NAME
McBain - Framework for building portable, auto-validating and
self-documenting APIs
SYNOPSIS
package MyAPI;
use McBain; # imports strict and warnings for you
get '/multiply' => (
description => 'Multiplies two integers',
params => {
one => { required => 1, integer => 1 },
two => { required => 1, integer => 1 }
},
cb => sub {
my ($api, $params) = @_;
return $params->{one} * $params->{two};
}
);
post '/factorial' => (
description => 'Calculates the factorial of an integer',
params => {
num => { required => 1, integer => 1, min_value => 0 }
},
cb => sub {
my ($api, $params) = @_;
# note how this route both uses another
# route and calls itself recursively
if ($params->{num} <= 1) {
return 1;
} else {
return $api->forward('GET:/multiply', {
one => $params->{num},
two => $api->forward('POST:/factorial', { num => $params->{num} - 1 })
});
}
}
);
1;
DESCRIPTION
"McBain" is a framework for building powerful APIs and applications.
Writing an API with "McBain" provides the following benefits:
* Lightweight-ness
"McBain" is extremely lightweight, with minimal dependencies on
non-core modules; only two packages; and a succinct, minimal syntax
that is easy to remember. Your APIs and applications will require
less resources and perform better. Maybe.
* Portability
"McBain" APIs can be run/used in a variety of ways with absolutely
no changes of code. For example, they can be used directly from Perl
code (see McBain::Directly), as fully fledged RESTful PSGI web
services (see McBain::WithPSGI), as Gearman workers (see
McBain::WithGearmanXS), or as ZeroMQ workers (see
McBain::WithZeroMQ). Seriously, no change of code required. More
McBain runners are yet to come (plus search CPAN to see if more are
available), and you can easily create your own, god knows I don't
have the time or motivation or talent. Why should I do it for you
anyway?
* Auto-Validation
No more tedious input tests. "McBain" will handle input validation
for you. All you need to do is define the parameters you expect to
get with the simple and easy to remember syntax provided by
Brannigan. When your API is used, "McBain" will automatically
validate input. If validation fails, "McBain" will return
appropriate errors and tell the users of your API that they suck.
* Self-Documentation
"McBain" also eases the burden of having to document your APIs, so
that other people can actually use it (and you, two weeks later when
you're drunk and can't remember why you wrote the thing in the first
place). Using simple descriptions you give to your API's methods,
and the parameter definitions, "McBain" can automatically create a
manual document describing your API (see the mcbain2pod command line
utility).
* Modularity and Flexibility
APIs written with "McBain" are modular and flexible. You can make
them object oriented if you want, or not, "McBain" won't care, it's
unobtrusive like that. APIs are hierarchical, and every module in
the API can be used as a complete API all by itself, detached from
its siblings, so you can actually load only the parts of the API you
need. Why is this useful? I don't know, maybe it isn't, what do I
care? It happened by accident anyway.
* No More World Hunger
It'll do that too, just give it a chance.
FUNCTIONS
The following functions are exported:
provide( $method, $route, %opts )
Define a method and a route. $method is one of "GET", "POST", "PUT" or
"DELETE". $route is a string that starts with a forward slash, like a
path in a URI. %opts can hold the following keys (only "cb" is
required):
* description
A short description of the method and what it does.
* params
A hash-ref of parameters in the syntax of Brannigan (see
Brannigan::Validations for a complete references).
* cb
An anonymous subroutine (or a subroutine reference) to run when the
route is called. The method will receive the root topic class (or
object, if the topics are written in object oriented style), and a
hash-ref of parameters.
get( $route, %opts )
Shortcut for "provide( 'GET', $route, %opts )"
post( $route, %opts )
Shortcut for "provide( 'POST', $route, %opts )"
put( $route, %opts )
Shortcut for "provide( 'PUT', $route, %opts )"
del( $route, %opts )
Shortcut for "provide( 'DELETE', $route, %opts )"
pre_route( $cb->( $self, $meth_and_route, \%params ) )
post_route( $cb->( $self, $meth_and_route, \$ret ) )
Define a post_route method to run before/after every request to a route
in the defining topic. See "PRE-ROUTES AND POST-ROUTES" for details.
METHODS
The following methods will be available on importing classes/objects:
call( @args )
Calls the API, requesting the execution of a certain route. This is the
main way your API is used. The arguments it expects to receive and its
behavior are dependent on the McBain runner used. Refer to the docs of
the runner you wish to use for more information.
forward( $namespace, [ \%params ] )
For usage from within API methods; this simply calls a method of the the
API with the provided parameters (if any) and returns the result. With
"forward()", an API method can call other API methods or even itself
(for recursive operations).
