=head1 NAME

Template::Alloy - TT2/3, HT, HTE, Tmpl, and Velocity Engine

=head1 SYNOPSIS

=head2 Template::Toolkit style usage

    my $t = Template::Alloy->new(
        INCLUDE_PATH => ['/path/to/templates'],
    );

    my $swap = {
        key1 => 'val1',
        key2 => 'val2',
        code => sub { 42 },
        hash => {a => 'b'},
    };

    # print to STDOUT
    $t->process('my/template.tt', $swap)
        || die $t->error;

    # process into a variable
    my $out = '';
    $t->process('my/template.tt', $swap, \$out);

    ### Alloy uses the same syntax and configuration as Template::Toolkit


=head2 HTML::Template::Expr style usage

    my $t = Template::Alloy->new(
        filename => 'my/template.ht',
        path     => ['/path/to/templates'],
    );

    my $swap = {
        key1 => 'val1',
        key2 => 'val2',
        code => sub { 42 },
        hash => {a => 'b'},
    };

    $t->param($swap);

    # print to STDOUT (errors die)
    $t->output(print_to => \*STDOUT);

    # process into a variable
    my $out = $t->output;

    ### Alloy can also use the same syntax and configuration as HTML::Template

=head2 Text::Tmpl style usage

    my $t = Template::Alloy->new;

    my $swap = {
        key1 => 'val1',
        key2 => 'val2',
        code => sub { 42 },
        hash => {a => 'b'},
    };

    $t->set_delimiters('#[', ']#');
    $t->set_strip(0);
    $t->set_values($swap);
    $t->set_dir('/path/to/templates');

    my $out = $t->parse_file('my/template.tmpl');

    my $str = "Foo #[echo $key1]# Bar";
    my $out = $t->parse_string($str);


    ### Alloy uses the same syntax and configuration as Text::Tmpl

=head2 Velocity (VTL) style usage

    my $t = Template::Alloy->new;

    my $swap = {
        key1 => 'val1',
        key2 => 'val2',
        code => sub { 42 },
        hash => {a => 'b'},
    };

    my $out = $t->merge('my/template.vtl', $swap);

    my $str = "#set($foo 1 + 3) ($foo) ($bar) ($!baz)";
    my $out = $t->merge(\$str, $swap);

=head2 Javascript style usage (requires Template::Alloy::JS)

    my $t = Template::Alloy->new;

    my $swap = {
        key1 => 'val1',
        key2 => 'val2',
        code => sub { 42 },
        hash => {a => 'b'},
    };

    my $out = '';
    $t->process_js('my/template.jstem', $swap, \$out);

    my $str = "[% var foo = 1 + 3; write('(' + foo + ') (' + get('key1') + ')'); %]";
    my $out = '';
    $t->process_js(\$str, $swap, \$out);


=head1 DESCRIPTION

"An alloy is a homogeneous mixture of two or more elements"
(http://en.wikipedia.org/wiki/Alloy).

Template::Alloy represents the mixing of features and capabilities
from all of the major mini-language based template systems (support
for non-mini-language based systems will happen eventually).  With
Template::Alloy you can use your favorite template interface and syntax
and get features from each of the other major template systems.  And
Template::Alloy is fast - whether your using mod_perl, CGI, or running
from the commandline.  There is even Template::Alloy::JS for getting
a little more speed when that is necessary.

Template::Alloy happened by accident (accidentally on purpose).  The
Template::Alloy (Alloy hereafter) was originally a part of the CGI::Ex
suite that performed simple variable interpolation.  It used TT2 style
variables in TT2 style tags "[% foo.bar %]".  That was all the
original Template::Alloy did.  This was fine and dandy for a couple
of years.  In winter of 2005-2006 Alloy was revamped to add a few
features.  One thing led to another and soon Alloy provided for most of
the features of TT2 as well as some from TT3.  Template::Alloy now
provides a full-featured implementation of the Template::Toolkit language.

After a move to a new company that was using HTML::Template::Expr and
Text::Tmpl templates, support was investigated and interfaces for
HTML::Template, HTML::Template::Expr, Text::Tmpl, and Velocity (VTL)
were added.  All of the various engines offer the same features - each
using a different syntax and interface.

More recently, the Template::Alloy::JS capabilities were introduced
to bring Javascript templates to the server side (along with an increase
in speed if ran in persistent environments).

Template::Toolkit brought the most to the table.  HTML::Template
brought the LOOP directive.  HTML::Template::Expr brought more vmethods
and using vmethods as top level functions.  Text::Tmpl brought the
COMMENT directive and encouraged speed matching (Text::Tmpl is almost
entirely C based and is very fast).  The Velocity engine brought
AUTO_EVAL and SHOW_UNDEFINED_INTERP.

Most of the standard Template::Toolkit documentation covering
directives, variables, configuration, plugins, filters, syntax, and
vmethods should apply to Alloy just fine (This pod tries to explain
everything - but there is too much).  See L<Template::Alloy::TT> for
a listing of the differences between Alloy and TT.

Most of the standard HTML::Template and
HTML::Template::Expr documentation covering methods, variables,
expressions, and syntax will apply to Alloy just fine as well.

Most of the standard Text::Tmpl documentation applies, as does
the documentation covering Velocity (VTL).

So should you use Template::Alloy ?  Well, try it out.  It may
give you no visible improvement.  Or it could.

=head1 BACKEND

Template::Alloy uses a recursive regex based grammar (early versions
during the CGI::Ex::Template phase did not).  This allows for the
embedding of opening and closing tags inside other tags (as in [% a =
"[% 1 + 2 %]" ; a|eval %]).  The individual methods such as parse_expr
and play_expr may be used by external applications to add TT style
variable parsing to other applications.

The regex parser returns an AST (abstract syntax tree) of the text,
directives, variables, and expressions.  All of the different template
syntax options compile to the same AST format.  The AST is composed
only of scalars and arrayrefs and is suitable for sending to
JavaScript via JSON or sharing with other languages.  The parse_tree
method is used for returning this AST.

Once at the AST stage, there are two modes of operation.  Alloy can
either operate directly on the AST using the Play role, or it can
compile the AST to perl code via the Compile role, and then execute
the code.  To use the perl code route, you must set the COMPILE_PERL
flag to 1.  If you are running in a cached-in-memory environment such
as mod_perl, this is the fastest option.  If you are running in a
non-cached-in-memory environment, then using the Play role to run the
AST is generally faster.  The AST method is also more secure as cached
AST won't ever eval any "perl" (assuming PERL blocks are disabled -
which is the default).

=head1 ROLES

Template::Alloy has split out its functionality into discrete roles.
In Template::Toolkit, this functionality is split into separate
classes.  The roles in Template::Alloy simply add on more methods to
the main class.  When Perl 6 arrives, these roles will be translated
into true Roles.

The following is a list of roles used by Template::Alloy.

    Template::Alloy::Compile  - Compile-to-perl role
    Template::Alloy::HTE      - HTML::Template::Expr role
    Template::Alloy::Operator - Operator role
    Template::Alloy::Parse    - Parse-to-AST role
    Template::Alloy::Play     - Play-AST role
    Template::Alloy::Stream   - Stream output role
    Template::Alloy::Tmpl     - Text::Tmpl role
    Template::Alloy::TT       - Template::Toolkit role
    Template::Alloy::Velocity - Velocity role
    Template::Alloy::VMethod  - Virtual methods role

    Template::Alloy::JS       - Javascript functionality - available separately

Template::Alloy automatically loads the roles when they are needed or
requested - but not sooner (with the exception of the Operator role
and the VMethod role which are always needed and always loaded).  This
is good for a CGI environment.  In mod_perl you may want to preload a
role to make the most of shared memory.  You may do this by passing
either the role name or a method supplied by that role.

