New IRC channel

There’s a new IRC channel for romanian Perl developers called #perl-ro … so come visit us in #perl-ro on Freenode and stay for a chat

Simple notifications with DBIx::Simple and Email::Simple

DBIx::Simple and Email::Simple are just what it says on the package: a simpler way of using DBI and handling email.


If you’re not familiar with it, start with this nice overview, summarized as follows:

  1. Connect using either an already connected DBI handle or an array of DSN, user, pass and options.
  2. Execute your query from:
    • a literal SQL string, a.k.a. write your own, with query, (a fall-back for overly complex SQL queries)
    • an SQL string interpolated with Perl variables, with SQL::Interp
    • a complex Perl data structure, describing an SQL query, with SQL::Abstract (my personal favorite)
  3. Fetch your data as a flat list (flat), simple hash (map), array of arrays (arrays), array of hashes (hashes), number of affected rows (rows), etc.
  4. PROFIT!


All this module does is email parsing and formatting (if you want fancy emails, there is also Email::MIME). To actually send the email you will also need Email::Sender.

Create your email by passing a hash to the create constructor. It’s keys are header, which is an array of headers (header name, followed by header value, etc.) and body which is the email text. The resulting object can be passed to the sendmail function exported by the Email::Sender::Simple module.

Notify Me, Please

A real-life example: daily notifications for users hitting soft quota disk limits. First, we setup a table (in a MySQL database) to track usage with the following columns: username (username), actual usage (usage), the quota limit (limit), email address (email) and a timestamp (last_notification). The primary key is username. The usage column is updated hourly by a script running from cron. After that script finishes, our notification script is ready to rock’n'roll!

First things, first, our simple modules:

use DBIx::Simple;
use Email::Simple;
use Email::Sender::Simple 'sendmail';

A List of Users

After that we lookup which users should get notification emails. They are the users whose usage is over the quota limit and whose last notification timestamp is older than 24 hours:

my $dbase = DBIx::Simple->connect( $dsn, $user, $pass );
my $users = $dbase->select( 'quotas', [qw/usage quota username email/], {
    usage => \">= limit",
    last_notification => \"< DATE_SUB( NOW(), INTERVAL '24 5' HOUR_MINUTE)",

So, what's going on here line by line?

Line 1: we're just connecting to the database. We skipped the error checking code for clarity. The error can be accessed with DBIx::Simple->error.

Line 2: we're telling DBIx::Simple we want to use SQL::Abstract and do a SELECT query by using the select method. The first argument is the table name: quotas, followed by a list of columns names (you can also use a literal string if you want to apply DB functions to columns), followed by a hash which defines the WHERE clause.

Line 3: this is the first WHERE condition. Notice the reference to a scalar, it tells SQL::Abstract that the string is not a value which should be escaped and quoted (with the use of ? placeholders) but literal SQL.

Line 4: the second WHERE condition. Again we use a scalar reference so we can use the DATE_SUB function. The condition subtracts 24 hours and 5 minutes from the current date and time and compares it with the timestamp in the last_notification column. The 5 minutes are there to account for the time it takes to update quota usage and send emails.

The resulting SQL query:

SELECT usage, quota, username, email FROM quotas WHERE usage >= limit AND last_notification < DATE_SUB( NOW(), INTERVAL '24 5' HOUR_MINUTE)

Some Emails

OK, now let's send compose and send those emails:

foreach my $user ($users->hashes) {
    my $email = Email::Simple->create(
        header => [
            From => '',
            To => $user->{email},
            Subject => 'Quota exceeded',
        body => "Dear $user->{username},\nYour current usage of $user->{usage} exceeds your quota of $user->{quota}.",
    sendmail( $email );

We just loop over all users who match the criteria, compose a personalized email for each one and send it on it's way. Note that the sendmail function could fail and throw an exception. You can use the nice Try::Tiny module to catch that exception and either abort all notifications or ignore it or even do the Macarena (if you're into that kind of exception handling techniques).

Final Bookkeeping

The last step (remember we're still in the user loop) is to update the last_notification column:

    $dbase->update( 'quotas',
        { last_notification => \"NOW()", },
        { username => $user->{username} },

Line by line:

Line 1: we tell DBIx::Simple again that we will use SQL::Abstract and do an UPDATE query on the quotas table.