$namespace is the method and route to execute, in the format
"<METHOD>:<ROUTE>", where "METHOD" is one of "GET", "POST", "PUT",
"DELETE", and "ROUTE" starts with a forward slash.
is_root( )
Returns a true value if the module is the root topic of the API. Mostly
used internally and in McBain runner modules.
MANUAL
ANATOMY OF AN API
Writing an API with "McBain" is easy. The syntax is short and easy to
remember, and the feature list is just what it needs to be - short and
sweet.
The main idea of a "McBain" API is this: a client requests the execution
of a method provided by the API, sending a hash of parameters. The API
then executes the method with the client's parameters, and produces a
response. Every runner module will enforce a different response format
(and even request format). When the API is used directly, for example,
whatever the API produces is returned as is. The PSGI and Gearman::XS
runners, however, are both JSON-in JSON-out interfaces.
A "McBain" API is built of one or more topics, in a hierarchical
structure. A topic is a class that provides methods that are
categorically similar. For example, an API might have a topic called
"math" that provides math-related methods such as add, multiply, divide,
etc.
Since topics are hierarchical, every API will have a root topic, which
may have zero or more child topics. The root topic is where your API
begins, and it's your decision how to utilize it. If your API is short
and simple, with methods that cannot be categorized into different
topics, then the entire API can live within the root topic itself, with
no child topics at all. If, however, you're building a larger API, then
the root topic might be empty, or it can provide general-purpose methods
that do not particularly fit in a specific topic, for example maybe a
status method that returns the status of the service, or an
authentication method.
The name of a topic is calculated from the name of the package itself.
The root topic is always called "/" (forward slash), and its child
topics are named like their package names, in lowercase, relative to the
root topic, with "/" as a separator instead of Perl's "::", and starting
with a slash. For example, lets look at the following API packages:
+------------------------+-------------------+------------------+
| Package Name | Topic Name | Description |
+========================+===================+==================+
| MyAPI | "/" | the root topic |
| MyAPI::Math | "/math" | a child topic |
| MyAPI::Math::Constants | "/math/constants" | a child-of-child |
| MyAPI::Strings | "/strings" | a child topic |
+------------------------+--------------------------------------+
You will notice that the naming of the topics is similar to paths in
HTTP URIs. This is by design, since I wrote "McBain" mostly for writing
web applications (with the PSGI runner), and the RESTful architecture
fits well with APIs whether they are HTTP-based or not.
CREATING TOPICS
To create a topic package, all you need to do is:
use McBain;
This will import "McBain" functions into the package, register the
package as a topic (possibly the root topic), and attempt to load all
child topics, if there are any. For convenience, "McBain" will also
import strict and warnings for you.
Notice that using "McBain" doesn't make your package an OO class. If you
want your API to be object oriented, you are free to form your classes
however you want, for example with Moo or Moose:
package MyAPI;
use McBain;
use Moo;
has 'some_attr' => ( is => 'ro' );
1;
CREATING ROUTES AND METHODS
The resemblance with HTTP continues as we delve further into methods
themselves. An API topic defines routes, and one or more methods that
can be executed on every route. Just like HTTP, these methods are "GET",
"POST", "PUT" and "DELETE".
Route names are like topic names. They begin with a slash, and every
topic *can* have a root route which is just called "/". Every method
defined on a route will have a complete name (or path, if you will), in
the format "<METHOD_NAME>:<TOPIC_NAME><ROUTE_NAME>". For example, let's
say we have a topic called "/math", and this topic has a route called
"/divide", with one "GET" method defined on this route. The complete
name (or path) of this method will be "GET:/math/divide".
By using this structure and semantics, it is easy to create CRUD
interfaces. Lets say your API has a topic called "/articles", that deals
with articles in your blog. Every article has an integer ID. The
"/articles" topic can have the following routes and methods:
+------------------------+--------------------------------------+
| Namespace | Description |
+========================+======================================+
| POST:/articles/ | Create a new article (root route /) |
| GET:/articles/(\d+) | Read an article |
| PUT:/articles/(\d+) | Update an article |
| DELETE:/articles/(\d+) | Delete an article |
+------------------------+--------------------------------------+
Methods are defined using the get(), post(), put() and del()
subroutines. The syntax is similar to Moose's antlers:
get '/multiply' => (
description => 'Multiplies two integers',
params => {
a => { required => 1, integer => 1 },
b => { required => 1, integer => 1 }
},
cb => sub {
my ($api, $params) = @_;
return $params->{a} * $params->{b};
}
);
Of the three keys above ("description", "params" and "cb"), only "cb" is
required. It takes the actual subroutine to execute when the method is
called. The subroutine will get two arguments: first, the root topic
(either its package name, or its object, if you're creating an object
oriented API), and a hash-ref of parameters provided to the method (if
any).
