Search code examples
sqlpostgresqldatetime-generation

How to generate a virtual table to generate a sequence of dates in PostgreSQL?


I'd like to generate a list of dates with the hopes of joining with another table, but I don't know what syntax to use, something similar to this:

SELECT dates.date, transactions.account_id, transactions.amount
  FROM (...) as dates
       LEFT JOIN transactions ON transactions.date = dates.date
 WHERE dates.date >= '2010-01-01' AND dates.date <= '2010-12-31'
 ORDER BY dates.date;

I want the dates so I don't have to further massage the data client-side. I'm using this to display a table similar to this:

Account    2010-01-01    2010-01-02    2010-01-03    Balance
============================================================
Chase 123        +100           -20           -70        +10
Chase 231                       +13            -9         +4

Solution

  • List of Dates

    Use the generate_series function to get a list of numbers that you can add to a date in order to get a list of dates:

    SELECT CURRENT_DATE + s.a AS dates 
      FROM generate_series(0,14,7) as s(a);
    

    Result:

    dates
    ------------
    2004-02-05
    2004-02-12
    2004-02-19
    

    Pivoting

    The latter part of your question deals with pivoting the result set -- converting row data into columnar data. PIVOT and UNPIVOT are ANSI, but I don't see them as supported by PostgreSQL currently. The most consistently supported means of pivoting a query is to use aggregate functions:

       SELECT t.account,
              SUM(CASE WHEN t.date = '2010-01-01' THEN t.amount END) AS '2010-01-01',
              SUM(CASE WHEN t.date = '2010-01-02' THEN t.amount END) AS '2010-01-02',
              SUM(CASE WHEN t.date = '2010-01-03' THEN t.amount END) AS '2010-01-03',
              SUM(t.amount) AS Balance
         FROM (SELECT CURRENT_DATE + s.a AS dates 
                 FROM generate_series(0,14,7) as s(a)) x
    LEFT JOIN TRANSACTIONS y ON y.date = x.date
     GROUP BY t.account
    

    Dynamic Columns

    ...means dynamic SQL.