    # import roles necessary for running TT
    use Template::Alloy qw(Parse Play Compile TT);

    # import roles based on methods
    use Template::Alloy qw(parse_tree play_tree compile_tree process);

Note: importing roles by method names does not import them into
that namespace - it is autoloading the role and methods into the
Template::Alloy namespace.  To help make this more clear you may use
the following syntax as well.

    # import roles necessary for running TT
    use Template::Alloy load => qw(Parse Play Compile TT);

    # import roles based on methods
    use Template::Alloy load => qw(process parse_tree play_tree compile_tree);

    # import roles based on methods
    use Template::Alloy
        Parse => 1,
        Play => 1,
        Compile => 1,
        TT => 1;

Even with all roles loaded Template::Alloy is still relatively small.
You can load all of the roles (except the JS role) by passing "all" to
the use statement.

    use Template::Alloy 'all';

    # or
    use Template::Alloy load => 'all';

    # or
    use Template::Alloy all => 1;

As a final option, Template::Alloy also includes the ability to
stand-in for other template modules.  It is able to do this because it
supports the majority of the interface of the other template systems.
You can do this in the following way:

    use Template::Alloy qw(Text::Tmpl HTML::Template);

    # or
    use Template::Alloy load => qw(Text::Tmpl HTML::Template);

    # or
    use Template::Alloy
        'Text::Tmpl'     => 1,
        'HTML::Template' => 1;

Note that the use statement will die if any of the passed module names
are already loaded and not subclasses of Template::Alloy.  This will
avoid thinking that you are using Template::Alloy when you really aren't.
Using the 'all' option won't automatically do this - you must mention
the "stood-in" modules by name.

The following modules may be "stood-in" for:

    Template
    Text::Tmpl
    HTML::Template
    HTML::Template::Expr

This feature is intended to make using Template::Alloy with existing
code easier.  Most cases should work just fine.  Almost all syntax will
just work (except Alloy may make some things work that were previously
broken).  However Template::Alloy doesn't support 100% of the interface
of any of the template systems.  If you are using "features-on-the-edge"
then you may need to re-write portions of your code that interact with
the template system.

=head1 PUBLIC METHODS

The following section lists most of the publicly available methods.  Some less
commonly used public methods are listed later in this document.

=over 4

=item C<new>

    my $obj = Template::Alloy->new({
        INCLUDE_PATH => ['/my/path/to/content', '/my/path/to/content2'],
    });

Arguments may be passed as a hash or as a hashref.  Returns a Template::Alloy object.

There are currently no errors during Template::Alloy object creation.  If you are
using the HTML::Template interface, this is different behavior.  The document is
not parsed until the output or process methods are called.

=item C<process>

This is the TT interface for starting processing.  Any errors that result in the
template processing being stopped will be stored and available via the ->error method.

    my $t = Template::Alloy->new;
    $t->process($in, $swap, $out)
        || die $t->error;

Process takes three arguments.

The $in argument can be any one of:

    String containing the filename of the template to be processed.
    The filename should be relative to INCLUDE_PATH.  (See
    INCLUDE_PATH, ABSOLUTE, and RELATIVE configuration items).  In
    memory caching and file side caching are available for this type.

    A reference to a scalar containing the contents of the template to be processed.

    A coderef that will be called to return the contents of the template.

    An open filehandle that will return the contents of the template when read.

The $swap argument should be hashref containing key value pairs that
will be available to variables swapped into the template.  Values can
be hashrefs, hashrefs of hashrefs and so on, arrayrefs, arrayrefs of
arrayrefs and so on, coderefs, objects, and simple scalar values such
as numbers and strings.  See the section on variables.

The $out argument can be any one of:

    undef - meaning to print the completed template to STDOUT.

    String containing a filename.  The completed template will be placed in the file.

    A reference to a string.  The contents will be appended to the scalar reference.

    A coderef.  The coderef will be called with the contents as a single argument.

    An object that can run the method "print".  The contents will be passed as
    a single argument to print.

    An arrayref.  The contents will be pushed onto the array.

    An open filehandle.  The contents will be printed to the open handle.

Additionally - the $out argument can be configured using the OUTPUT configuration
item.

The process method defaults to using the "cet" syntax which will parse TT3 and most
TT2 documents.  To parse HT or HTE documents, you must pass the SYNTAX configuration
item to the "new" method.  All calls to process would then default to HTE syntax.

    my $obj = Template::Alloy->new(SYNTAX => 'hte');

=item C<process_simple>

Similar to the process method but with the following restrictions:

The $in parameter is limited to a filename or a reference a string
containing the contents.

The $out parameter may only be a reference to a scalar string that
output will be appended to.

Additionally, the following configuration variables will be ignored:
VARIABLES, PRE_DEFINE, BLOCKS, PRE_PROCESS, PROCESS, POST_PROCESS,
AUTO_RESET, OUTPUT.

=item C<error>

Should something go wrong during a "process" command, the error that
occurred can be retrieved via the error method.

    $obj->process('somefile.html', {a => 'b'}, \$string_ref)
        || die $obj->error;

=item C<output>

HTML::Template way to process a template.  The output method requires
that a filename, filehandle, scalarref, or arrayref argument was
passed to the new method.  All of the HT calling conventions for new
are supported.  The key difference is that Alloy will not actually
process the template until the output method is called.

    my $obj = Template::Alloy->new(filename => 'myfile.html');
    $obj->param(\%swap);
    print $obj->output;

See the HTML::Template documentation for more information.

The output method defaults to using the "hte" syntax which will parse
HTE and HT documents.  To parse TT3 or TT2 documents, you must pass
the SYNTAX configuration item to the "new" method.  All calls to
process would then default to TT3 syntax.

    my $obj = Template::Alloy->new(SYNTAX => 'tt3');

Any errors that occur during the output method will die with the error
as the die value.

=item C<param>

HTML::Template way to get or set variable values that will be used by
the output method.

    my $val = $obj->param('key'); # get one value

    $obj->param(key => $val);     # set one value

    $obj->param(key => $val, key2 => $val2);   # set multiple

    $obj->param({key => $val, key2 => $val2}); # set multiple

See the HTML::Template documentation for more information.

Note: Alloy does not support the die_on_bad_params configuration.
This is because Alloy does not resolve variable names until the output
method is called.

=item C<define_vmethod>

This method is available for defining extra Virtual methods or
filters.  This method is similar to Template::Stash::define_vmethod.

    Template::Alloy->define_vmethod(
        'text',
        reverse => sub { my $item = shift; return scalar reverse $item },
    );

=item C<register_function>

This is the HTML::Template way of defining text vmethods.  It is the same as
calling define_vmethod with "text" as the first argument.

    Template::Alloy->register_function(
        reverse => sub { my $item = shift; return scalar reverse $item },
    );

=item C<define_directive>

This method can be used for adding new directives or overridding existing
ones.

   Template::Alloy->define_directive(
       MYDIR => {
           parse_sub => sub {}, # parse additional items in the tag
           play_sub  => sub {
               my ($self, $ref, $node, $out_ref) = @_;
               $$out_ref .= "I always say the same thing!";
               return;
           },
           is_block  => 1,  # is this block like
           is_postop => 0,  # not a post operative directive
           no_interp => 1,  # no interpolation in this block
           continues => undef, # it doesn't "continue" any other directives
       },
   );

Now with a template like:

   my $str = "([% MYDIR %]This is something[% END %])";
   Template::Alloy->new->process(\$str);

You will get:

   (I always say the same thing!)