You can provide "McBain" with a short "description" of the method, so
that "McBain" can use it when documenting the API with mcbain2pod.
You can also tell "McBain" which parameters your method takes. The
"params" key will take a hash-ref of parameters, in the format defined
by Brannigan (see Brannigan::Validations for a complete references).
These will be both enforced and documented.
As you may have noticed in the "/articles" example, routes can be
defined using regular expressions. This is useful for creating proper
RESTful URLs:
# in topic /articles
get '/(\d+)' => (
description => 'Returns an article by its integer ID',
cb => sub {
my ($api, $params, $id) = @_;
return $api->db->get_article($id);
}
);
If the regular expression contains captures, and a call to the API
matches the regular expressions, the values captured will be passed to
the method, after the parameters hash-ref (even if the method does not
define parameters, in which case the parameters hash-ref will be empty -
this may change in the future).
It is worth understanding how "McBain" builds the regular expression. In
the above example, the topic is "/articles", and the route is "/(\d+)".
Internally, the generated regular expression will be
"^/articles/(\d+)$". Notice how the topic and route are concatenated,
and how the "^" and "$" metacharacters are added to the beginning and
end of the regex, respectively. This means it is impossible to create
partial regexes, which only pose problems in my experience.
OPTIONS REQUESTS
Every route defined by the API also automatically gets an "OPTIONS"
method, again just like HTTP. This method returns a list of HTTP-style
methods allowed on the route. The return format depends on the runner
module used. The direct runner will return a hash-ref with keys being
the HTTP methods, and values being hash-refs holding the "description"
and "params" definitions (if any).
For example, let's look at the following route:
get '/something' => (
description => 'Gets something',
cb => sub { }
);
put '/something' => (
description => 'Updates something',
params => { new_content => { required => 1 } },
cb => sub { }
);
Calling "OPTIONS:/something" will return:
{
GET => {
description => "Gets something"
},
PUT => {
description => "Updates something",
params => {
new_content => { required => 1 }
}
}
}
CALLING METHODS FROM WITHIN METHODS
Methods are allowed to call other methods (whether in the same route or
not), and even call themselves recursively. This can be accomplished
easily with the forward() method. For example:
get '/factorial => (
description => 'Calculates the factorial of a number',
params => {
num => { required => 1, integer => 1 }
},
cb => sub {
my ($api, $params) = @_;
if ($params->{num} <= 1) {
return 1;
} else {
return $api->forward('GET:/multiply', {
one => $params->{num},
two => $api->forward('GET:/factorial', { num => $params->{num} - 1 })
});
}
}
);
In the above example, notice how the "GET:/factorial" method calls both
"GET:/multiply" and itself.
EXCEPTIONS
"McBain" APIs handle errors in a graceful way, returning proper error
responses to callers. As always, the way errors are returned depends on
the runner module used. When used directly from Perl code, McBain will
confess (i.e. die) with a hash-ref consisting of two keys:
* "code" - An HTTP status code indicating the type of the error (for
example 404 if the route doesn't exist, 405 if the route exists but
the method is not allowed, 400 if parameters failed validation,
etc.).
* "error" - The text/description of the error.
Depending on the type of the error, more keys might be added to the
exception. For example, the parameters failed validation error will also
include a "rejects" key holding Brannigan's standard rejects hash,
describing which parameters failed validation.
When writing APIs, you are encouraged to return exceptions in this
format to ensure proper handling by "McBain". If "McBain" encounters an
exception that does not conform to this format, it will generate an
exception with "code" 500 (indicating "Internal Server Error"), and the
"error" key will hold the exception as is.
PRE-ROUTES AND POST-ROUTES
*New in v1.3.0*
Every topic in your API can define pre and post routes. The pre route is
called right before a route is executed, while the post route is called
immediately after.
You should note that the pre and post routes are called on every route
execution (when applicable), even when forwarding from one route to
another.
Pre and post routes are hierarchical. When a route is executed, "McBain"
will analyze the entire chain of topics leading up to that route, and
execute all pre and post routes on the way (if any, of course). So, for
example, if the route "/math/factorial" is to be executed, "McBain" will
look for pre and post routes and the root topic ("/"), the "/math"
topic, and the "/math/factorial" topic (if it exists). Whichever ones it
finds will be executed, in order.
