software development – Page 5

sqlunit & dbUnit

Now that I have my database schema creation and maintenance automated with Incanto and dbMaintain I have been looking at testing tools for stored procedures. Perhaps the best known for Oracle PL/SQL are Feuerstein’s utPLSQL and subsequent Quest Code Tester for Oracle. Other options include plunit, PLUTO and ruby-pl-spec. My main requirement, however, has been to invoke tests from an Ant project continuously integrated using Hudson and for now I have started using sqlunit, which is both mature and portable across different databases, as well as allowing relatively simple testcases to be specified in entirely XML.

My build-common.xml now contains


 

 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 classname="net.sourceforge.sqlunit.ant.SqlunitTask"

 classpathref="sqlunit-lib"/>

In addition to sqlunit I’ve found
dbUnit to be a useful tool for complex data setup prior to running a test.
Initilization of the dbUnit ant task is done as follows:


 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 

 classname="org.dbunit.ant.DbUnitTask"

 classpathref="dbunit-lib"/>

Hudson

This week I’ve been upgrading from CruiseControl to Hudson. CruiseControl has served well for the past eighteen months, completing hundreds of builds with only the occasional restart. Once configured with a couple of XML files CruiseControl is low maintenance, but my impression is that now Thoughtworks is offering a commercial product it is no longer getting as much attention as its competitors. Hudson in particular is an open source CI solution that has received good reviews, and with the need to configure new Maven projects I thought this a good time to switch.

The best way to use Hudson seems to be with a dedicated VM including any infrastructure required to test the project being built, in my case for example Java 6, Maven2 and a dedicated instance of Oracle XE. Although it is possible to simply start Hudson from the command line it makes more sense to install it as a service, and this is exactly what the Debian/Ubuntu package does. I use Hudson with a Nexus maven repository, with settings.xml stored in …/tools/_usr_share_maven2/conf/, so far it seems to be working well.

dbMaintain

Since reading about database refactoring some while ago I’ve been seeking to adopt a tool to support automated tracking and application of database changes. The best known solution for this is LiquiBase (fka Sundog Refactoring Tool), with over 10,000 downloads of LiquiBase Core recorded on SourceForge as I write this. In order to adopt LiquiBase on an existing project, however, I was faced with replacing dozens of SQL DDL scripts with LiquiBase specific XML files. There is a generateChangeLog command, but it has limitations: for example it does not recognize the primary key of Oracle index organized tables, and the output is not arranged anything like the documented best practices.

Instead, for now, I have adopted a newer tool called dbMaintain, which was presented at a Belgian Java User Group last April (see below) and a ServerSide news article last July. dbMaintain provides similar functionality to LiquiBase, but instead of XML it simply passes through native SQL scripts using JDBC, making it much easier to adopt with confidence on an existing project. I did run into one known issue with setting the default schema, but I’m happy to say Tim Ducheyne responded to my emails and provided a fixed snapshot within days. I hope this project gains momentum, and I’m looking forward to the next release.

Unit testing: unitils & dbmaintain

Technorati Tags: Database development

PS. Other less mature alternatives include SQL based Thoughtworks dbdeploy and Carbon Five’s c5-db-migration. Another custom XML based solution is MIGRATEdb. Sony Pictures Imageworks have ported ActiveRecord migrations to Scala and there is a relatively inactive port of ActiveRecord migrations to Groovy called Bering.

Oracle SQLDeveloper 2.1 vs Quest TOAD

This past week I’ve been seeing whether I can wean myself off TOAD to the new SQLDeveloper 2.1 that was officially released last month. SQLDeveloper is a Java based IDE and so, unlike TOAD, it can be used on OS X and Ubuntu. SQLDeveloper is not open source, but it is licenced for free, whereas TOAD costs almost a thousand dollars per developer for the entry level version.

