UNIX for Dummies Questions & Answers

I want to calculate day-to-day changes in a data set. In an ideal situation the following would suffice:
Date x
2007-01-01 5
2007-01-02 8
2007-01-03 4
2007-01-04 3
2007-01-05 10
2007-01-06 10
2007-01-07 2
2007-01-08 10
2007-01-09 1
2007-01-10 8

SELECT D1.date, (D1.x-D2.x)/D2.x AS change
FROM data AS D1 INNER JOIN data AS D2 ON D1.date=D2.date+1;

This will not work however for non-contiguous dates. In other words , what would be the suggested SQL query for a data set as follows ?

2007-01-01 5
2007-01-02 8
2007-01-03 4
2007-01-04 3
2007-01-05 10
2007-01-08 10
2007-01-09 2
2007-01-10 10
2007-01-11 1
2007-01-12 8

where day 6 and 7 constitute a weekend for which no samples are present. My initial thought would be to have a temporary table with the following structure : id - date - x
where id is a contiguous autonumber and the resulting query would be:

SELECT TD1.date, (TD1.x-TD2.x)/TD2.x AS change
FROM tempdata AS TD1 INNER JOIN tempdata AS TD2 ON TD1.id=TD2.id+1;

But the question is, how should this table be constructed, particularly because not all databases support autonumbers? Any other approaches which are perhaps more straightforward?

If using Oracle, then try the LAG analytic function. It provides access to more than one row of a table at the same time without a self join.

Thank you for your response. Would appreciate an ANSI compliant method for portability purposes.
I have also thought about a procedural method, whereby the data is read into an array and sorted from there. Not too appealing either, prefer pure SQL.