Find all those columns which have only null values, in a MySQL table
mysql, sql
Solution
You can avoid using a procedure by dynamically creating (from the `INFORMATION_SCHEMA.COLUMNS` table) a string that contains the SQL you wish to execute, then preparing a statement from that string and executing it.
The SQL we wish to build will look like:
SELECT * FROM (
SELECT 'tableA' AS `table`,
IF(COUNT(`column_a`), NULL, 'column_a') AS `column`
FROM tableA
UNION ALL
SELECT 'tableB' AS `table`,
IF(COUNT(`column_b`), NULL, 'column_b') AS `column`
FROM tableB
UNION ALL
-- etc.
) t WHERE `column` IS NOT NULL
This can be done using the following:
SET group_concat_max_len = 4294967295; -- to overcome default 1KB limitation
SELECT CONCAT(
'SELECT * FROM ('
, GROUP_CONCAT(
'SELECT ', QUOTE(TABLE_NAME), ' AS `table`,'
, 'IF('
, 'COUNT(`', REPLACE(COLUMN_NAME, '`', '``'), '`),'
, 'NULL,'
, QUOTE(COLUMN_NAME)
, ') AS `column` '
, 'FROM `', REPLACE(TABLE_NAME, '`', '``'), '`'
SEPARATOR ' UNION ALL '
)
, ') t WHERE `column` IS NOT NULL'
)
INTO @sql
FROM INFORMATION_SCHEMA.COLUMNS
WHERE TABLE_SCHEMA = DATABASE();
PREPARE stmt FROM @sql;
EXECUTE stmt;
DEALLOCATE PREPARE stmt;
See it on sqlfiddle.
Problem
The situation is as follows: I have a substantial number of tables, with each a substantial number of columns. I need to deal with this old and to-be-deprecated database for a new system, and I'm looking for a way to eliminate all columns that have - apparently - never been in use. I wanna do this by filtering out all columns that have a value on any given row, leaving me with a set of columns where the value is NULL in all rows. Of course I could manually sort every column descending, but that'd take too long as I'm dealing with loads of tables and columns. I estimate it to be 400 tables with up to 50 (!) columns per table. Is there any way I can get this information from the information_schema? EDIT: Here's an example: ``` column_a column_b column_c column_d NULL NULL NULL 1 NULL 1 NULL 1 NULL 1 NULL NULL NULL NULL NULL NULL ``` The output should be 'column_a' and 'column_c', for being the only columns without any filled in values.