We'll add more details in later revisions of this document.

=item C<define_syntax>

This method can be used for adding another syntax to or overriding
existing ones in the list of choices available in Alloy.  The syntax can
be chosen by the SYNTAX configuration item.

    Template::Alloy->define_syntax(
        my_uber_syntax => sub {
            my $self = shift;
            local $self->{'V2PIPE'}      = 0;
            local $self->{'V2EQUALS'}    = 0;
            local $self->{'PRE_CHOMP'}   = 0;
            local $self->{'POST_CHOMP'}  = 0;
            local $self->{'NO_INCLUDES'} = 0;
            return $self->parse_tree_tt3(@_);
        },
    );

The subroutine that is used must return an opcode tree (AST) that
can be played by the execute_tree method.

=item C<define_operator>

This method allows for adding new operators or overriding existing ones.

    Template::Alloy->define_operator({
        type       => 'right', # can be one of prefix, postfix, right, left, none, ternary, assign
        precedence => 84,      # relative precedence for resolving multiple operators without parens
        symbols    => ['foo', 'FOO'], # any mix of chars can be used for the operators
        play_sub   => sub {
            my ($one, $two) = @_;
            return "You've been foo'ed ($one, $two)";
        },
    });

You can then use it in a template as in the following:

   my $str = "[% 'ralph' foo 1 + 2 * 3 %]";
   Template::Alloy->new->process(\$str);

You will get:

   You've been foo'ed (ralph, 7)

Future revisions of this document will include more samples.
This is an experimental feature and the API will probably change.

=item C<dump_parse_tree>

This method allows for returning a Data::Dumper dump of a parsed
template.  It is mainly used for testing.

=item C<dump_parse_expr>

This method allows for returning a Data::Dumper dump of a parsed
variable.  It is mainly used for testing.

=item C<import>

All of the arguments that can be passed to "use" that are listed
above in the section dealing with ROLES, can be used with the
import method.

    # import by role
    Template::Alloy->import(qw(Compile Play Parse TT));

    # import by method
    Template::Alloy->import(qw(compile_tree play_tree parse_tree process));

    # import by "stand-in" class
    Template::Alloy->import('Text::Tmpl', 'HTML::Template::Expr');

As mentioned in the ROLE section - arguments passed to import are
not imported into current namespace.  Roles and methods are only
imported into the Template::Alloy namespace.

=back

=head1 VARIABLES

This section discusses how to use variables and expressions in the TT
mini-language.

A variable is the most simple construct to insert into the TT mini
language.  A variable name will look for the matching value inside
Template::Alloys internal stash of variables which is essentially a
hash reference.  This stash is initially populated by either passing a
hashref as the second argument to the process method, or by setting
the "VARIABLES" or "PRE_DEFINE" configuration variables.

If you are using either the HT or the HTE syntax, the VAR, IF, UNLESS,
LOOP, and INCLUDE directives will accept a NAME attribute which may
only be a single level (non-chained) HTML::Template variable name, or
they may accept an EXPR attribute which may be any valid TT3 variable
or expression.

The following are some sample ways to access variables.

    ### some sample variables
    my %vars = (
        one       => '1.0',
        foo       => 'bar',
        vname     => 'one',
        some_code => sub { "You passed me (".join(', ', @_).")" },
        some_data => {
            a     => 'A',
            bar   => 3234,
            c     => [3, 1, 4, 1, 5, 9],
            vname => 'one',
        },
        my_list   => [20 .. 50],
        cet       => Template::Alloy->new,
    );

    ### pass the variables into the Alloy process
    $cet->process($template_name, \%vars)
         || die $cet->error;

    ### pass the variables during object creation (will be available to every process call)
    my $cet = Template::Alloy->new(VARIABLES => \%vars);

=head2 GETTING VARIABLES

Once you have variables defined, they can be used directly in the
template by using their name in the stash.  Or by using the GET
directive.

    [% foo %]
    [% one %]
    [% GET foo %]

Would print when processed:

    bar
    1.0
    bar

To access members of a hashref or an arrayref, you can chain together
the names using a ".".

    [% some_data.a %]
    [% my_list.0] [% my_list.1 %] [% my_list.-1 %]
    [% some_data.c.2 %]

Would print:

    A
    20 21 50
    4

If the value of a variable is a code reference, it will be called.
You can add a set of parenthesis and arguments to pass arguments.
Arguments are variables and can be as complex as necessary.

    [% some_code %]
    [% some_code() %]
    [% some_code(foo) %]
    [% some_code(one, 2, 3) %]

Would print:

    You passed me ().
    You passed me ().
    You passed me (bar).
    You passed me (1.0, 2, 3).

If the value of a variable is an object, methods can be called using
the "." operator.

    [% cet %]

    [% cet.dump_parse_expr('1 + 2').replace('\s+', ' ') %]

Would print something like:

    Template::Alloy=HASH(0x814dc28)

    $VAR1 = [ [ undef, '+', '1', '2' ], 0 ];

Each type of data (string, array and hash) have virtual methods
associated with them.  Virtual methods allow for access to functions
that are commonly used on those types of data.  For the full list of
built in virtual methods, please see the section titled VIRTUAL
METHODS

    [% foo.length %]
    [% my_list.size %]
    [% some_data.c.join(" | ") %]

Would print:

    3
    31
    3 | 1 | 4 | 5 | 9

It is also possible to "interpolate" variable names using a "$".  This
allows for storing the name of a variable inside another variable.  If
a variable name is a little more complex it can be embedded inside of
"${" and "}".

    [% $vname %]
    [% ${vname} %]
    [% ${some_data.vname} %]
    [% some_data.$foo %]
    [% some_data.${foo} %]

Would print:

    1.0
    1.0
    1.0
    3234
    3234

In Alloy it is also possible to embed any expression (non-directive) in
"${" and "}" and it is possible to use non-integers for array access.
(This is not available in TT2)

    [% ['a'..'z'].${ 2.3 } %]
    [% {ab => 'AB'}.${ 'a' ~ 'b' } %]
    [% color = qw/Red Blue/; FOR [1..4] ; color.${ loop.index % color.size } ; END %]

Would print:

    c
    AB
    RedBlueRedBlue

=head2 SETTING VARIABLES.

To define variables during processing, you can use the = operator.  In
most cases this is the same as using the SET directive.

    [% a = 234 %][% a %]
    [% SET b = "Hello" %][% b %]

Would print:

    234
    Hello

It is also possible to create arrayrefs and hashrefs.

    [% a = [1, 2, 3] %]
    [% b = {key1 => 'val1', 'key2' => 'val2'} %]

    [% a.1 %]
    [% b.key1 %] [% b.key2 %]

Would print:

    2
    val1 val2

It is possible to set multiple values in the same SET directive.

    [% SET a = 'A'
           b = 'B'
           c = 'C' %]
    [% a %]    [% b %]    [% c %]

Would print:

    A    B    C

It is also possible to unset variables, or to set members of
nested data structures.

    [% a = 1 %]
    [% SET a %]

    [% b.0.c = 37 %]

    ([% a %])
    [% b.0.c %]

Would print

    ()
    37

=head1 LITERALS AND CONSTRUCTORS

The following are the types of literals (numbers and strings) and
constructors (hash and array constructs) allowed in Alloy.  They can be
used as arguments to functions, in place of variables in directives,
and in place of variables in expressions.  In Alloy it is also possible
to call virtual methods on literal values.

=over 4

=item Integers and Numbers.

