Net::XMPP::Namespaces - In depth discussion on how namespaces are handled
Net::XMPP::Namespaces provides an depth look at how Net::XMPP handles namespacs, and how to add your own custom ones. It also serves as the storage bin for all of the Namespace information Net::XMPP requires.
XMPP as a protocol is very well defined. There are three main top
level packets (message, iq, and presence). There is also a way to
extend the protocol in a very clear and strucutred way, via namespaces.
Two major ways that namespaces are used in Jabber is for making the
<iq/> a generic wrapper, and as a way for adding data to any packet via
a child tag <x/>. We will use <x/> to represent the packet, but in
reality it could be any child tag: <foo/>, <data/>, <error/>, etc.
The Info/Query <iq/> packet uses namespaces to determine the type of
information to access. Usually there is a <query/> tag in the <iq/>
that represents the namespace, but in fact it can be any tag. The
definition of the Query portion, is the first tag that has a namespace.
<iq type="get"><query xmlns="..."/></iq>
or
<iq type="get"><foo xmlns="..."/></iq>
After that Query stanza can be any number of other stanzas (<x/> tags)
you want to include. The Query packet is represented and available by
calling GetQuery() or GetChild(), and the other namespaces are
available by calling GetChild().
The X tag is just a way to piggy back data on other packets. Like
embedding the timestamp for a message using jabber:x:delay, or signing
you presence for encryption using jabber:x:signed.
To this end, Net::XMPP has sought to find a way to easily, and clearly
define the functions needed to access the XML for a namespace. We will
go over the full docs, and then show two examples of real namespaces so
that you can see what we are talking about.
To avoid a lot of nasty modules populating memory that are not used,
and to avoid having to change 15 modules when a minor change is
introduced, the Net::XMPP modules have taken AUTOLOADing to the
extreme. Namespaces.pm is nothing but a set of function calls that
generates a big hash of hashes. The hash is accessed by the Stanza.pm
AUTOLOAD function to do something. (This will make sense, I promise.)
Before going on, I highly suggest you read a Perl book on AUTOLOAD and
how it works. From this point on I will assume that you understand it.
When you create a Net::XMPP::IQ object and add a Query to it (NewChild)
several things are happening in the background. The argument to
NewChild is the namespace you want to add. (custom-namespace)
Now that you have a Query object to work with you will call the GetXXX
functions, and SetXXX functions to set the data. There are no defined
GetXXX and SetXXXX functions. You cannot look in the Namespaces.pm
file and find them. Instead you will find something like this:
&add_ns(ns => "mynamespace",
tag => "mytag",
xpath => {
JID => { type=>'jid', path => '@jid' },
Username => { path => 'username/text()' },
Test => { type => 'master' }
}
);
When the GetUsername() function is called, the AUTOLOAD function looks
in the Namespaces.pm hash for a "Username" key. Based on the "type" of
the field (scalar being the default) it will use the "path" as an XPath
to retrieve the data and call the XPathGet() method in Stanza.pm.
Confused yet?
Now this is where this starts to get a little sticky. When you see a namespace with __netxmpp__, or __netjabber__ from Net::Jabber, at the beginning it is usually something custom to Net::XMPP and NOT part of the actual XMPP protocol. There are some places where the structure of the XML allows for multiple children with the same name. The main places you will see this behavior is where you have multiple tags with the same name and those have children under them (jabber:iq:roster). In jabber:iq:roster, the <item/> tag can be repeated multiple times, and is sort of like a mini-namespace in itself. To that end, we treat it like a seperate namespace and defined a __netxmpp__:iq:roster:item namespace to hold it. What happens is this, in my code I define that the <item/>s tag is "item" and anything with that tag name is to create a new Net::XMPP::Stanza object with the namespace __netxmpp__:iq:roster:item which then becomes a child of the jabber:iq:roster Stanza object. Also, when you want to add a new item to a jabber:iq:roster project you call NewQuery with the private namespace. I know this sounds complicated. And if after reading this entire document it is still complicated, email me, ask questions, and I will monitor it and adjust these docs to answer the questions that people ask.
