQL username (default: discogsography) |
| 35 | +# POSTGRES_DB PostgreSQL database (default: discogsography) |
| 36 | + |
| 37 | +set -euo pipefail |
| 38 | + |
| 39 | +NEO4J_CONTAINER="${NEO4J_CONTAINER:-discogsography-neo4j}" |
| 40 | +NEO4J_USER="${NEO4J_USER:-neo4j}" |
| 41 | +NEO4J_PASSWORD="${NEO4J_PASSWORD:-discogsography}" |
| 42 | +POSTGRES_CONTAINER="${POSTGRES_CONTAINER:-discogsography-postgres}" |
| 43 | +POSTGRES_USER="${POSTGRES_USER:-discogsography}" |
| 44 | +POSTGRES_DB="${POSTGRES_DB:-discogsography}" |
| 45 | + |
| 46 | +MIN_YEAR=1860 |
| 47 | +MAX_YEAR=$(($(date +%Y) + 1)) |
| 48 | + |
| 49 | +APPLY=0 |
| 50 | +if [[ "${1:-}" == "--apply" ]]; then |
| 51 | + APPLY=1 |
| 52 | +elif [[ -n "${1:-}" ]]; then |
| 53 | + echo "β Unknown argument '${1}'. Use --apply to perform the cleanup (omit for a dry run)." |
| 54 | + exit 1 |
| 55 | +fi |
| 56 | + |
| 57 | +if [[ "${APPLY}" -eq 1 ]]; then |
| 58 | + echo "β οΈ APPLY mode: implausible years will be set to null." |
| 59 | +else |
| 60 | + echo "π§ͺ DRY RUN: counting affected records only. Re-run with --apply to make changes." |
| 61 | +fi |
| 62 | +echo "π
Plausible year range: [${MIN_YEAR}, ${MAX_YEAR}]" |
| 63 | +echo "" |
| 64 | + |
| 65 | +run_neo4j() { |
| 66 | + docker exec "${NEO4J_CONTAINER}" cypher-shell -u "${NEO4J_USER}" -p "${NEO4J_PASSWORD}" "$@" |
| 67 | +} |
| 68 | + |
| 69 | +run_pg() { |
| 70 | + docker exec "${POSTGRES_CONTAINER}" psql -U "${POSTGRES_USER}" -d "${POSTGRES_DB}" -t -A "$@" |
| 71 | +} |
| 72 | + |
| 73 | +# ββ Neo4j ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ |
| 74 | +echo "π Checking Neo4j container '${NEO4J_CONTAINER}' is running..." |
| 75 | +if ! docker inspect --format '{{.State.Running}}' "${NEO4J_CONTAINER}" 2>/dev/null | grep -q true; then |
| 76 | + echo "β Container '${NEO4J_CONTAINER}' is not running. Start it first with: docker compose up -d neo4j" |
| 77 | + exit 1 |
| 78 | +fi |
| 79 | + |
| 80 | +for label in Release Master; do |
| 81 | + COUNT=$(run_neo4j --format plain \ |
| 82 | + "MATCH (n:${label}) WHERE n.year IS NOT NULL AND (n.year < ${MIN_YEAR} OR n.year > ${MAX_YEAR}) RETURN count(n) AS n;" | |
| 83 | + tail -1) |
| 84 | + echo " ${label}: ${COUNT} node(s) with an implausible year." |
| 85 | + if [[ "${APPLY}" -eq 1 && "${COUNT}" -gt 0 ]]; then |
| 86 | + echo " π Nulling ${label}.year for out-of-range values..." |
| 87 | + run_neo4j \ |
| 88 | + "MATCH (n:${label}) |
| 89 | + WHERE n.year IS NOT NULL AND (n.year < ${MIN_YEAR} OR n.year > ${MAX_YEAR}) |
| 90 | + CALL { WITH n SET n.year = null } IN TRANSACTIONS OF 50000 ROWS;" |
| 91 | + fi |
| 92 | +done |
| 93 | +echo "" |
| 94 | + |
| 95 | +# ββ PostgreSQL βββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ |
| 96 | +echo "π Checking PostgreSQL container '${POSTGRES_CONTAINER}' is running..." |
| 97 | +if ! docker inspect --format '{{.State.Running}}' "${POSTGRES_CONTAINER}" 2>/dev/null | grep -q true; then |
| 98 | + echo "β Container '${POSTGRES_CONTAINER}' is not running. Start it first with: docker compose up -d postgres" |
| 99 | + exit 1 |
| 100 | +fi |
| 101 | + |
| 102 | +for table in releases masters; do |
| 103 | + COUNT=$(run_pg -c \ |
| 104 | + "SELECT count(*) FROM ${table} |
| 105 | + WHERE data->>'year' ~ '^[0-9]+\$' |
| 106 | + AND ((data->>'year')::int < ${MIN_YEAR} OR (data->>'year')::int > ${MAX_YEAR});") |
| 107 | + echo " ${table}: ${COUNT} row(s) with an implausible year." |
| 108 | + if [[ "${APPLY}" -eq 1 && "${COUNT}" -gt 0 ]]; then |
| 109 | + echo " π Setting ${table}.data->'year' to null for out-of-range values..." |
| 110 | + run_pg -c \ |
| 111 | + "UPDATE ${table} |
| 112 | + SET data = jsonb_set(data, '{year}', 'null'::jsonb) |
| 113 | + WHERE data->>'year' ~ '^[0-9]+\$' |
| 114 | + AND ((data->>'year')::int < ${MIN_YEAR} OR (data->>'year')::int > ${MAX_YEAR});" |
| 115 | + fi |
| 116 | +done |
| 117 | +echo "" |
| 118 | + |
| 119 | +if [[ "${APPLY}" -eq 1 ]]; then |
| 120 | + echo "β
Cleanup complete." |
| 121 | +else |
| 122 | + echo "β
Dry run complete. Re-run with --apply to null the years counted above." |
| 123 | +fi |
0 commit comments