    [% 23423   %]        Prints an integer.
    [% 3.14159 %]        Prints a number.
    [% pi = 3.14159 %]   Sets the value of the variable.
    [% 3.13159.length %] Prints 7 (the string length of the number)

Scientific notation is supported.

    [% 314159e-5 + 0 %]      Prints 3.14159.

    [% .0000001.fmt('%.1e') %]  Prints 1.0e-07

Hexadecimal input is also supported.

    [% 0xff + 0 %]    Prints 255

    [% 48875.fmt('%x') %]  Prints beeb

=item Single quoted strings.

Returns the string.  No variable interpolation happens.

    [% 'foobar' %]          Prints "foobar".
    [% '$foo\n' %]          Prints "$foo\\n".  # the \\n is a literal "\" and an "n"
    [% 'That\'s nice' %]    Prints "That's nice".
    [% str = 'A string' %]  Sets the value of str.
    [% 'A string'.split %]  Splits the string on ' ' and returns the list.

Note: virtual methods can only be used on literal strings in Alloy, not in TT.

You may also embed the current tags in strings (Alloy only).

    [% '[% 1 + 2 %]' | eval %]  Prints "3"

=item Double quoted strings.

Returns the string.  Variable interpolation happens.

    [% "foobar" %]                   Prints "foobar".
    [% "$foo"   %]                   Prints "bar" (assuming the value of foo is bar).
    [% "${foo}" %]                   Prints "bar" (assuming the value of foo is bar).
    [% "foobar\n" %]                 Prints "foobar\n".  # the \n is a newline.
    [% str = "Hello" %]              Sets the value of str.
    [% "foo".replace('foo','bar') %] Prints "bar".

Note: virtual methods can only be used on literal strings in Alloy, not in TT.

You may also embed the current tags in strings (Alloy only).

    [% "[% 1 + 2 %]" | eval %]  Prints "3"

=item Array Constructs.

    [% [1, 2, 3] %]               Prints something like ARRAY(0x8309e90).
    [% array1 = [1 .. 3] %]       Sets the value of array1.
    [% array2 = [foo, 'a', []] %] Sets the value of array2.
    [% [4, 5, 6].size %]          Prints 3.
    [% [7, 8, 9].reverse.0 %]     Prints 9.

Note: virtual methods can only be used on array contructs in Alloy, not in TT.

=item Quoted Array Constructs.

    [% qw/1 2 3/ %]                Prints something like ARRAY(0x8309e90).
    [% array1 = qw{Foo Bar Baz} %] Sets the value of array1.
    [% qw[4 5 6].size %]           Prints 3.
    [% qw(Red Blue).reverse.0 %]   Prints Blue.

Note: this works in Alloy and is planned for TT3.

=item Hash Constructs.

    [% {foo => 'bar'} %]                 Prints something like HASH(0x8305880)
    [% hash = {foo => 'bar', c => {}} %] Sets the value of hash.
    [% {a => 'A', b => 'B'}.size %]      Prints 2.
    [% {'a' => 'A', 'b' => 'B'}.size %]  Prints 2.
    [% name = "Tom" %]
    [% {Tom => 'You are Tom',
        Kay => 'You are Kay'}.$name %]   Prints You are Tom

Note: virtual methods can only be used on hash contructs in Alloy, not in TT.

=item Regex Constructs.

    [% /foo/ %]                              Prints (?-xism:foo)
    [% a = /(foo)/i %][% "FOO".match(a).0 %] Prints FOO

Note: this works in Alloy and is planned for TT3.

=back

=head1 VIRTUAL METHODS

Virtual methods (vmethods) are a TT feature that allow for operating on the
swapped template variables.

This document shows some samples of using vmethods.  For a full
listing of available virtual methods, see L<Template::Alloy::VMethod>.

=head1 EXPRESSIONS

Expressions are one or more variables or literals joined together with
operators.  An expression can be used anywhere a variable can be used
with the exception of the variable name in the SET directive, and the
filename of PROCESS, INCLUDE, WRAPPER, and INSERT.

For a full listing of operators, see L<Template::Alloy::Operator>.

The following section shows some samples of expressions.  For a full
list of available operators, please see the section titled OPERATORS.

    [% 1 + 2 %]           Prints 3
    [% 1 + 2 * 3 %]       Prints 7
    [% (1 + 2) * 3 %]     Prints 9

    [% x = 2 %]                      # assignments don't return anything
    [% (x = 2) %]         Prints 2   # unless they are in parens
    [% y = 3 %]
    [% x * (y - 1) %]     Prints 4

=head1 DIRECTIVES

This section contains the alphabetical list of DIRECTIVES available in
Alloy.  DIRECTIVES are the "functions" and control structures that
work in the various mini-languages.  For further discussion and
examples beyond what is listed below, please refer to the TT
directives documentation or to the appropriate documentation
for the particular directive.

The examples given in this section are done using the
Template::Toolkit syntax, but can be done in any of the various
syntax options.  See L<Template::Alloy::TT>, L<Template::Alloy::HTE>,
L<Template::Alloy::Tmpl>, and L<Template::Alloy::Velocity>.

    [% IF 1 %]One[% END %]
    [% FOREACH a = [1 .. 3] %]
        a = [% a %]
    [% END %]

    [% SET a = 1 %][% SET a = 2 %][% GET a %]

In TT multiple directives can be inside the same set of '[%' and '%]'
tags as long as they are separated by space or semi-colons (;) (The
Alloy version of Tmpl allows multiple also - but none of the other
syntax options do).  Any block directive that can also be used as a
post-operative directive (such as IF, WHILE, FOREACH, UNLESS, FILTER,
and WRAPPER) must be separated from preceding directives with a
semi-colon if it is being used as a block directive.  It is more safe
to always use a semi-colon.  Note: separating by space is only
available in Alloy but is a planned TT3 feature.

    [% SET a = 1 ; SET a = 2 ; GET a %]
    [% SET a = 1
       SET a = 2
       GET a
     %]

    [% GET 1
         IF 0   # is a post-operative
       GET 2 %] # prints 2

    [% GET 1;
       IF 0     # it is block based
         GET 2
       END
     %]         # prints 1

The following is the list of directives.

=over 4

=item C<BLOCK>

Saves a block of text under a name for later use in PROCESS, INCLUDE,
and WRAPPER directives.  Blocks may be placed anywhere within the
template being processed including after where they are used.

    [% BLOCK foo %]Some text[% END %]
    [% PROCESS foo %]

    Would print

    Some text

    [% INCLUDE foo %]
    [% BLOCK foo %]Some text[% END %]

    Would print

    Some text

Anonymous BLOCKS can be used for capturing.

    [% a = BLOCK %]Some text[% END %][% a %]

    Would print

    Some text

Anonymous BLOCKS can be used with macros.


=item C<BREAK>

Alias for LAST.  Used for exiting FOREACH and WHILE loops.

=item C<CALL>

Calls the variable (and any underlying coderefs) as in the GET method, but
always returns an empty string.

=item C<CASE>

Used with the SWITCH directive.  See the L</"SWITCH"> directive.

=item C<CATCH>

Used with the TRY directive.  See the L</"TRY"> directive.

=item C<CLEAR>

Clears any of the content currently generated in the innermost block
or template.  This can be useful when used in conjunction with the TRY
statement to clear generated content if an error occurs later.

=item C<COMMENT>

Will comment out any text found between open and close tags.  Note, that
the intermediate items are still parsed and END tags must align - but the
parsed content will be discarded.

    [% COMMENT %]
       This text won't be shown.
       [% IF 1 %]And this won't either.[% END %]
    [% END %]

=item C<CONFIG>

Allow for changing the value of some compile time and runtime configuration
options.

