Question

Using PHP/MySQL

I’m trying to create a select statement that gets the data from the least day of the current month and year (I’m using it to show data from on a certain player ‘this month’). The 1st of the months data may not always exist therefore if the first isn’t found then it would use the 2nd or 3rd or so on, whichever is the least month.

The query would need to be something like:

SELECT * FROM table_name WHERE name = '$username' AND
[...theDate = the least day of data in the current month/year]

It would return a single row of data.

This is a query I use for getting the ‘this week’ data:

SELECT name, data, theDate
FROM table_name
WHERE name = '$username'
AND YEARWEEK(theDate) = YEARWEEK(CURRENT_DATE)
ORDER BY theDate;

Answer 1

If you are interested in returning only one row, the easiest way to do this would be:

SELECT t.*
  FROM table_name t
 WHERE t.name = '$username'
   AND t.theDate >= CAST(DATE_FORMAT(NOW(),'%Y-%m-01') AS DATE)
   AND t.theDate < DATE_ADD(DATE_FORMAT(NOW(),'%Y-%m-01'), INTERVAL 1 MONTH)
 ORDER BY s.name DESC, s.theDate DESC
 LIMIT 1

The ORDER BY is based on the assumption that you have an index on (or with leading columns of) (name,theDate). That would be the most appropriate index for the predicates (i.e. conditions in the WHERE clause. There’s really no need for us to sort the name column, since we know it’s going to be equal to something… but specifying the ORDER BY in this way makes it more likely MySQL will do a reverse scan operation on the index, to return the rows in the correct order, avoiding a filesort operation.

NOTE: I specify the bare theDate column in the conditions in the WHERE clause, rather than wrapping that in any function… by specifying the bare column and a bounded range, we enable MySQL to make use of an index range scan operation. There are other possible ways to include this condition in the WHERE clause, for example…

DATE_FORMAT(t.theDate,'%Y-%m') = DATE_FORMAT(NOW(),'%Y-%m')

which will return an equivalent result, but a predicate like this is not sargable. That is, MySQL can’t/won’t do a range scan on an index to satisfy this.

If you are intending to get all the rows for the “least” date in a month for a given user (your question doesn’t seem to indicate that you need only one row), here’s one way get that result:

SELECT t.* 
  FROM table_name t
  JOIN ( SELECT s.name
              , s.theDate
           FROM table_name s 
          WHERE s.name = '$username'
            AND s.theDate >= CAST(DATE_FORMAT(NOW(),'%Y-%m-01') AS DATE)
            AND s.theDate < DATE_ADD(DATE_FORMAT(NOW(),'%Y-%m-01'), INTERVAL 1 MONTH)
          ORDER BY s.name DESC, s.theDate DESC
          LIMIT 1
       ) r
    ON r.name = t.name
   AND r.theDate = t.theDate 

Again, MySQL can make use of an index (if available) with leading columns (name,theDate) to satisfy the predicates, and to do a reverse scan operation (avoiding a sort), and to do the JOIN operation.

NOTE: We’re assuming here that ‘theDate’ is datatype DATE (with no time component). If it’s a DATETIME or a TIMESTAMP, there’s a potential for a time component, and that query may not return all rows for a given “date” value, if the time components are different for the rows with the same “date”. (e.g. ‘2012-07-13 17:30’ and ‘2012-07-13 19:55’ are different datetime values.) If we want to return both of those rows (because both are a date of “July 13”), we need to do a range scan instead of an equality test.

SELECT t.* 
  FROM table_name t
  JOIN ( SELECT s.name
              , s.theDate
           FROM table_name s 
          WHERE s.name = '$username'
            AND s.theDate >= CAST(DATE_FORMAT(NOW(),'%Y-%m-01') AS DATE)
            AND s.theDate < DATE_ADD(DATE_FORMAT(NOW(),'%Y-%m-01'), INTERVAL 1 MONTH)
          ORDER BY s.name DESC, s.theDate DESC
          LIMIT 1
       ) r
    ON t.name = r.name 
   AND t.theDate >= r.theDate
   AND t.theDate < DATE_FORMAT(DATE_ADD(r.theDate,INTERVAL 1 DAY),'%Y-%m-%d')

Note those last two lines… we’re looking for any rows with a theDate value that is greater than or equal to the “least” value found for the current month AND that is ALSO less than midnight of the following day.