Add index_candidates_to_deduplicate.sql query to list indexes which could benefit from "btree index deduplication" feature in pg 13+

sql/index_candidates_to_deduplicate.sql

@@ -0,0 +1,41 @@
+--list reindex queries for btree indexes which could potentially benefit from postgresql 13 feature "btree index deduplication"
+--duplicates_estimate represents estimated part with duplicates for each indexed column (including nulls) according to MCV list in statistics
+--requires pageinspect extension to check if index already use deduplication
+
+WITH selected_indexes AS materialized (
+SELECT n.nspname AS schema_name,
+  c.relname AS table_name,
+  i.relname AS index_name,
+  pg_relation_size(x.indexrelid) AS index_size_bytes,
+  pg_relation_size(c.oid) AS table_size_bytes,
+  replace(pg_get_indexdef(i.oid), 'CREATE INDEX ', '') AS index_def,
+  (SELECT array_agg(round(f::numeric,4)) FROM
+  (
+    SELECT max(null_frac) + (1 - max(null_frac))*SUM(u) AS f FROM pg_attribute a JOIN pg_stats s ON (s.schemaname = n.nspname AND s.tablename = c.relname AND s.attname = a.attname) LEFT JOIN LATERAL unnest(s.most_common_freqs) u ON TRUE WHERE a.attrelid = x.indrelid AND a.attnum = ANY(x.indkey) GROUP BY a.attnum
+    UNION ALL 
+    SELECT max(null_frac) + (1 - max(null_frac))*SUM(u2) AS f FROM pg_stats s2 LEFT JOIN LATERAL unnest(s2.most_common_freqs) u2 ON true WHERE s2.schemaname = n.nspname AND s2.tablename = i.relname GROUP BY s2.attname
+  ) t
+  ) AS sum_most_common_freqs
+ FROM pg_index x
+   JOIN pg_class c ON c.oid = x.indrelid
+   JOIN pg_class i ON i.oid = x.indexrelid
+   LEFT JOIN pg_namespace n ON n.oid = c.relnamespace
+   JOIN pg_am am ON am.oid = i.relam AND am.amname = 'btree'--only btree indexes
+WHERE
+c.relkind IN ('r', 'm') AND i.relkind = 'i'--skip partitioned indexes
+AND (c.relpersistence = 'p' or not pg_is_in_recovery())--skip unlogged indexes
+AND x.indisunique = 'f'--skip unique indexes
+AND pg_relation_size(x.indexrelid) > 1024*1024--skip indexes smaller than 1MB
+AND n.nspname != 'pg_catalog'
+)
+SELECT
+format('-- index size: %s, table size: %s, duplicates_estimate: %s
+-- %s
+reindex index concurrently %s;
+',
+pg_size_pretty(index_size_bytes), pg_size_pretty(table_size_bytes), sum_most_common_freqs, index_def, (case when schema_name = 'public' then format('%I', index_name) else format('%I.%I', schema_name, index_name) end)) as reindex_queries
+
+FROM selected_indexes WHERE
+(bt_metap(schema_name || '.' || index_name)).allequalimage = 'f'--skip indexes already in new format
+AND 0.05 <= ALL(sum_most_common_freqs)--require at least 5% of duplicates on each indexed column
+ORDER BY index_size_bytes DESC;