The "pre_route" subroutine gets as parameters the API package (or
object, if writing object-oriented APIs, or the context object, if
writing in contextual mode), the full route name (the method and the
path, e.g. "POST:/math/factorial"), and the parameters hash-ref, after
validation has occurred.
package MyApi::Math;
post '/factorial' => (
...
);
pre_route {
my ($self, $meth_and_route, $params) = @_;
# do something here
}
The "post_route" subroutine gets the same parameters, except the
parameters hash-ref, in which place a reference to the result returned
by the actual route is passed. So, for example, if the
"POST:/math/factorial" method returned 13, then "post_route" will get a
reference to a scalar variable whose value is 13.
post_route {
my ($self, $meth_and_route, $ret) = @_;
if ($$ret == 13) {
# change the result to 14, because
# 13 is an unlucky number
$$ret = 14;
}
}
CONTEXTUAL MODE
*Note: contextual mode is an experimental feature introduced in v1.2.0
and may change in the future.*
Contextual mode is an optional way of writing "McBain" APIs, reminiscent
of web application frameworks such as Catalyst and Leyland. The main
idea is that a context object is created for every request, and follows
it during its entire life.
In regular mode, the API methods receive the class of the root package
(or its object, if writing object oriented APIs), and a hash-ref of
parameters. This is okay for simple APIs, but many APIs need more, like
information about the user who sent the request.
In contextual mode, the context object can contain user information,
methods for checking authorization (think role-based and ability-based
authorization systems), database connections, and anything else your API
might need in order to fulfill the request.
Writing APIs in contextual mode is basically the same as in regular
mode, only you need to build a context class. Since "McBain" doesn't
intrude on your OO system of choice, constructing the class is your
responsibility, and you can use whatever you want (like Moo, Moose,
Class::Accessor). "McBain" only requires your context class to implement
a subroutine named "create_from_env( $runner, \%env, @args_to_call )".
This method will receive the name of the runner module used, the
standard environment hash-ref of "McBain" (which includes the keys
"METHOD", "ROUTE" and "PAYLOAD"), plus all of the arguments that were
sent to the "call( @args )" method. These are useful for certain runner
modules, such as the PSGI runner, which gets the PSGI hash-ref, from
which you can extract session data, user information, HTTP headers, etc.
Note that this means that if you plan to use your API with different
runner modules, your "create_from_env()" method should be able to parse
differently formatted arguments.
Note that currently, the context class has to be named
"__ROOT__::Context", where "__ROOT__" is the name of your API's root
package. So, for example, if your API's root package is named "MyAPI",
then "McBain" will expect "MyAPI::Context".
When writing in contextual mode, your API methods will receive the
context object instead of the root package/object, and the parameters
hash-ref.
Let's look at a simple example for writing APIs in contextual mode. Say
our API is called "MyAPI". Let's begin with the context class,
"MyAPI::Context":
package MyAPI::Context;
use Moo;
use Plack::Request;
has 'user_agent' => (
is => 'ro',
default => sub { 'none' }
);
sub create_from_env {
my ($class, $runner, $mcbain_env, @call_args) = @_;
my $user_agent;
if ($runner eq 'McBain::WithPSGI') {
# extract user agent from the PSGI env,
# which will be the first item in @call_args
$user_agent = Plack::Request->new($call_args[0])->user_agent;
}
return $class->new(user_agent => $user_agent);
}
1;
Now let's look at the API itself:
package MyAPI;
use McBain -contextual;
get '/' => (
cb => sub {
my ($c, $params) = @_;
if ($c->user_agent =~ m/Android/) {
# do it this way
} else {
# do it that way
}
# you can still forward to other methods
$c->forward('GET:/something_else', \%other_params);
}
);
1;
So as you can see, the only real change for API packages is the need to
write "use McBain -contextual" instead of "use McBain". The only
"challenge" is writing the context class.
MCBAIN RUNNERS
*NOTE: since v2.0.0 the way runner modules are used has changed. The
"MCBAIN_WITH" environment variable is no longer used. Read on for more
information.*
A runner module is in charge of loading "McBain" APIs in a specific way.
The default runner, McBain::Directly, is the simplest runner there is,
and is meant for using APIs directly from Perl code.
The runner module is in charge of whatever heavy lifting is required in
order to turn your API into a "service", or an "app", or whatever it is
you think your API needs to be.
The following runners are currently available:
* McBain::Directly - Directly use an API from Perl code.
* McBain::WithPSGI - Turn an API into a Plack based, JSON-to-JSON
RESTful web application.
* McBain::WithGearmanXS - Turn an API into a JSON-to-JSON Gearman
worker.