    [% CONFIG
        ANYCASE   => 1
        PRE_CHOMP => '-'
    %]

The following compile time configuration options may be set:

    ANYCASE
    AUTO_EVAL
    AUTO_FILTER
    CACHE_STR_REFS
    ENCODING
    INTERPOLATE
    POST_CHOMP
    PRE_CHOMP
    SEMICOLONS
    SHOW_UNDEFINED_INTERP
    SYNTAX
    V1DOLLAR
    V2EQUALS
    V2PIPE

The following runtime configuration options may be set:

    ADD_LOCAL_PATH
    CALL_CONTEXT
    DUMP
    VMETHOD_FUNCTIONS
    STRICT (can only be enabled, cannot be disabled)

If non-named parameters as passed, they will show the current configuration:

   [% CONFIG ANYCASE, PRE_CHOMP %]

   CONFIG ANYCASE = undef
   CONFIG PRE_CHOMP = undef

=item C<DEBUG>

Used to reset the DEBUG_FORMAT configuration variable, or to turn
DEBUG statements on or off.  This only has effect if the DEBUG_DIRS or
DEBUG_ALL flags were passed to the DEBUG configuration variable.

    [% DEBUG format '($file) (line $line) ($text)' %]
    [% DEBUG on %]
    [% DEBUG off %]

=item C<DEFAULT>

Similar to SET, but only sets the value if a previous value was not
defined or was zero length.

    [% DEFAULT foo = 'bar' %][% foo %] => 'bar'

    [% foo = 'baz' %][% DEFAULT foo = 'bar' %][% foo %] => 'baz'

=item C<DUMP>

DUMP inserts a Data::Dumper printout of the variable or expression.
If no argument is passed it will dump the entire contents of the
current variable stash (with private keys removed).

The output also includes the current file and line number that the
DUMP directive was called from.

See the DUMP configuration item for ways to customize and control
the output available to the DUMP directive.

    [% DUMP %] # dumps everything

    [% DUMP 1 + 2 %]

=item C<ELSE>

Used with the IF directive.  See the L</"IF"> directive.

=item C<ELSIF>

Used with the IF directive.  See the L</"IF"> directive.

=item C<END>

Used to end a block directive.

=item C<EVAL>

Same as the EVALUATE directive.

=item C<EVALUATE>

Introduced by the Velocity templating language.  Parses and processes the
contents of the passed item.  This is similar to the eval filter, but
Velocity needs a directive.  Named arguments may be used for
re-configuring the parser.  Any of the items that can be passed to
the CONFIG directive may be passed here.

    [% EVALUATE "[% 1 + 3 %]" %]

    [% foo = "bar" %]
    [% EVALUATE "<TMPL_VAR foo>" SYNTAX => 'ht' %]

=item C<FILTER>

Used to apply different treatments to blocks of text.  It may operate as a BLOCK
directive or as a post operative directive.  Alloy supports all of the filters in
Template::Filters.  The lines between scalar virtual methods and filters is blurred (or
non-existent) in Alloy.  Anything that is a scalar virtual method may be used as a FILTER.

TODO - enumerate the at least 7 ways to pass and use filters.

=item C<'|'>

Alias for the FILTER directive.  Note that | is similar to the
'.' in Template::Alloy.  Therefore a pipe cannot be used directly after a
variable name in some situations (the pipe will act only on that variable).
This is the behavior employed by TT3.  To get the TT2 behavior for a PIPE, use
the V2PIPE configuration item.

=item C<FINAL>

Used with the TRY directive.  See the L</"TRY"> directive.

=item C<FOR>

Alias for FOREACH

=item C<FOREACH>

Allows for iterating over the contents of any arrayref.  If the variable is not an
arrayref, it is automatically promoted to one.

    [% FOREACH i IN [1 .. 3] %]
        The variable i = [% i %]
    [%~ END %]

    [% a = [1 .. 3] %]
    [% FOREACH j IN a %]
        The variable j = [% j %]
    [%~ END %]

Would print:

        The variable i = 1
        The variable i = 2
        The variable i = 3

        The variable j = 1
        The variable j = 2
        The variable j = 3

You can also use the "=" instead of "IN" or "in".

    [% FOREACH i = [1 .. 3] %]
        The variable i = [% i %]
    [%~ END %]

    Same as before.

Setting into a variable is optional.

    [% a = [1 .. 3] %]
    [% FOREACH a %] Hi [% END %]

Would print:

     hi  hi  hi 

If the item being iterated is a hashref and the FOREACH does not
set into a variable, then values of the hashref are copied into
the variable stash.

    [% FOREACH [{a => 1}, {a => 2}] %]
        Key a = [% a %]
    [%~ END %]

Would print:

        Key a = 1
        Key a = 2

The FOREACH process uses the Template::Alloy::Iterator class to handle
iterations (It is compatible with Template::Iterator).  During the FOREACH
loop an object blessed into the iterator class is stored in the variable "loop".

The loop variable provides the following information during a FOREACH:

    index  - the current index
    max    - the max index of the list
    size   - the number of items in the list
    count  - index + 1
    number - index + 1
    first  - true if on the first item
    last   - true if on the last item
    next   - return the next item in the list
    prev   - return the previous item in the list
    odd    - return 1 if the current count is odd, 0 otherwise
    even   - return 1 if the current count is even, 0 otherwise
    parity - return "odd" if the current count is odd, "even" otherwise

The following:

    [% FOREACH [1 .. 3] %] [% loop.count %]/[% loop.size %] [% END %]

Would print:

     1/3  2/3  3/3 

The iterator is also available using a plugin.  This allows for access
to multiple "loop" variables in a nested FOREACH directive.

    [%~ USE outer_loop = Iterator(["a", "b"]) %]
    [%~ FOREACH i = outer_loop %]
        [%~ FOREACH j = ["X", "Y"] %]
           [% outer_loop.count %]-[% loop.count %] = ([% i %] and [% j %])
        [%~ END %]
    [%~ END %]

Would print:

           1-1 = (a and X)
           1-2 = (a and Y)
           2-1 = (b and X)
           2-2 = (b and Y)

FOREACH may also be used as a post operative directive.

    [% "$i" FOREACH i = [1 .. 5] %] => 12345

=item C<GET>

Return the value of a variable or expression.

    [% GET a %]

The GET keyword may be omitted.

    [% a %]

    [% 7 + 2 - 3 %] => 6

See the section on VARIABLES.

=item C<IF (IF / ELSIF / ELSE)>

Allows for conditional testing.  Expects an expression as its only
argument.  If the expression is true, the contents of its block are
processed.  If false, the processor looks for an ELSIF block.  If an
ELSIF's expression is true then it is processed.  Finally it looks for
an ELSE block which is processed if none of the IF or ELSIF's
expressions were true.

    [% IF a == b %]A equaled B[% END %]

    [% IF a == b -%]
        A equaled B
    [%- ELSIF a == c -%]
        A equaled C
    [%- ELSE -%]
        Couldn't determine that A equaled anything.
    [%- END %]

IF may also be used as a post operative directive.

    [% 'A equaled B' IF a == b %]

Note: If you are using HTML::Template style documents, the TMPL_IF
tag parses using the limited HTML::Template parsing rules.  However,
you may use EXPR="" to embed a TT3 style expression.

=item C<INCLUDE>

Parse the contents of a file or block and insert them.  Variables defined
or modifications made to existing variables are discarded after
a template is included.