Version 2.1 of SQLDeveloper shows considerable improvement compared to the original release, but there are still some rough edges. TOAD presents database connections (connection configuration) and sessions (active connections) clearly and separately, with schema browsers and worksheets for each session in separate windows, and highlighting of windows that share the same session. SQLDeveloper has a single connection navigator, perhaps more familiar to users of other IDEs, but each connection may have only one session: you have to duplicate and rename a connection if you want to establish another session using the same connection parameters. Working with TOAD there is nearly always a default current session context, so it was annoying with SQLDeveloper reports to be prompted repeatedly for the connection to use to run a report.

One of the features of SQLDeveloper that first interested me was Subversion support. Again, this is similar to other Java IDEs. I’ve used Team Coding with TOAD in the past but the integration with CVS or PVCS was not as seamless as one would like. Now I just use TortoiseSVN and make sure I open the local source file whenever I want to edit a stored procedure and never edit the stored procedure source in the database directly. Oracle provides a tutorial on Subversion with SQLDeveloper but in my opinion it completely misses the point by failing to explain how to maintain stored procedures.

In general TOAD still appears a little more polished: for example favorites support in the File Open dialog, and a popup calendar control for setting date bind variables when executing a query. TOAD provides better feedback: for example it automatically displays the row count in a status line for every worksheet query, but with SQLDeveloper you have to select a popup menu option. For long running procedure execs TOAD provides a script runner with animated graphics and a timer, but the default behavior for SQLDeveloper is to exec the procedure like sqlplus in a small log window that is not immediately obvious. On the other hand the dockable tabs of SQLDeveloper are easier to manage than the more dated MDI design of TOAD.

Perhaps the biggest surprise, however, was when I clicked Tools/Monitor SQL and got ‘The use of this feature requires that the Oracle Tuning Pack be licensed.’ Oracle is not giving everything away for free after all, and it looks like for more advanced features I may still be going back to TOAD for a little while.

Incanto

After months of delay I’ve finally switched all my database creation scripts to Incanto. The most obvious benefit is that my Ant build scripts are now more portable, as I am no longer exec’ing a shell script wrapper to sqlplus, and I no longer need Cygwin to run the build scripts on my Windows workstation. One of the nice features about Incanto is that it allows you to pass in an Ant propertyset which is used to define SQL*Plus variables for the invoked scripts. I find this useful for overriding schema names and allowing multiple sets of application schemas to be created in the same developer database, for example when researching support issues alongside new development. The Incanto website has a page of best practices including a <macrodef> which I placed in a shared build file and <import>ed into my various Ant scripts.

Incanto requires sqlplus to be on the PATH, so I handle that and setting login parameters in .antrc. One strange issue I encountered on SLES while making this transition was that even although sqlplus was on my PATH and executable from my shell scripts, when I used Incanto I got

java.io.IOException: java.io.IOException: sqlplus: not found

Changing the permissions on sqlplus from 750 to 755 resolved the issue, even though the executable already belonged to a secondary group of which I was a member. Note that if you don’t need the advanced features of sqlplus you can still execute PL/SQL with Ant’s sql task.

Subversion on Ubuntu – Multiple Repositories

Eighteen months have passed since I setup Subversion on Ubuntu and I’ve found multiple repositories are valuable for providing finer grained access control and easier storage management. Here is a revised incantation:

su -
adduser --system --home /srv/svnrepos --gecos "System account to run svnserve" svn
apt-get install subversion xinetd
cat >> /etc/xinetd.d/svn << "EOF"
service svn
{
        port                    = 3690
        socket_type             = stream
        protocol                = tcp
        wait                    = no
        user                    = svn
        server                  = /usr/bin/svnserve
        server_args             = -i -r /srv/svnrepos
}
EOF
/etc/init.d/xinetd restart

# for each repository, eg. myrepo1
$repo=myrepo1
svnadmin create /srv/svnrepos/$repo
chown -R svn:nogroup /srv/svnrepos

# uncomment line to use default password file
vi ~svn/$repo/conf/svnserve.conf

cat >> ~svn/$repo/conf/passwd << "EOF"
[users]
fred = *****
...