* McBain::WithWebSocket - Turn an API into a WebSocket server.
* McBain::WithZeroMQ - Turn an API into a JSON-to-JSON ZeroMQ REP
worker.
The latter four completely change the way your API is used, and yet you
can see their code is very short.
To tell "McBain" which runner module to use, you must provide the name
of the runner when loading your API:
use MyAPI -withPSGI; # can also write -WithPSGI
In the above example, "McBain::WithPSGI" will be the runner module used.
The default runner module is "McBain::Directly". If you "use" an API
with no parameter, it will be the loaded runner module:
use MyAPI;
use MyAPI -directly; # the same as above
You can easily create your own runner modules, so that your APIs can be
used in different ways. A runner module needs to implement the following
interface:
init( $runner_class, $target_class )
This method is called when "McBain" is first imported into an API topic.
$target_class will hold the name of the class currently being imported
to.
You can do whatever initializations you need to do here, possibly
manipulating the target class directly. You will probably only want to
do this on the root topic, which is why "is_root( )" is available on
$target_class.
You can look at "WithPSGI" and "WithGearmanXS" to see how they're using
the "init()" method. For example, in "WithPSGI", Plack::Component is
added to the @ISA array of the root topic, so that it turns into a Plack
app. In "WithGearmanXS", the "init()" method is used to define a
"work()" method on the root topic, so that your API can run as any
standard Gearman worker.
generate_env( $runner_class, @call_args )
This method receives whatever arguments were passed to the "call( @args
)" method. It is in charge of returning a standard hash-ref that
"McBain" can use in order to determine which route the caller wants to
execute, and with what parameters. Remember that the way "call()" is
invoked depends on the runner used.
The hash-ref returned *must* have the following key-value pairs:
* ROUTE - The route to execute (string).
* METHOD - The method to call on the route (string).
* PAYLOAD - A hash-ref of parameters to provide for the method. If no
parameters are provided, an empty hash-ref should be given.
The returned hash-ref is called $env, inspired by PSGI.
generate_res( $runner_class, \%env, $result )
This method formats the result from a route before returning it to the
caller. It receives the $env hash-ref (if needed), and the result from
the route. In the "WithPSGI" runner, for example, this method encodes
the result into JSON and returns a proper PSGI response array-ref.
handle_exception( $runner_class, $error, @args )
This method will be called whenever a route raises an exception, or
otherwise your code fails. The $error variable will always be a standard
exception hash-ref, with "code" and "error" keys, and possibly more.
Read the discussion above.
The method should format the error before returning it to the user,
similar to what "generate_res()" above performs, but it allows you to
handle exceptions gracefully.
Whatever arguments were provided to "call()" will be provided to this
method as-is, so that you can inspect or use them if need be.
"WithGearmanXS", for example, will get the Gearman::XS::Job object and
call the "send_fail()" method on it, to properly indicate the job
failed.
CONFIGURATION AND ENVIRONMENT
No configuration files or environment variables required.
DEPENDENCIES
"McBain" depends on the following CPAN modules:
* Brannigan
* Carp
* File::Spec
* Scalar::Util
* Try::Tiny
The command line utility, mcbain2pod, depends on the following CPAN
modules:
* IO::Handle
* Getopt::Compact
* Module::Load
INCOMPATIBILITIES WITH OTHER MODULES
None reported.
BUGS AND LIMITATIONS
Please report any bugs or feature requests to "
[email protected]",
or through the web interface at
<
http://rt.cpan.org/NoAuth/ReportBug.html?Queue=McBain>.
SUPPORT
You can find documentation for this module with the perldoc command.
perldoc McBain
You can also look for information at:
* RT: CPAN's request tracker
<
http://rt.cpan.org/NoAuth/Bugs.html?Dist=McBain>
* AnnoCPAN: Annotated CPAN documentation
<
http://annocpan.org/dist/McBain>
* CPAN Ratings
<
http://cpanratings.perl.org/d/McBain>
* Search CPAN
<
http://search.cpan.org/dist/McBain/>
AUTHOR
Ido Perlmuter <
[email protected]>
LICENSE AND COPYRIGHT
Copyright (c) 2013-2014, Ido Perlmuter "
[email protected]".
This module is free software; you can redistribute it and/or modify it
under the same terms as Perl itself, either version 5.8.1 or any later
version. See perlartistic and perlgpl.
The full text of the license can be found in the LICENSE file included
with this module.
DISCLAIMER OF WARRANTY
BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER
EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE
ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH
YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
NECESSARY SERVICING, REPAIR, OR CORRECTION.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE LIABLE
TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL, OR
CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE
SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
DAMAGES.