    [% INCLUDE path/to/template.html %]

    [% INCLUDE "path/to/template.html" %]

    [% file = "path/to/template.html" %]
    [% INCLUDE $file %]

    [% BLOCK foo %]This is foo[% END %]
    [% INCLUDE foo %]

Arguments may also be passed to the template:

    [% INCLUDE "path/to/template.html" a = "An arg" b = "Another arg" %]

Filenames must be relative to INCLUDE_PATH unless the ABSOLUTE
or RELATIVE configuration items are set.

Multiple filenames can be passed by separating them with a plus, a space,
or commas (TT2 doesn't support the comma).  Any supplied arguments will
be used on all templates.

    [% INCLUDE "path/to/template.html",
               "path/to/template2.html" a = "An arg" b = "Another arg" %]

On Perl 5.6 on some platforms there may be some issues with the variable
localization.  There is no problem on 5.8 and greater.

=item C<INSERT>

Insert the contents of a file without template parsing.

Filenames must be relative to INCLUDE_PATH unless the ABSOLUTE
or RELATIVE configuration items are set.

Multiple filenames can be passed by separating them with a plus, a space,
or commas (TT2 doesn't support the comma).

    [% INSERT "path/to/template.html",
              "path/to/template2.html" %]

=item C<JS>

Only available if the COMPILE_JS configuration item is true (default is false).
This requires the L<Template::Alloy::JS> module to be installed.

Allow eval'ing the block of text as javascript.  The block will be parsed and then eval'ed.

    [% a = "BimBam" %]
    [%~ JS %]
        write('The variable a was "' + get('a') + '"');
        set('b', "FooBar");
    [% END %]
    [% b %]

Would print:

    The variable a was "BimBam"
    FooBar

=item C<LAST>

Used to exit out of a WHILE or FOREACH loop.

=item C<LOOP>

This directive operates similar to the HTML::Template loop directive.
The LOOP directive expects a single variable name.  This variable name
should point to an arrayref of hashrefs.  The keys of each hashref
will be added to the variable stash when it is iterated.

    [% var a = [{b => 1}, {b => 2}, {b => 3}] %]

    [% LOOP a %] ([% b %]) [% END %]

Would print:

     (1)  (2)  (3) 

If Alloy is in HT mode and GLOBAL_VARS is false, the contents of
the hashref will be the only items available during the loop iteration.

If LOOP_CONTEXT_VARS is true, and $QR_PRIVATE is false (default when
called through the output method), then the variables __first__, __last__,
 __inner__, __odd__, and __counter__ will be set.  See the HTML::Template
loop_context_vars configuration item for more information.

=item C<MACRO>

Takes a directive and turns it into a variable that can take arguments.

    [% MACRO foo(i, j) BLOCK %]You passed me [% i %] and [% j %].[% END %]

    [%~ foo("a", "b") %]
    [% foo(1, 2) %]

Would print:

    You passed me a and b.
    You passed me 1 and 2.

Another example:

    [% MACRO bar(max) FOREACH i = [1 .. max] %]([% i %])[% END %]

    [%~ bar(4) %]

Would print:

    (1)(2)(3)(4)

Starting with version 1.012 of Template::Alloy there is also
a macro operator.

    [% foo = ->(i,j){ "You passed me $i and $j" } %]

    [% bar = ->(max){ FOREACH i = [1 .. max]; i ; END } %]

See the Template::Alloy::Operator documentation for more examples.

=item C<META>

Used to define variables that will be available via either the
template or component namespace.

Once defined, they cannot be overwritten.

    [% template.foobar %]
    [%~ META foobar = 'baz' %]
    [%~ META foobar = 'bing' %]

Would print:

    baz

=item C<NEXT>

Used to go to the next iteration of a WHILE or FOREACH loop.

=item C<PERL>

Only available if the EVAL_PERL configuration item is true (default is false).

Allow eval'ing the block of text as perl.  The block will be parsed and then eval'ed.

    [% a = "BimBam" %]
    [%~ PERL %]
        my $a = "[% a %]";
        print "The variable \$a was \"$a\"";
        $stash->set('b', "FooBar");
    [% END %]
    [% b %]

Would print:

    The variable $a was "BimBam"
    FooBar

During execution, anything printed to STDOUT will be inserted into the template.  Also,
the $stash and $context variables are set and are references to objects that mimic the
interface provided by Template::Context and Template::Stash.  These are provided for
compatibility only.  $self contains the current Template::Alloy object.

=item C<PROCESS>

Parse the contents of a file or block and insert them.  Unlike INCLUDE,
no variable localization happens so variables defined or modifications made
to existing variables remain after the template is processed.

    [% PROCESS path/to/template.html %]

    [% PROCESS "path/to/template.html" %]

    [% file = "path/to/template.html" %]
    [% PROCESS $file %]

    [% BLOCK foo %]This is foo[% END %]
    [% PROCESS foo %]

Arguments may also be passed to the template:

    [% PROCESS "path/to/template.html" a = "An arg" b = "Another arg" %]

Filenames must be relative to INCLUDE_PATH unless the ABSOLUTE
or RELATIVE configuration items are set.

Multiple filenames can be passed by separating them with a plus, a space,
or commas (TT2 doesn't support the comma).  Any supplied arguments will
be used on all templates.

    [% PROCESS "path/to/template.html",
               "path/to/template2.html" a = "An arg" b = "Another arg" %]

=item C<RAWPERL>

Only available if the EVAL_PERL configuration item is true (default is false).
Similar to the PERL directive, but you will need to append
to the $output variable rather than just calling PRINT.

=item C<RETURN>

Used to exit the innermost block or template and continue processing
in the surrounding block or template.

There are two changes from TT2 behavior.  First, In Alloy, a RETURN during
a MACRO call will only exit the MACRO.  Second, the RETURN directive
takes an optional variable name or expression, if passed, the MACRO will
return this value instead of the normal text from the MACRO.  The process_simple
method will also return this value.

You can also use the item, list, and hash return vmethods.

    [% RETURN %]       # just exits
    [% RETURN "foo" %] # return value is foo
    [% "foo".return %] # same thing

=item C<SET>

Used to set variables.

   [% SET a = 1 %][% a %]             => "1"
   [% a = 1 %][% a %]                 => "1"
   [% b = 1 %][% SET a = b %][% a %]  => "1"
   [% a = 1 %][% SET a %][% a %]      => ""
   [% SET a = [1, 2, 3] %][% a.1 %]   => "2"
   [% SET a = {b => 'c'} %][% a.b %]  => "c"

=item C<STOP>

Used to exit the entire process method (out of all blocks and templates).
No content will be processed beyond this point.

=item C<SWITCH>

Allow for SWITCH and CASE functionality.

   [% a = "hi" %]
   [% b = "bar" %]
   [% SWITCH a %]
       [% CASE "foo"           %]a was foo
       [% CASE b               %]a was bar
       [% CASE ["hi", "hello"] %]You said hi or hello
       [% CASE DEFAULT         %]I don't know what you said
   [% END %]

Would print:

   You said hi or hello

=item C<TAGS>

Change the type of enclosing braces used to delineate template tags.  This
remains in effect until the end of the enclosing block or template or until
the next TAGS directive.  Either a named set of tags must be supplied, or
two tags themselves must be supplied.