To repeat, here is an example call to add_ns():
&add_ns(ns => "mynamespace",
tag => "mytag",
xpath => {
JID => { type=>'jid', path => '@jid' },
Username => { path => 'username/text()' },
Test => { type => 'master' }
}
);
ns - This is the new namespace that you are trying to add.
tag - This is the root tag to use for objects based on this namespace.
xpath - The hash reference passed in the add_ns call to each name of
entry tells Net::XMPP how to handle subsequent GetXXXX(), SetXXXX(),
DefinedXXXX(), RemoveXXXX(), AddXXXX() calls. The basic options you
can pass in are:
type - This tells Stanza how to handle the call. The possible
values are:
array - The value to set and returned is an an array
reference. For example, <group/> in jabber:iq:roster.
child - This tells Stanza that it needs to look for the
__netxmpp__ style namesapced children. AddXXX() adds
a new child, and GetXXX() will return a new Stanza
object representing the packet.
flag - This is for child elements that are tags by themselves:
<foo/>. Since the presence of the tag is what is
important, and there is no cdata to store, we just call
it a flag.
jid - The value is a Jabber ID. GetXXX() will return a
Net::XMPP::JID object unless you pass it "jid", then it
returns a string.
master - The GetXXX() and SetXXX() calls return and take a
hash representing all of the GetXXX() and SetXXX()
calls. For example:
SetTest(foo=>"bar",
bar=>"baz");
Translates into:
SetFoo("bar");
SetBar("baz");
GetTest() would return a hash containing what the
packet contains:
{ foo=>"bar", bar=>"baz" }
raw - This will stick whatever raw XML you specify directly
into the Stanza at the point where the path specifies.
scalar - This will set and get a scalar value. This is the
main workhorse as attributes and CDATA is represented
by a scalar. This is the default setting if you do
not provide one.
special - The special type is unique in that instead of a
string "special", you actually give it an array:
[ "special" , <subtype> ]
This allows Net::XMPP to be able to handle the
SetXXXX() call in a special manner according to your
choosing. Right now this is mainly used by
jabber:iq:time to automatically set the time info in
the correct format, and jabber:iq:version to set the
machine OS and add the Net::Jabber version to the
return packet. You will likely NOT need to use
this, but I wanted to mention it.
timestamp - If you call SetXXX() but do not pass it anything,
or pass it "", then Net::XMPP will place a
timestamp in the xpath location.
path - This is the XPath path to where the bit data lives. The
difference. Now, this is not full XPath due to the nature
of how it gets used. Instead of providing a rooted path
all the way to the top, it's a relative path ignoring what
the parent is. For example, if the "tag" you specified was
"foo", and the path is "bar/text()", then the XPath will be
rooted in the XML of the <foo/> packet. It will set and get
the CDATA from:
<foo><bar>xxxxx</bar></foo>
For a flag and a child type, just specify the child element.
Take a look at the code in this file for more help on what
this means. Also, read up on XPath if you don't already know
what it is.
child - This is a hash reference that tells Net::XMPP how to handle
adding and getting child objects. The keys for the hash are
as follows:
ns - the real or custom (__netxmpp__) namesapce to use for
this child packet.
skip_xmlns => 1 - this tells Net::XMPP not to add an
xmlns='' into the XML for the child
object.
specify_name => 1 - allows you to call NewChild("ns","tag")
and specify the tag to use for the child
object. This, IMHO, is BAD XML
practice. You should always know what
the tag of the child is and use an
attribute or CDATA to change the type
of the stanza. You do not want to use
this.
tag - If you use specify_name, then this is the default tag
to use. You do not want to use this.
calls - Array reference telling Net::XMPP what functions to create
for this name. For most of the types above you will get
Get, Set, Defined, and Remove. For child types you need to
decide how you API will look and specify them yourself:
["Get","Defined"]
["Add"]
["Get","Add","Defined"]
It all depends on how you want your API to look.
Once more... The following:
&add_ns(ns => "mynamespace",
tag => "mytag",
xpath => {
JID => { type=>'jid', path => '@jid' },
Username => { path => 'username/text()' },
Test => { type => 'master' }
}
);
generates the following API calls:
GetJID()
SetJID()
DefinedJID()
RemoveJID()
GetUsername()
SetUsername()
DefinedUsername()
RemoveUsername()
GetTest()
SetTest()
Well. I hope that I have not scared you off from writing a custom namespace for you application and use Net::XMPP. Look in the Net::XMPP::Protocol manpage for an example on using the add_ns() function to register your custom namespace so that Net::XMPP can properly handle it.
Ryan Eatmon
This module is free software, you can redistribute it and/or modify it under the LGPL.