RPC::XML(3) A set of classes for core data, message and XML handling

SYNOPSIS


use RPC::XML;
$req = RPC::XML::request->new('fetch_prime_factors',
RPC::XML::int->new(985_120_528));
...
$resp = RPC::XML::ParserFactory->new()->parse(STREAM);
if (ref($resp))
{
return $resp->value->value;
}
else
{
die $resp;
}

DESCRIPTION

The RPC::XML package is an implementation of the XML-RPC standard. The package as a whole provides classes for data, for clients, for servers and for parsers (based on the XML::Parser and XML::LibXML packages from CPAN).

This module provides a set of classes for creating values to pass to the constructors for requests and responses. These are lightweight objects, most of which are implemented as blessed scalar references so as to associate specific type information with the value. Classes are also provided for requests, responses and faults (errors).

This module does not actually provide any transport implementation or server basis. For these, see RPC::XML::Client and RPC::XML::Server, respectively.

SUBROUTINES/METHODS

At present, two subroutines are available for import. They must be explicitly imported as part of the "use" statement, or with a direct call to "import":
time2iso8601([$time])
Convert the integer time value in $time (which defaults to calling the built-in "time" if not present) to a (pseudo) ISO 8601 string in the UTC time zone. This is a convenience function for occassions when the return value needs to be of the dateTime.iso8601 type, but the value on hand is the return from the "time" built-in. Note that the format of this string is not strictly compliant with ISO 8601 due to the way the dateTime.iso8601 data-type was defined in the specification. See ``DATES AND TIMES'', below.
smart_encode(@args)
Converts the passed-in arguments to datatype objects. Any that are already encoded as such are passed through unchanged. The routine is called recursively on hash and array references. Note that this routine can only deduce a certain degree of detail about the values passed. Boolean values will be wrongly encoded as integers. Pretty much anything not specifically recognizable will get encoded as a string object. Thus, for types such as "fault", the ISO time value, base-64 data, etc., the program must still explicitly encode it. However, this routine will hopefully simplify things a little bit for a majority of the usage cases.

If an argument is a blessed reference (an object), smart_encode will generally treat it as a non-blessed reference of the underlying type. That is, objects based on hash references will be encoded as if they are unblessed hash references (becoming RPC::XML::struct objects), objects based on array references are encoded as array references (RPC::XML::array), etc. Only hash references, array references and scalar references are treated in this fashion; any other blessed references cannot be down-graded and will cause an exception to be thrown.

The exception to this are objects of the DateTime class: this package does not utilize DateTime directly, but if you pass in a reference to an existing object of that class, it is properly converted to an object of the RPC::XML::datetime_iso8601 class.

In addition to these, the following ``helper'' functions are also available. They may be imported explicitly, or all may be imported via the tag ":types":

    RPC_BOOLEAN RPC_INT RPC_I4 RPC_I8 RPC_DOUBLE
    RPC_DATETIME_ISO8601 RPC_BASE64 RPC_STRING RPC_NIL

Each creates a data object of the appropriate type from a single value (or, in the case of RPC_NIL, from no value). They are merely short- hand for calling the constructors of the data classes directly.

All of the above (helpers and the first two functions) may be imported via the tag ":all".

CLASSES

The classes provided by this module are broken into two groups: data classes and message classes.

Data Classes

The following data classes are provided by this library. Each of these provide at least the set of methods below. Note that these classes are designed to create throw-away objects. There is currently no mechanism for changing the value stored within one of these object after the constructor returns. It is assumed that a new object would be created, instead.

The common methods to all data classes are:

new($value)
Constructor. The value passed in is the value to be encapsulated in the new object.
value
Returns the value kept in the object. Processes recursively for "array" and "struct" objects.
as_string
Returns the value as a XML-RPC fragment, with the proper tags, etc.
serialize($filehandle)
Send the stringified rendition of the data to the given file handle. This allows messages with arbitrarily-large base-64 data within them to be sent without having to hold the entire message within process memory.
length
Returns the length, in bytes, of the object when serialized into XML. This is used by the client and server classes to calculate message length.
type
Returns the type of data being stored in an object. The type matches the XML-RPC specification, so the normalized form "datetime_iso8601" comes back as "dateTime.iso8601".
is_fault
All types except the fault class return false for this. This is to allow consistent testing of return values for fault status, without checking for a hash reference with specific keys defined.