    [% TAGS html %]

    [% TAGS <!-- --> %]

The named tags are (duplicated from TT):

    asp       => ['<%',     '%>'    ], # ASP
    default   => ['\[%',    '%\]'   ], # default
    html      => ['<!--',   '-->'   ], # HTML comments
    mason     => ['<%',     '>'     ], # HTML::Mason
    metatext  => ['%%',     '%%'    ], # Text::MetaText
    php       => ['<\?',    '\?>'   ], # PHP
    star      => ['\[\*',   '\*\]'  ], # TT alternate
    template  => ['\[%',    '%\]'   ], # Normal Template Toolkit
    template1 => ['[\[%]%', '%[%\]]'], # allow TT1 style
    tt2       => ['\[%',    '%\]'   ], # TT2

If custom tags are supplied, by default they are escaped using
quotemeta.  You may also pass explicitly quoted strings,
or regular expressions as arguments as well (if your
regex begins with a ', ", or / you must quote it.

    [% TAGS [<] [>] %]          matches "[<] tag [>]"

    [% TAGS '[<]' '[>]' %]      matches "[<] tag [>]"

    [% TAGS "[<]" "[>]" %]      matches "[<] tag [>]"

    [% TAGS /[<]/ /[>]/ %]      matches "< tag >"

    [% TAGS ** ** %]            matches "** tag **"

    [% TAGS /**/ /**/ %]        Throws an exception.

You should be sure that the start tag does not include grouping
parens or INTERPOLATE will not function properly.

=item C<THROW>

Allows for throwing an exception.  If the exception is not caught
via the TRY DIRECTIVE, the template will abort processing of the directive.

    [% THROW mytypes.sometime 'Something happened' arg1 => val1 %]

See the TRY directive for examples of usage.

=item C<TRY>

The TRY block directive will catch exceptions that are thrown
while processing its block (It cannot catch parse errors unless
they are in included files or evaltt'ed strings.   The TRY block
will then look for a CATCH block that will be processed.  While
it is being processed, the "error" variable will be set with the thrown
exception as the value.  After the TRY block - the FINAL
block will be ran whether or not an error was thrown (unless a CATCH
block throws an error).

Note: Parse errors cannot be caught unless they are in an eval FILTER, or are
in a separate template being INCLUDEd or PROCESSed.

    [% TRY %]
    Nothing bad happened.
    [% CATCH %]
    Caught the error.
    [% FINAL %]
    This section runs no matter what happens.
    [% END %]

Would print:

    Nothing bad happened.
    This section runs no matter what happens.

Another example:

    [% TRY %]
    [% THROW "Something happened" %]
    [% CATCH %]
      Error:               [% error %]
      Error.type:          [% error.type %]
      Error.info:          [% error.info %]
    [% FINAL %]
      This section runs no matter what happens.
    [% END %]

Would print:

      Error:               undef error - Something happened
      Error.type:          undef
      Error.info:          Something happened
      This section runs no matter what happens.

You can give the error a type and more information including named arguments.
This information replaces the "info" property of the exception.

    [% TRY %]
    [% THROW foo.bar "Something happened" "grrrr" foo => 'bar' %]
    [% CATCH %]
      Error:               [% error %]
      Error.type:          [% error.type %]
      Error.info:          [% error.info %]
      Error.info.0:        [% error.info.0 %]
      Error.info.1:        [% error.info.1 %]
      Error.info.args.0:   [% error.info.args.0 %]
      Error.info.foo:      [% error.info.foo %]
    [% END %]

Would print something like:

      Error:               foo.bar error - HASH(0x82a395c)
      Error.type:          foo.bar
      Error.info:          HASH(0x82a395c)
      Error.info.0:        Something happened
      Error.info.1:        grrrr
      Error.info.args.0:   Something happened
      Error.info.foo:      bar

You can also give the CATCH block a type to catch.  And you
can nest TRY blocks.  If types are specified, Alloy will try and
find the closest matching type.  Also, an error object can
be re-thrown using $error as the argument to THROW.

    [% TRY %]
      [% TRY %]
        [% THROW foo.bar "Something happened" %]
      [% CATCH bar %]
        Caught bar.
      [% CATCH DEFAULT %]
        Caught default - but re-threw.
        [% THROW $error %]
      [% END %]
    [% CATCH foo %]
      Caught foo.
    [% CATCH foo.bar %]
      Caught foo.bar.
    [% CATCH %]
      Caught anything else.
    [% END %]

Would print:

        Caught default - but re-threw.

      Caught foo.bar.

=item C<UNLESS>

Same as IF but condition is negated.

    [% UNLESS 0 %]hi[% END %]  => hi

Can also be a post operative directive.

=item C<USE>

Allows for loading a Template::Toolkit style plugin.

    [% USE iter = Iterator(['foo', 'bar']) %]
    [%~ iter.get_first %]
    [% iter.size %]

Would print:

    foo
    2

Note that it is possible to send arguments to the new object
constructor.  It is also possible to omit the variable name being
assigned.  In that case the name of the plugin becomes the variable.

    [% USE Iterator(['foo', 'bar', 'baz']) %]
    [%~ Iterator.get_first %]
    [% Iterator.size %]

Would print:

    foo
    3

Plugins that are loaded are looked up for in the namespace listed in
the PLUGIN_BASE directive which defaults to Template::Plugin.  So in
the previous example, if Template::Toolkit was installed, the iter
object would loaded by the class Template::Plugin::Iterator.  In Alloy,
an effective way to disable plugins is to set the PLUGIN_BASE to a
non-existent base such as "_" (In TT it will still fall back to look
in Template::Plugin).

Note: The iterator plugin will fall back and use
Template::Alloy::Iterator if Template::Toolkit is not installed.  No
other plugins come installed with Template::Alloy.

The names of the Plugin being loaded from PLUGIN_BASE are case
insensitive.  However, using case insensitive names is bad as it
requires scanning the @INC directories for any module matching the
PLUGIN_BASE and caching the result (OK - not that bad).

If the plugin is not found and the LOAD_PERL directive is set, then
Alloy will try and load a module by that name (note: this type of lookup
is case sensitive and will not scan the @INC dirs for a matching
file).

    # The LOAD_PERL directive should be set to 1
    [% USE ta = Template::Alloy %]
    [%~ ta.dump_parse_expr('2 * 3') %]

Would print:

    [[undef, '*', 2, 3], 0];

See the PLUGIN_BASE, and PLUGINS configuration items.

See the documentation for Template::Manual::Plugins.

=item C<VIEW>

Implement a TT style view.  For more information, please
see the Template::View documentation.  This DIRECTIVE
will correctly parse the arguments and then pass them
along to a newly created Template::View object.  It
will fail if Template::View can not be found.

=item C<WHILE>

Will process a block of code while a condition is true.

    [% WHILE i < 3 %]
        [%~ i = i + 1 %]
        i = [% i %]
    [%~ END %]

Would print:

        i = 1
        i = 2
        i = 3

You could also do:

    [% i = 4 %]
    [% WHILE (i = i - 1) %]
        i = [% i %]
    [%~ END %]

Would print:

        i = 3
        i = 2
        i = 1

Note that (f = f - 1) is a valid expression that returns the value
of the assignment.  The parenthesis are not optional.

WHILE has a built in limit of 1000 iterations.  This is controlled by the
global variable $WHILE_MAX in Template::Alloy.

WHILE may also be used as a post operative directive.

    [% "$i" WHILE (i = i + 1) < 7 %] => 123456

=item C<WRAPPER>

Block directive.  Processes contents of its block and then passes them
in the [% content %] variable to the block or filename listed in the
WRAPPER tag.

    [% WRAPPER foo b = 23 %]
    My content to be processed ([% b %]).[% a = 2 %]
    [% END %]

    [% BLOCK foo %]
    A header ([% a %]).
    [% content %]
    A footer ([% a %]).
    [% END %]

This would print.

    A header (2).
    My content to be processed (23).
    A footer (2).

The WRAPPER directive may also be used as a post operative directive.

