First install the oracle instant client somewhere - I normally install it to /usr/local/oracle/instantclient_10_2. Then…

cd /usr/local/oracle/instantclient_10_2
ln -s libclntsh.dylib.10.1 libclntsh.dylib

Following this, run mkdir -p /usr/local/oracle/network/admin/, and place a copy of the tnsnames.ora file in this directory.

Now you need to add the following to your .profile or .bashrc file:

# Oracle env variables
export ORACLE_HOME="/usr/local/oracle/instantclient_10_2"
export DYLD_LIBRARY_PATH="/usr/local/oracle/instantclient_10_2:$DYLD_LIBRARY_PATH"
export SQLPATH="/usr/local/oracle/instantclient_10_2"
export TNS_ADMIN="/usr/local/oracle/network/admin"
export NLS_LANG="AMERICAN_AMERICA.UTF8"
export PATH="$PATH:$DYLD_LIBRARY_PATH"

That’s all of the set-up done, now you just need to install the gem:

gem install ruby-oci8

Done.

In SQL…

1
2

ALTER session SET NLS_SORT=BINARY_CI;
ALTER session SET NLS_COMP=LINGUISTIC;

For use in Perl::DBI…

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18

use strict;
use DBI;
 
my $dbh = DBI->connect(
    'dbi:Oracle:schema_name',
    'user',
    'password',
    {
        RaiseError    => 1,
        AutoCommit    => 0,
        on_connect_do => [
            'alter session set NLS_SORT=BINARY_CI',
            'alter session set NLS_COMP=LINGUISTIC'
        ]
    }
) || die "Database connection not made: $DBI::errstr";
 
$dbh->disconnect();

About once a year I need to migrate a medium-sized web application using MySQL to Oracle.

Usually it’s to prepare a successful intranet application for a move to a formally supported production environment - and in Silicon Valley, that usually means Oracle.

I was apologizing to Monty at OSCON this year about my latest migration project away from MySQL.

His surprising response was, “It’s not a bad thing. I don’t mind hearing about conversions because successful migrations demonstrate that people don’t have to worry about database lock-in.”

Here’s some notes on doing migrations.

read more

My first thought was to do this in Perl (as the rest of the system is in Perl), but this would mean adding extra methods and calls in the Perl code to update the database (both the original tables and the new log tables). That seemed like a solution - a pain in the arse to implement, but a solution.

Thankfully one of the helpful chaps in my department suggested doing it all in the database with triggers as this is quite common in banks and the like. What a damn fine idea! Only a little SQL to write and no extra Perl.

Here’s an example…

First, here’s the table that we want to create an audit log for:

1
2
3
4
5
6
7

CREATE TABLE "COMMENT" (
  "COMMENT_ID" NUMBER(10,0) NOT NULL, 
  "COMMENT_BODY" VARCHAR2(1000), 
  "USER" VARCHAR2(100), 
  "CREATED_DATE" DATE, 
  CONSTRAINT "PK_COMMENT" PRIMARY KEY ("COMMENT_ID")
);

Now, we create the audit table.¹

1
2
3
4
5
6
7

CREATE TABLE "COMMENT_AUDIT" (
  "COMMENT_ID" NUMBER(10,0) NOT NULL, 
  "COMMENT_BODY" VARCHAR2(1000), 
  "USER" VARCHAR2(100), 
  "CREATED_DATE" DATE,
  "AUDIT_DATE" DATE DEFAULT SYSTIMESTAMP NOT NULL
);

Finally, we create the trigger to keep a log of changes to our original table.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

CREATE OR REPLACE TRIGGER "TA_COMMENT" 
before UPDATE OR DELETE ON COMMENT FOR each row
begin
IF deleting OR updating then 
  INSERT INTO comment_audit(
    COMMENT_ID,
    COMMENT_BODY,
    USER,
    CREATED_DATE
  )
  VALUES (
    :old.COMMENT_ID,
    :old.COMMENT_BODY,
    :old.USER,
    :old.CREATED_DATE
  );
end IF;
end;
/
ALTER TRIGGER "TA_COMMENT" ENABLE;

Now whenever a delete or update action is performed on the “COMMENT” table, a record of the old values and the date/time at which they were changed is recorded in the audit table.

First we need to create a sequence for the ID’s:

1
2
3
4

CREATE SEQUENCE "S_COMMENT_ID"
START WITH 1
INCREMENT BY 1
CACHE 10;

Then we create the table and the required trigger:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22

CREATE TABLE "COMMENT" (
  "COMMENT_ID" NUMBER(10,0) NOT NULL, 
  "COMMENT_BODY" VARCHAR2(1000), 
  "USER" VARCHAR2(100), 
  "CREATED_DATE" DATE, 
  CONSTRAINT "PK_COMMENT" PRIMARY KEY ("COMMENT_ID")
);
 
CREATE OR REPLACE TRIGGER "TR_COMMENT_ID" 
BEFORE INSERT 
ON COMMENT 
REFERENCING NEW AS NEW OLD AS OLD 
FOR EACH ROW 
BEGIN 
  IF(:new.comment_id IS NULL) then 
  SELECT S_COMMENT_ID.NEXTVAL 
  INTO :new.COMMENT_ID 
  FROM dual; 
  end IF; 
END;
/
ALTER TRIGGER "TR_COMMENT_ID" ENABLE;

And there you go - an auto-incrementing primary key… Damn that’s a lot more code to write than a simple ‘auto-increment’.

Edit: An alternative when using SQL Developer is to create your table, select it, then go to the ‘Triggers’ tab. In there you will find a useful ‘Actions…’ button where you can create a new trigger linking your tables primary key to a sequence.