The classes themselves are:

RPC::XML::int
Creates an integer value. Constructor expects the integer value as an argument.
RPC::XML::i4
This is like the "int" class. Note that services written in strictly-typed languages such as C, C++ or Java may consider the "i4" and "int" types as distinct and different.
RPC::XML::i8
This represents an 8-byte integer, and is not officially supported by the XML-RPC specification. This has been added to accommodate services already in use that have chosen to add this extension.
RPC::XML::double
Creates a floating-point value.
RPC::XML::string
Creates an arbitrary string. No special encoding is done to the string (aside from XML document encoding, covered later) with the exception of the "<", ">" and "&" characters, which are XML-escaped during object creation, and then reverted when the "value" method is called.
RPC::XML::boolean
Creates a boolean value. The value returned will always be either of 1 or 0, for true or false, respectively. When calling the constructor, the program may specify any of: 0, "no", "false", 1, "yes", "true".
RPC::XML::datetime_iso8601
Creates an instance of the XML-RPC "dateTime.iso8601" type. The specification for ISO 8601 may be found elsewhere. No processing is done to the data. Note that the XML-RPC specification actually got the format of an ISO 8601 date slightly wrong. Because this is what is in the published spec, this package produces dates that match the XML-RPC spec, not the the ISO 8601 spec. However, it will read date-strings in proper ISO 8601 format. See ``DATES AND TIMES'', below.
RPC::XML::nil
Creates a "nil" value. The value returned will always be undef. No value should be passed when calling the constructor.

Note that nil is an extension to XML-RPC, which is not supported by all implementations. $RPC::XML::ALLOW_NIL must be set to a non-false value before objects of this type can be constructed. See ``GLOBAL VARIABLES''. However, even if $RPC::XML::ALLOW_NIL is set to a false value, the parsers will recognize the "<nil />" tag and construct an object.

In practice, this type is only useful to denote the equivalent of a ``void'' return value from a function. The type itself is not interchangeable with any of the other data-types.

RPC::XML::base64
Creates an object that encapsulates a chunk of data that will be treated as base-64 for transport purposes. The value may be passed in as either a string or as a scalar reference. Additionally, a second (optional) parameter may be passed, that if true identifies the data as already base-64 encoded. If so, the data is decoded before storage. The "value" method returns decoded data, and the "as_string" method encodes it before stringification.

Alternately, the constructor may be given an open filehandle argument instead of direct data. When this is the case, the data is never read into memory in its entirety, unless the "value" or "as_string" methods are called. This allows the manipulation of arbitrarily-large Base-64-encoded data chunks. In these cases, the flag (optional second argument) is still relevant, but the data is not pre-decoded if it currently exists in an encoded form. It is only decoded as needed. Note that the filehandle passed must be open for reading, at least. It will not be written to, but it will be read from. The position within the file will be preserved between operations.

Because of this, this class supports a special method called "to_file", that takes one argument. The argument may be either an open, writable filehandle or a string. If it is a string, "to_file" will attempt to open it as a file and write the decoded data to it. If the argument is a an open filehandle, the data will be written to it without any pre- or post-adjustment of the handle position (nor will it be closed upon completion). This differs from the "serialize" method in that it always writes the decoded data (where the other always writes encoded data), and in that the XML opening and closing tags are not written. The return value of "to_file" is the size of the data written in bytes.

RPC::XML::array
Creates an array object. The constructor takes zero or more data-type instances as arguments, which are inserted into the array in the order specified. "value" returns an array reference of native Perl types. If a non-null value is passed as an argument to "value()", then the array reference will contain datatype objects (a shallow rather than deep copy).
RPC::XML::struct
Creates a struct object, the analogy of a hash table in Perl. The keys are ordinary strings, and the values must all be data-type objects. The "value" method returns a hash table reference, with native Perl types in the values. Key order is not preserved. Key strings are now encoded for special XML characters, so the use of such ("<", ">", etc.) should be transparent to the user. If a non-null value is passed as an argument to "value()", then the hash reference will contain the datatype objects rather than native Perl data (a shallow vs. deep copy, as with the array type above).

When creating RPC::XML::struct objects, there are two ways to pass the content in for the new object: Either an existing hash reference may be passed, or a series of key/value pairs may be passed. If a reference is passed, the existing data is copied (the reference is not re-blessed), with the values encoded into new objects as needed.

RPC::XML::fault
A fault object is a special case of the struct object that checks to ensure that there are two keys, "faultCode" and "faultString".

As a matter of convenience, since the contents of a RPC::XML::fault structure are specifically defined, the constructor may be called with exactly two arguments, the first of which will be taken as the code, and the second as the string. They will be converted to RPC::XML types automatically and stored by the pre-defined key names.

Also as a matter of convenience, the fault class provides the following accessor methods for directly retrieving the integer code and error string from a fault object:

code
string

Both names should be self-explanatory. The values returned are Perl values, not RPC::XML class instances.

