Table Structure

Table Structure

• longer, narrower tables are generally better than wide, short tables
• tradeoff between query speed and index size
• NULLs
• 3NF
  • Idea: don’t duplicate data, use keys and more tables; keep like with like

Keys & Indices

• PRIMARY: every row should have a unique identifier
  • a table may have only one PRIMARY key
  • generally bad practice not to have a PRIMARY KEY
• INDEX (see below)
• UNIQUE: unlike PRIMARY, can have NULLs
• search may be \(O(\log{n})\) instead of \(O(n)\)
• Optimizing queries with EXPLAIN

Foreign Keys

• Referential Integrity:
  • every value in a FOREIGN KEY must correspond to an entry in another table’s PRIMARY KEY
  • unidirectional
  • one-to-one, one-to-many
• Entity Relationship Schema

Partitioning

• Inheritance
• split the data into separate logical blocks

Housekeeping

• VACUUM ANALYZE

Example

DROP INDEX IF EXISTS individuals_state_idx;
DROP INDEX IF EXISTS committees_cmte_st_idx;

-- wicked slow, running time forever...
-- note sequential scans
EXPLAIN SELECT sum(1) as N FROM committees c
JOIN individuals i ON c.cmte_st = i.state
WHERE i.state = 'MA';

-- Build index on big table
CREATE INDEX ON individuals (state);

-- now it uses the index, running time 
EXPLAIN SELECT sum(1) as N FROM committees c
JOIN individuals i ON c.cmte_st = i.state
WHERE i.state = 'MA';

-- Build index on small table
CREATE INDEX ON committees (cmte_st);

-- now it uses BOTH indexes, running time 
EXPLAIN SELECT sum(1) as N FROM committees c
JOIN individuals i ON c.cmte_st = i.state
WHERE i.state = 'MA';