#!/usr/bin/env perl

# ABSTRACT: Command-line interface for RESTful web services
# PODNAME: presto

use App::Presto;

App::Presto->run(@ARGV);


__END__
=pod

=head1 NAME

presto - Command-line interface for RESTful web services

=head1 VERSION

version 0.008

=head1 SYNOPSIS

Invoke from the shell:

	bash$ presto http://my-server.com

Very basic usage:

	http://my-server.com> GET /product/1.json
	{"id":1,"name":"My Product"}

	http://my-server.com> HEAD /product/1.json
	HTTP/1.1 200 OK
	Connection: close
	Date: Thu, 28 Jun 2012 21:05:33 GMT
	Content-Length: 0
	Content-Type: application/json
	Client-Date: Thu, 28 Jun 2012 21:05:44 GMT
	Client-Response-Num: 1

=head1 DESCRIPTION

C<App::Presto> provides a command-line interface (CLI) for RESTful
web services.  When looking for a way to interact with RESTful services
answers typically point to some horrible GUI or (on the complete opposite
end of the spectrum) just using C<curl> directly on the command-line.
This tool attempts to find some sort of middle ground by providing a
quasi-DSL for interacting with a RESTful service in an interactive way.

=head1 FEATURES

=head2 Basic HTTP methods

All HTTP methods are implemented as commands in presto.  The URL that is
given is appended to the endpoint specified when presto is invoked as
shown in the SYNOPSIS above.

=head2 Request Building

If the endpoint contains a C<*> character the URL fragment specified in
the GET/POST/etc command is inserted at that point. This allows you to
do things like auto-append a file extension to all URLs.  For instance:

	bash$ presto http://my-server.com*.json
	http://my-server.com> GET /product/1

In this case, the full URL would be
C<http://my-server.com/product/1.json>.  If no C<*> is found in the URL,
the URL fragment is simply appended at the end of the endpoint.

All arguments after the first will be treated as query parameters (for
GET/HEAD/DELETE requests) or request content (for POST/PUT requests). For
instance:

	http://my-server.com> GET /products limit=10 offset=20
  # request goes to http://my-sever.com/products?limit=10&offset=20

	http://my-server.com> POST /products '{"name":"A New Product"}'
  # request goes to http://my-sever.com/products with the body as specified

You can also specify additional headers you would like included in the request:

	# the ":" is optional
	http://my-server.com> header Accept: application/json

	# shortcut for "header Content-Type application/json"
	http://my-server.com> type application/json

	# shortcut for "header Authorization: Basic QWxhZGRpbjpvcGVuIHNlc2FtZQ==
	http://my-server.com> authorization Aladdin 'open sesame'

	# view all headers
	http://my-server.com> headers

	# view specific header
	http://my-server.com> header Authorization

	# clear all headers
	http://my-server.com> headers --clear

	# unset specific header
	http://my-server.com> header --unset Authorization

If you are creating form URL-encoded data, a shortcut has been made to avoid having
to manually URL-encode everything manually:

	http://my-server.com> form foo=bar baz=1,2,3

	# outputs
	foo=bar&baz=1%2C2%2C3

=head2 Response Handling

By default, presto will just dump the response body to the screen after
a request is completed. There are additional options, however:

	# dump full request/response to the screen (exactly as transmitted over the wire)
	http://my-server.com> config verbose 1

	# parse the response according to the content-type and use
	# Data::Dumper to display it
	http://my-server.com> config deserialize_response 1

	# use something other than Data::Dumper to dump a parsed
	# response body
	http://my-server.com> config pretty_printer JSON
	http://my-server.com> config pretty_printer Data::Dump

	# send the output to a file (the '>' must not be followed by any white-space!)
	http://my-server.com> GET /some-image.png >some-image.png

Pretty-printing can be especially helpful for making XML or JSON response
bodies more human-readable.

When C<deserialize_response> is set, if the content-type of the
response is "text/html", the HTML is automatically stripped with
L<HTML::FormatText::WithLinks> and displayed as formatted text.

If the request or response body is binary (using a simple heuristic
like the C<-B> file-test operator), the output is not printed to STDOUT.
Instead, you may want to use output redirection as show above and send
the response body to a file.

	http://my-server.com> GET /some-image.jpg >foo.jpg

=head2 Persistent Configuration

As demonstrated above, you can use the C<config> command to change the
behavior of presto.  These configuration options are persisted in a
config file specific to the endpoint provided at the command-line and
will be reloaded the next time you invoke presto with the same endpoint.

Current valid config keys are:

=over 4

=item * verbose

Boolean, when enabled, dumps request/response to STDOUT (defaults to "0")

=item * deserialize_response 

Boolean, when enabled response body is parsed based on the C<Content-Type>
header (defaults to "1")

=item * pretty_printer

Must be one of the supported modules (i.e. Data::Dumper or JSON).
Use tab completion to see currently supported values (defaults to "JSON").

=item * binmode

Used to set encoding of STDIN and STDOUT handles (defaults to "utf8")

=back

B<TODO:> provide a means for aliasing endpoints so that configuration
is shared across multiple endpoints.

