Question about Primary Key

Question

HI
I'm going to build a Database to automate center in my university
The Database version is Oracle 10g v2
Which data type of primary key should I user?
I used number in past but I heard there is more efficient data type for this purpose
Thanks

Answer 1

BINARY_FLOAT and BINARY_DOUBLE are designed to store floating point numbers. If you are generating synthetic primary keys using a sequence, you are generating integers. It makes no sense from a data modeling perspective to store integer data in floating point fields. And it is exceptionally dangerous if a value gets rounded unexpectedly (floating point numbers lose precision for the trailing bits)
You'll have to clarify for me what "more helpful for indexing" means. As far as storage space goes, there may be some space savings using a BINARY_FLOAT or BINARY_DOUBLE if you have very large primary key values, but there will be a space penalty for most "normal" sequence-generated values, i.e.
SQL> create table x (
  2    col1 number,
  3    col2 binary_float,
  4    col3 binary_double
  5  );
Table created.
SQL> insert into x values( 1, 2, 3 );
1 row created.
SQL> select length(col1), length(col2), length(col3) from x;
LENGTH(COL1) LENGTH(COL2) LENGTH(COL3)
           1            8            8
SQL> insert into x values (1e20, 1e20, 1e20 );
1 row created.
SQL> select length(col1), length(col2), length(col3) from x;
LENGTH(COL1) LENGTH(COL2) LENGTH(COL3)
           1            8            8
          21           15            8
SQL> select * from x;
      COL1       COL2       COL3
         1   2.0E+000   3.0E+000
1.0000E+20   1.0E+020   1.0E+020So it takes 8 times the storage to store a 1 in a BINARY_FLOAT or BINARY_DOUBLE than it does to store that in a NUMBER column. But it takes less space if you're storing values like 1e20 (a 1 with 20 0's). If you're storing values large enough to get the space benefit, however, you're going to have rounding issues that will cause the application not to function as you'd expect which is presumably a much greater problem than a bit of wasted disk space.
Justin