Line 2: the first hash describes what will be updated: hash keys are column names and hash values will be the record values. We use the MySQL function: NOW() to set the timestamp to the current date and time, therefore we have to use a reference to a scalar again. Otherwise, the column will be set to the string "NOW()".

Line 3: the next hash describes a WHERE clause. Since the username is the primary key of the table, it is enough to identify the record we want to update.

This will make sure the no further notification emails will be sent for another 24 hours.

That's all folks, pretty simple, ain't it?

Modern Perl Book – draft

If you’re programming in Perl it’s a must read!

XML::LibXML – building XML: intro

XML::Simple is good enough for simple tasks (i.e. config files), but if you want to get down to business you might prefer XML::LibXML

The distribution is both a XML parser (and validator) and a XML generator. However the roles are not clearly separated and the methods are not orthogonal. The docs are OK, but unwieldy – there are many classes, and the methods are listed in the synopsis instead of the TOC. This also means you have to grep / search in the browser to get to a method’s description. The method you’re looking for might be hidden in one of the many classes, between other methods that are somehow unrelated to your purpose. In this series I’ll try to point out and illustrate the methods for the most common tasks.

What’s what

In the XML::LibXML namespace we have:

  • ::Document – this will hold everything, help you to set XML version, encoding and compression, create nodes, set the root node and last, but not least validate and serialize your document into a string
  • ::Node – base class for XML nodes, you never create a node directly (you create elements mostly). It has, however many methods which you can use
  • ::Element – this class represents an XML element, inherits the methods from ::Node and is your workhorse for XML building

First Example

use XML::LibXML;

$version = '1.0'; # is there any other?
$encoding = 'UTF-8'; # cause it's the standard, that's why!
$doc = XML::LibXML::Document->new( $version, $encoding ); # my XML DOM

# Element creation
$x = $doc->createElement( 'x' ); # <x/> element
$y = $doc->createElement( 'y' ); # <y/> element

# Attributes
$x->setAttribute( 'lang', 'en' ); # <x lang="en" />
$y->setAttribute( 'baz', 1 ); # <y baz="1" />

# XML structure and text nodes
$x->appendChild( $y ); # <x><y/></x>
$y->appendTextNode( 'foo' ); # <y>foo<y/>
$x->appendTextChild( 'name', 'bar' ); # <x><name>bar</name></x>

$doc->setDocumentElement( $x ); # make <x/> the root element

print $doc->toString( 1 ); # serialize it

This is the result:


What happened?

The first step is to create a DOM document class (line 5). I like to think of it as a context object for all things XML. Then I create two nodes (lines 8, 9). You can create them using the DOM object or directly from their class (XML::LibXML::Element->new).

After that I add an attribute to each node (lines 12, 13). This is a short-cut method, it creates an attribute node with the given name and content and appends it to the node. You can also do this step by step, using the XML::LibXML::Attr class.

Then I arrange the nodes so that <y/> is the child node of <x/> (line 16). A text node is created and appended to <y/> with another short-cut method (line 17). I use another short-cut to create two nodes: a child element node (<name/>) and a text node as its child, and append them to <x/>.

The last step in building my XML is to declare <x/> as the document (or root) node, i.e. the top-most node in the document.

Then I serialize and print the whole thing out. The parameter passed to toString detemines how compact the serialization is:

  • 0 – no line breaks or indentation
  • 1 – line breaks and indentation for nested element nodes
  • 2 – line breaks for element and text nodes and indentation for nested element nodes (it’s like 1, but text nodes are printed on separate lines from element nodes)

One last note: the order of the steps does not matter. I could’ve added the attributes last and set the document node first thing, the end result would the same. Well, at least as long as you don’t lookup nodes before they are added. For example, you can’t refer to <x/> with the $doc->documentElement method before you add it as a document element.

Less than CPAN

Most of you are familiar with CPAN. It’s everything you wanted. It’s got recursive parser generators, ORMs, XML libraries, advanced OO systems, network servers and games.

You are probably also familiar with its command line client, which is also called cpan. It’s not nice. There’s a more advanced client, cpanp (cpan plus), but it’s not better either. Both are very nosy and pushy, they ask you all kind of intimate details, like where do you live, which mirror do like to gaze into, would you like french fries with your coffee and would you like extra flags for your compiler (no, I don’t and I guess I never will, thank you and would you please just install this module!). Then they make smalltalk to you:

Oooh, I’m fetching an index, look at me, now I’m compiling your module, and holy shit, here comes a dependency, look I’m compiling it too, ain’t I just great?.