Message Classes

The message classes are used both for constructing messages for outgoing communication as well as representing the parsed contents of a received message. Both implement the following methods:
new
This is the constructor method for the two message classes. The response class may have only a single value (as a response is currently limited to a single return value), and requests may have as many arguments as appropriate. In both cases, the arguments are passed to the exported "smart_encode" routine described earlier.
as_string
Returns the message object expressed as an XML document. The document will be lacking in linebreaks and indention, as it is not targeted for human reading.
serialize($filehandle)
Serialize the message to the given file-handle. This avoids creating the entire XML message within memory, which may be relevant if there is especially-large Base-64 data within the message.
length
Returns the total size of the message in bytes, used by the client and server classes to set the Content-Length header.

The two message-object classes are:

RPC::XML::request
This creates a request object. A request object expects the first argument to be the name of the remote routine being called, and all remaining arguments are the arguments to that routine. Request objects have the following methods (besides "new" and "as_string"):
name
The name of the remote routine that the request will call.
args
Returns a list reference with the arguments that will be passed. No arguments will result in a reference to an empty list.
RPC::XML::response
The response object is much like the request object in most ways. It may take only one argument, as that is all the specification allows for in a response. Responses have the following methods (in addition to "new" and "as_string"):
value
The value the response is returning. It will be a RPC::XML data-type.
is_fault
A boolean test whether or not the response is signalling a fault. This is the same as taking the "value" method return value and testing it, but is provided for clarity and simplicity.

DATES AND TIMES

The XML-RPC specification refers to the date/time values as ISO 8601, but unfortunately got the syntax slightly wrong in the examples. However, since this is the published specification it is necessary to produce time-stamps that conform to this format. The specification implies that the only format for date/time values is:

    YYYYMMDDThh:mm:ss

(Here, the "T" is literal, the rest represent elements of the date and time.) However, the ISO 8601 specification does not allow this particular format, and in generally is considerably more flexible than this. Yet there are implementations of the XML-RPC standard in other languages that rely on a strict interpretation of this format.

To accommodate this, the RPC::XML package only produces dateTime.iso8601 values in the format given in the spec, with the possible addition of timezone information if the string used to create a RPC::XML::datetime_iso8601 instance included a timezone offset. The string passed in to the constructor for that class must match:

    \d\d\d\d-?\d\d-?\d\dT?\d\d:\d\d:\d\d([.,]\d+)?(Z|[-+]\d\d:\d\d)?

This pattern is also used by smart_encode to distinguish a date/time string from a regular string. Note that the "T" is optional here, as it is in the ISO 8601 spec. The timezone is optional, and if it is not given then UTC is assumed. The XML-RPC specification says not to assume anything about the timezone in the absence of one, but the format of ISO 8601 declares that that absence of an explicit timezone dictates UTC.

If you have DateTime::Format::ISO8601 installed, then RPC::XML::datetime_iso8601 will fall back on it to try and parse any input strings that do not match the above pattern. If the string cannot be parsed by the DateTime::Format::ISO8601 module, then the constructor returns undef and $RPC::XML::ERROR is set.

DIAGNOSTICS

All constructors (in all data classes) return "undef" upon failure, with the error message available in the package-global variable $RPC::XML::ERROR.

GLOBAL VARIABLES

The following global variables may be changed to control certain behavior of the library. All variables listed below may be imported into the application namespace when you "use" RPC::XML:
$ENCODING
This variable controls the character-set encoding reported in outgoing XML messages. It defaults to "us-ascii", but may be set to any value recognized by XML parsers.
$FORCE_STRING_ENCODING
By default, "smart_encode" uses heuristics to determine what encoding is required for a data type. For example, 123 would be encoded as "int", where 3.14 would be encoded as "double". In some situations it may be handy to turn off all these heuristics, and force encoding of "string" on all data types encountered during encoding. Setting this flag to "true" will do just that.

Defaults to "false".

$ALLOW_NIL
By default, the XML-RPC "nil" extension is not supported. Set this to a non-false value to allow use of nil values. Data objects that are "nil" are represented as undef by Perl. See ``The nil Datatype''.

CAVEATS

This began as a reference implementation in which clarity of process and readability of the code took precedence over general efficiency. It is now being maintained as production code, but may still have parts that could be written more efficiently.

BUGS

Please report any bugs or feature requests to "bug-rpc-xml at rt.cpan.org", or through the web interface at <http://rt.cpan.org/NoAuth/ReportBug.html?Queue=RPC-XML>. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.

SUPPORT

LICENSE AND COPYRIGHT

This file and the code within are copyright (c) 2011 by Randy J. Ray.

Copying and distribution are permitted under the terms of the Artistic License 2.0 (<http://www.opensource.org/licenses/artistic-license-2.0.php>) or the GNU LGPL 2.1 (<http://www.opensource.org/licenses/lgpl-2.1.php>).

CREDITS

The XML-RPC standard is Copyright (c) 1998-2001, UserLand Software, Inc. See <http://www.xmlrpc.com> for more information about the XML-RPC specification.

AUTHOR

Randy J. Ray <[email protected]>