=head2 History and Scripting

Just like configuration, command history is maintained separately for each
endpoint specified on the command-line and is persisted across sessions
(assuming you have a capable Term::Readline library installed).  You can
interrogate the history using the (surprisingly named) C<history> command.
It supports a small subset of the C<bash> history command:

	# dump all history
	http://my-server.com> history

	# dump last 5 entries
	http://my-server.com> history 5

	# delete specific history entries
	http://my-server.com> history -d 4

	# clear history
	http://my-server.com> history -c

Presto also provides a way of saving and replaying bits of your command
history. Here are some examples:

	# save all history to script file "my-script"
	http://my-server.com> save my-script

	# save the last 5 history entries
	http://my-server.com> save my-script 5

	# save entries 3-7
	http://my-server.com> save my-script 3..7 

To replay scripts:

	http://my-server.com> source my-script

	# prompt before each command
	http://my-server.com> source -i my-script

=head2 Variable interpolation

At times (especially when working with scripts) it might be handy to
use elements from a previous response to affect a subsequent request.
Anything inside a balanced C<$(...)> will be interpolated for you.
For instance, a very contrived example:

	# hypothetical authentication protocal that returns a token in the response headers
	http://my-server.com> POST /auth.json username=jdoe&password=s3cr3t
	{"authenticated":true}

	# see the authentication token
	http://my-server.com> echo $(HEADER[X-Auth-Token])
	2c26b46b68ffc68ff99b453c1d30413413422d706483bfa0f98a5e886266e7ae

If you need to include that in subsequent request, you can use the "stash" feature:

	# store the value
	http://my-server.com> stash auth-token $(HEADER[X-Auth-Token])

	# use the value later
	http://my-server.com> header X-Auth-Token $(STASH[auth-token])

Those variable substitutions can be used anywhere in a command.  C<HEADER>
and C<BODY> always refer to the most recent request while the C<STASH>
is a persisted for the life of the process.

One useful feature for scripting is to prompt for user input.  You can do
this by using the C<PROMPT> pseudo-variable.  The first set of brackets
specify the prompt value.  The second (optional) set of brackets specify
the initial value.  An example:

	# collect the username/password from the user
	http://my-server.com> stash username $(PROMPT[username:])
	http://my-server.com> stash password $(PROMPT[password:])

	# use the stashed values
	http://my-server.com> authorization $(STASH[username]) $(STASH[password])
	http://my-server.com> GET /$(STASH[username])/profile

	# or use a value that was prompted for directly (without stashing it)
	http://my-server.com> GET /products 'created_on=$(PROMPT[Created on (YYYY-MM-DD):])'

	# you can also specify initial values
	http://my-server.com> GET /products 'status=$(PROMPT[Product status:][active])'

You may also specify a local file to use as an argument to a command.  An example:

	http://my-server.com> POST /products $(FILE[my-product.xml])

The file is assumed to be in the same encoding as the C<binmode>
configuration.  If it is using a different character set, you can specify
that in a second bracketed parameter:

	http://my-server.com> POST /products $(FILE[my-product.xml][latin-1])

The contents of the file will be slurped, decoded and included as an
argument to the command as if you had typed it on the command-line
directly.

B<TODO:> Allow data structure references (from C<STASH> or even C<BODY>)
to be passed to a POST or PUT command which is then serialized based
on the content-type of the request before being sent over the wire.

=head2 (EXPERIMENTAL) Data::DPath integration

As an add-on to the variable interpolated described above, you can
use dpath expressions to further process the data returned from the
REST service.  Another very contrived example:

	http://my-server.com> GET /products.json
	[{"id":"1","name":"My Product"},{"id":"2","name":"Another Product"}]

	# issue a request to /product/2.json
	http://my-server.com> GET /product/$(BODY/id[-1]).json
	{"id":2,"name":"Another Product"}

In this example, anything after C<BODY> (including the C</>) is passed
to L<Data::DPath> and the result is then injected in it's place (the target
data for C<BODY> being the previous request's response data).

This feature will work on C<$(STASH)> values as well.

=head1 CAVEAT EMPTOR

This is beta-quality code and while I use it in my own daily workflow,
it is likely riddled with horribly obvious bugs and missing functionality
(let alone undocumented features).

=head1 ACKNOWLEDGEMENTS

Much of this was inspired by L<resty|https://github.com/micha/resty>
which is a rather magical (aka convoluted) set of bash functions (at least
for this occassional bash programmer).  After attempting to understand
and enhance L<resty|https://github.com/micha/resty>, I decided to try
my hand at creating something a little more perlish.

A big thank you to L<Shutterstock Images|http://shutterstock.com> for
allowing me to work on this on company time and release it to the CPAN.

=head1 AUTHOR

Brian Phillips <bphillips@cpan.org>

=head1 COPYRIGHT AND LICENSE

This software is copyright (c) 2012 by Brian Phillips and Shutterstock Images (http://shutterstock.com).

This is free software; you can redistribute it and/or modify it under
the same terms as the Perl 5 programming language system itself.

=cut

