Curtis Summers

It’s nice to see Oracle tipping it’s hat to PHP over the last year or so. PHP is very good at meeting specific web needs in the enterprise, and Oracle acknowledges this by including it in their Oracle HTTP Server (Apache, really) package. See the PHP Developer Center for more Oracle/PHP goodness.

I’ve been learning the Ruby language lately. This language started in Japan, but I think it’s catching on like wildfire over here now. It seems to have such a pure way about implementing its features.

Some helpful Ruby references:

• Programming Ruby – online book
• Why’s (Poignant) Guide to Ruby – book with cartoon foxes!
• Things That Newcomers to Ruby Should Know – helpful tips
• Ruby-GNOME2 – GNOME/GTK bindings for Ruby
• Ruby Articles – I suggest reading Blocks and Closures and Dynamic Productivity

I’ve recently been reading Linux Kernel Development (2nd Edition) by Robert Love.

I’m not a kernel hacker, and it’s been a while since I’ve coded in C, but I picked this up to learn more about the nitty gritty details of the operating system I use everyday. I like the special attention given to overall concepts and how certain aspects of the kernel came into being. A good read.

In my last sql post I showed a simple example of pivoting data in SQL when we know the values that we would like to pivot. Often, however, I run into the case where I must pivot data for which:

1. I don’t know the values I need to pivot, or
2. there are so many possible values to pivot that I can’t possibly give each its own column

A good example of this is summarizing Financial Aid awards by student for a particular term. The university may have thousands of different awards to give out, but a student usually has no more than 10 awards per term.

So, my source table looks something like this:

TERM    ID    AWARD    AMOUNT
----    --    -----    ------
FA04     1     PELL      4050
FA04     1   ESCHOL       500
FA04     1   WKSTDY      2000
FA04     2     PELL      3500
FA04     2   WKSTDY      2000
FA04     3     PELL      4000

Since I can’t actually pivot the values into their own columns, I’m going to create AWARD columns 1..n containing the first n awards for each student per term. In order to do this, I need to be able to number the awards per student, and then pivot the number. Oracle’s RANK functionality makes this easy:

SELECT
    TERM,
    ID,
    MAX(CASE WHEN RN = 1    THEN AWARD
             ELSE NULL END) AS   AWARD_1,
    MAX(CASE WHEN RN = 1    THEN AMOUNT
             ELSE NULL END) AS   AMOUNT_1,
    MAX(CASE WHEN RN = 2    THEN AWARD
             ELSE NULL END) AS   AWARD_2,
    MAX(CASE WHEN RN = 2    THEN AMOUNT
             ELSE NULL END) AS   AMOUNT_2,
    MAX(CASE WHEN RN = 3    THEN AWARD
             ELSE NULL END) AS   AWARD_3,
    MAX(CASE WHEN RN = 3    THEN AMOUNT
             ELSE NULL END) AS   AMOUNT_3,
FROM
    (SELECT
         RANK() OVER(PARTITION BY
                         TERM,
                         ID
                     ORDER BY
                         TERM,
                         ID,
                         AWARD)  AS RN,
         AWARD,
         AMOUNT
     FROM
         AWARD_TABLE)
GROUP BY
    TERM,
    ID)

And the results:

TERM  ID  AWARD_1  AMOUNT_1  AWARD_2  AMOUNT_2  AWARD_3  AMOUNT_3
----  --  -------  --------  -------  --------  -------  --------
FA04   1   ESCHOL       500     PELL      4050   WKSTDY      2000
FA04   2     PELL      3500   WKSTDY      2000
FA04   3     PELL      4000

My job requires lots of fun SQL magic. Below is a technique I use frequently to pivot data in SQL.

1) Pivoting values based on a known value

I often need to condense records listed in a table into one record per id by pivoting up some of the values into columns. A good example would be a test score table like so:

ID    TEST_CODE    TEST_SCORE
--    ---------    ----------
 1          ACT            21
 2          ACT            21
 3          SAT          1200
 3          ACT            24
 4          ACT            19
 4          SAT          1010
 4          SAT          1100

Since I know the TEST_CODE values that I want to pivot, I can use CASE statements to pivot the values into ACT and SAT columns. Also note that I’m grouping by ID and taking the MAX value of the specified test score. If my table only has one test score per test type, then I could just as well use MIN and get the same result. Otherwise, I’m electing to get the highest score available per person.

SELECT
    ID,
    MAX(CASE WHEN TEST_CODE = 'ACT'
             THEN TEST_SCORE
             ELSE NULL END) AS ACT_SCORE,
    MAX(CASE WHEN TEST_CODE = 'SAT'
             THEN TEST_SCORE
             ELSE NULL END) AS SAT_SCORE
FROM
    TEST_TABLE
GROUP BY
    ID;

The results:

ID    ACT_SCORE    SAT_SCORE
--    ---------    ---------
 1           21
 2           21
 3           24         1200
 4           19         1100

Pretty simple, right? Ok, but what if you need to pivot data and you don’t know the values that you want to pivot? Or, what if the there are so many possible values to pivot that you can’t afford to give each one it’s own column? I’ll answer that next time…