But fear not, now there’s another alternative, cpanm (a.k.a. cpan minus). Minus stands for less talk, more work. More work means: zero configuration, automatic dependency resolution, local::lib support, and if you want it bad enough, even zero install.

Besides installing modules, it offers you a self upgrading command, installing just the dependencies without building the distribution or just downloading and unpacking the source.

All software sucks though, so while it works for 99% of the people, it still might fail on complex modules or apps. For me, it failed a Padre install, namely it choked on the Wx library from what was probably a wx toolkit version mismatch.

Still, in the spirit of Worse is Better, cpanm is my new CPAN client of choice.

Try it, and if you like it, go and vote for your favorite features.

Zero install CPAN client

Sometimes I can’t be bothered to install cpanm. Luckily there is such a thing as zero-install cpanm:

wget -O - | perl - My::Module

This uses wget to fetch the latest cpanm source (from the easy to remember URL redirect, or from the equivalent, spits it out to STDOUT, which in turned is piped to perl for execution (additional options and command line arguments must be placed after perl’s - argument).

If you don’t like wget’s progress bar and messages, just tell it to be --quiet:

wget --quiet -O - | perl - My::Module

Be warned however that this is fetching the whole source (which currently weighs in at approx. 55 KB). Thus it’s kinda wasteful of network bandwidth if you use it constantly, so consider installing it (either for local::lib or for your whole system).

Better than grep

Don’t you hate it when you just type away at your command line, and you think “Who the hell uses this class called Lights::On and Lights::Out ?”, and you just reflexively type:

project$ grep -rE "use Lights::O" lib

and then it hits you like a brick wall of text:

lib/Bed/ Lights::Out;
lib/Bed/.svn/text-base/ Lights::Out;
lib/ Lights::Out;
lib/Sinners/ Lights::On;
lib/Sinners/ Lights::On;
lib/Sinners/ Lights::On;
lib/Sinners/ Lights::On;
lib/Sinners/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/.svn/text-base/ Lights::On;
lib/Sinners/ Lights::On;
lib/Sinners/ Lights::On;
lib/ Lights::Out;
lib/.svn/text-base/ Lights::Out;
lib/.svn/text-base/ Lights::Out;

Ugly as sin, ain’t it? Stupid SOB, that grep, can’t he see these are subversion files you have absolutely no interest in? Well, as a matter of fact, he’s so stupid, he can’t.

That’s why you, smart Perl Monger that you are, fall back to ack!

Because ack is better than grep! Yep, it’s so smart it almost hurts! Just try it:

project$ ack "use Lights::O"

No -r (we know you want -r every time), it skips those pesky svn base files as well as backup files, binary files, even core dumps. And for your convenience it will print line numbers and highlights the search terms for you (except when you pipe or redirect the output). Because it’s written in Perl, it comes with all the goodies of Perl regexes (no more -E or puzzling over extended vs. regular regexes and which metacharacters to escape).

There’s one common catch though. Unlike grep, ack will by default only search known filetypes, so if you’ve got some weird files (let’s say .wango files), ack wouldn’t know about and they will be silently skipped.

But don’t despair:

  • you can make it display a list of file types it knows about:
    ack --help=types
  • you can force it to search unknown file types too (subversion funky files will still be skipped):
    ack -a "use Lights::O"
  • even better, you can make “search unknown files too” your default:
    echo "--all" >> ~/.ackrc

There is a standalone version, which requires only Perl. Yes, it runs on Windows too (pure Perl), just install Strawberry Perl, cpan App::Ack and you’re good to go (provided you’ve got all your paths set up)

And it comes with a cat: ack --thpppt

Monthly meeting

Ce ziceţi ca odată pe lună (de ex. în prima vineri din fiecare lună) să ne întâlnim undeva în oraş?

Propuneri pentru dată resp. locaţie?

How to read a file in a string in one move. (Chuck Norris style)

my $file = '/files/example.xml';
my $xml;
#use a local block to reset the input record separator.
  local $/=undef;
  open(XML, "$file") || die("Could not open $file");
  binmode XML;
  $xml = <XML>;
  close XML;
print $xml;

suus nomen est Padre

Padre logo

suus nomen est Padre. novus perl ide

Padre is a Perl IDE, an Integrated Development environment, or in other words a text editor that is simple to use for new Perl programmers but also supports large multi-lingual and multi-technology projects.

You can fetch it from here