Data normalization

From SQLZOO
Revision as of 11:32, 17 July 2012 by Marek (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

To select the data from 20 different columns into one changing the values according to one other column consisting the positions of the 20 columns.

schema:scott

Sometimes we have un-normailsed data that we want to normalise. Suppose we have data in 20 columns F1 to F20:

Line    F1    F2   F3   F4 ...
------------------------------
A       11    10   13   15 ...
B       20    22   23   28 ...

But we want a table that has just one data column. Like this...

Line   Col   Val
----------------
A      1     11
A      2     10
A      3     13
A      4     15
...
B      1     20
B      2     22
B      3     23
B      4     28
...

You can use INSERT ... SELECT ... statement

DROP TABLE normal
DROP TABLE unnormal;
CREATE TABLE unnormal
  (Line CHAR(1) PRIMARY KEY,
   F1 INTEGER,
   F2 INTEGER,
   F3 INTEGER,
   F4 INTEGER);
--Put the bad data in
INSERT INTO unnormal VALUES
 ('A', 11, 10, 13, 15);
INSERT INTO unnormal VALUES
 ('B', 20, 22, 23, 28);
--Create the good table
CREATE TABLE normal
 (Line CHAR(1), col INTEGER, val INTEGER,
  PRIMARY KEY (Line, col) );
--Copy the data into it
INSERT INTO normal(Line, col, val)
  SELECT Line, 1, F1 FROM unnormal
UNION
  SELECT Line, 2, F2 FROM unnormal
UNION
  SELECT Line, 3, F3 FROM unnormal
UNION
  SELECT Line, 4, F4 FROM unnormal;
SELECT * FROM normal;
Personal tools
Namespaces

Variants
Actions
Reference
Toolbox
Google AdSense