Google Trends – Databases & Languages

Google Trends is a great tool for graphing search term popularity over time. Here is a comparison of searches for several makes of database.

Perhaps it’s a quirk of the phrases I used, but I was surprised to see Oracle so dominant, and even MySQL more popular than SQL Server. PostgreSQL languishes in obscurity compared to MySQL, and Ingres barely registers at all.

Here’s some languages that I’m interested in. Java is hugely dominant.

Let’s take out Java and zoom in. I was surprised to see Ruby’s recent decline relative to Python. Books on Scala have come out only recently.

Looking more specifically at JVM language dialects I’m surprised to see Jython competing so strongly with JRuby. I’m not sure if I picked the right phrase for Groovy.

Choosing tools solely on the basis of popularity is obviously not a great idea, but I do feel better now about continuing to use Oracle, Java and Python.

P.S. See also the TIOBE Programming Community Index

cx_Oracle callfunc()

I like wrapping database queries in stored procedures: it keeps the SQL in the database, where Oracle can track its dependencies and flag if some change makes it invalid. Here’s an example using Oracle’s demo schema SCOTT:

CREATE OR REPLACE PACKAGE emp_extract_k
AS
   TYPE emp_t
   IS
      RECORD( ename  varchar2( 10 ), hiredate date, sal     number( 7, 2 ) );
   TYPE emp_csr_t IS REF CURSOR
      RETURN emp_t;
   FUNCTION get_emps
      RETURN emp_csr_t;
END;
/
CREATE OR REPLACE PACKAGE BODY emp_extract_k
AS
   FUNCTION get_emps
      RETURN emp_csr_t
   IS
      c   emp_csr_t;
   BEGIN
      OPEN c FOR SELECT   ename, hiredate, sal FROM scott.emp;
      RETURN c;
   END;
END;/

Now suppose I want to take the output from a stored procedure like the one above and write it to a CSV file. In Python this is trivially easy:

 import cx_Oracle
 import csv
 reportName = 'emp_extract_k.get_emps'
 connection = cx_Oracle.connect(userid)
 cursor = connection.cursor()
 rs = cursor.callfunc(reportName, cx_Oracle.CURSOR)
 fd = open(reportName+'.csv', 'w')
 csvFile = csv.writer(fd)
 header = []
 [header.append(col_defn[0]) for col_defn in rs.description]
 csvFile.writerow(header)
 for r in rs:
  csvFile.writerow(r)
 fd.close()

VBScript will never die, just fade away

I was looking for an official announcement of the end of life of VBScript but could not find one. Instead I stumbled across the following MSDN blog post with a vague reference: "I believe that it was [at the] IT Forum in Nov 2005 that this was announced." A colorful blog post from Ryan Stemkoski made me laugh, the comments are interesting also. The closest I could find to an official Microsoft announcement was a blog post five years ago by Eric Lippert, who admitted to being the last person to ever add a feature to VBScript, in November 2000.

Adding a surrogate key to an Oracle table

This works for me for a mid-sized table.

ALTER TABLE foo ADD ( foo_id integer )
/
CREATE SEQUENCE foo_id_seq
/
CREATE OR REPLACE TRIGGER before_update_foo
   BEFORE INSERT OR UPDATE
   ON FOO
   REFERENCING OLD AS old NEW AS new
   FOR EACH ROW
DECLARE
   l_foo_id   foo.foo_id%TYPE;
BEGIN
   IF :new.foo_id IS NULL
   THEN
      SELECT   foo_id_seq.nextval INTO l_foo_id FROM dual;

      :new.foo_id := l_foo_id;
   END IF;
END;
/
UPDATE   foo
   SET   foo_id = NULL
/
COMMIT
/
CREATE UNIQUE INDEX foo_ak
   ON foo( foo_id )
/
ALTER TABLE foo MODIFY ( foo_id NOT NULL )
/
ALTER TABLE foo ADD (CONSTRAINT foo_ak UNIQUE (foo_id))
/

1 2 3 4 5 6

Posts navigation