    [% BLOCK baz %]([% content %])[% END -%]
    [% "foobar" WRAPPER baz %]

Would print

    (foobar)');

Multiple filenames can be passed by separating them with a plus, a space,
or commas (TT2 doesn't support the comma).  Any supplied arguments will
be used on all templates.  Wrappers are processed in reverse order, so
that the first wrapper listed will surround each subsequent wrapper listed.
Variables from inner wrappers are available to the next wrapper that
surrounds it.

    [% WRAPPER "path/to/outer.html",
               "path/to/inner.html" a = "An arg" b = "Another arg" %]


=back

=head1 DIRECTIVES (HTML::Template Style)

HTML::Template templates use directives that look similar to the
following:

    <TMPL_VAR NAME="foo">

    <TMPL_IF NAME="bar">
      BAR
    </TMPL_IF>

The normal set of HTML::Template directives are TMPL_VAR,
TMPL_IF, TMPL_ELSE, TMPL_UNLESS, TMPL_INCLUDE, and TMPL_LOOP.
These tags should have either a NAME attribute, an EXPR attribute,
or a bare variable name that is used to specify the value to
be operated.  If a NAME is specified, it may only be a single
level value (as opposed to a TT chained variable).  In the case
of the TMPL_INCLUDE directive, the NAME is the file to be included.

In Alloy, the EXPR attribute can be used with any of these types to
specify TT compatible variable or expression that will be used for
the value.

    <TMPL_VAR NAME="foo">          Prints the value contained in foo
    <TMPL_VAR foo>                 Prints the value contained in foo
    <TMPL_VAR EXPR="foo">          Prints the value contained in foo

    <TMPL_VAR NAME="foo.bar.baz">  Prints the value contained in {'foo.bar.baz'}
    <TMPL_VAR EXPR="foo.bar.baz">  Prints the value contained in {foo}->{bar}->{baz}

    <TMPL_IF foo>                  Prints FOO if foo is true
      FOO
    </TMPL_IF

    <TMPL_UNLESS foo>              Prints FOO unless foo is true
      FOO
    </TMPL_UNLESS

    <TMPL_INCLUDE NAME="foo.ht">   Includes the template in "foo.ht"

    <TMPL_LOOP foo>                Iterates on the arrayref foo
      <TMPL_VAR name>
    </TMPL_LOOP>

Template::Alloy makes all of the other TT3 directives available
in addition to the normal set of HTML::Template directives.  For
example, the following is valid in Alloy.

    <TMPL_MACRO bar(n) BLOCK>You said <TMPL_VAR n></TMPL_MACRO>
    <TMPL_GET bar("hello")>

The TMPL_VAR tag may also include an optional ESCAPE attribute.
This specifies how the value of the tag should be escaped prior
to substituting into the template.

    Escape value |   Type of escape
    ---------------------------------
    HTML, 1      |   HTML encoding
    URL          |   URL encoding
    JS           |   basic javascript encoding (\n, \r, and \")
    NONE, 0      |   No encoding (default).

The TMPL_VAR tag may also include an optional DEFAULT attribute
that contains a string that will be used if the variable returns
false.

    <TMPL_VAR foo DEFAULT="Foo was false">

=head1 CHOMPING

Chomping refers to the handling of whitespace immediately before and
immediately after template tags.  By default, nothing happens to this
whitespace.  Modifiers can be placed just inside the opening and just
before the closing tags to control this behavior.

Additionally, the PRE_CHOMP and POST_CHOMP configuration variables can
be set and will globally control all chomping behavior for tags that
do not have their own chomp modifier.  PRE_CHOMP and POST_CHOMP can
be set to any of the following values:

    none:      0   +   Template::Constants::CHOMP_NONE
    one:       1   -   Template::Constants::CHOMP_ONE
    collapse:  2   =   Template::Constants::CHOMP_COLLAPSE
    greedy:    3   ~   Template::Constants::CHOMP_GREEDY

=over 4

=item CHOMP_NONE

Don't do any chomping.  The "+" sign is used to indicate CHOMP_NONE.

    Hello.

    [%+ "Hi." +%]

    Howdy.

Would print:

    Hello.

    Hi.

    Howdy.

=item CHOMP_ONE (formerly known as CHOMP_ALL)

Delete any whitespace up to the adjacent newline.  The "-" is used to indicate CHOMP_ONE.

    Hello.

    [%- "Hi." -%]

    Howdy.

Would print:

    Hello.
    Hi.
    Howdy.

=item CHOMP_COLLAPSE

Collapse adjacent whitespace to a single space.  The "=" is used to indicate CHOMP_COLLAPSE.

    Hello.

    [%= "Hi." =%]

    Howdy.

Would print:

    Hello. Hi. Howdy.

=item CHOMP_GREEDY

Remove all adjacent whitespace.  The "~" is used to indicate CHOMP_GREEDY.

    Hello.

    [%~ "Hi." ~%]

    Howdy.

Would print:

    Hello.Hi.Howdy.

=back

=head1 CONFIGURATION

The following configuration variables are supported (in
alphabetical order).  Note: for further discussion you can refer to
the TT config documentation.

Items may be passed in upper or lower case.  If lower case names
are passed they will be resolved to uppercase during the "new"
method.

All of the variables in this section can be passed to the "new" constructor.

    my $obj = Template::Alloy->new(
        VARIABLES  => \%hash_of_variables,
        AUTO_RESET => 0,
        TRIM       => 1,
        POST_CHOMP => "=",
        PRE_CHOMP  => "-",
    );

=over 4

=item ABSOLUTE

Boolean.  Default false.  Are absolute paths allowed for included files.

=item ADD_LOCAL_PATH

If true, allows calls include_filename to temporarily add the
directory of the current template being processed to the INCLUDE_PATHS
arrayref.  This allows templates to refer to files in the local
template directory without specifying the local directory as part of
the filename.  Default is 0.  If set to a negative value, the current
directory will be added to the end of the current INCLUDE_PATHS.

This property may also be set in the template using the CONFIG directive.

    [% CONFIG ADD_LOCAL_PATH => 1 %]

=item ANYCASE

Allow directive matching to be case insensitive.

    [% get 23 %] prints 23 with ANYCASE => 1

=item AUTO_RESET

Boolean.  Default 1.  Clear blocks that were set during the process method.

=item AUTO_EVAL

Boolean.  Default 0 (default 1 in Velocity syntax).  If set to true,
double quoted strings will automatically be passed to the eval filter.
This configuration option may also be passed to the CONFIG directive.

=item AUTO_FILTER

Can be the name of any filter.  Default undef.  Any variable returned by a
GET directive (including implicit GET) will be passed to the named filter.
This configuration option may also be passed to the CONFIG directive.

    # with AUTO_FILTER => 'html'

    [% f = "&"; GET f %] prints &amp;
    [% f = "&"; f %]     prints &amp; (implicit GET)

If a variable already has another filter applied the AUTO_FILTER is not applied.
The "none" scalar virtual method has been added to allow for using variables
without reapplying filters.

    # with AUTO_FILTER => 'html'

    [% f = "&";  f | none %] prints &
    [% f = "&"; g = f; g %]  prints &amp;
    [% f = "&"; g = f; g | none %]  prints & (because g = f is a SET directive)
    [% f = "&"; g = GET f; g | none %]  prints &amp; (because the actual GET directive was called)

=item BLOCKS

Only available via when using the process interface.

A hashref of blocks that can be used by the process method.

    BLOCKS => {
        block_1 => sub { ... }, # coderef that returns a block
        block_2 => 'A String',  # simple string
    },

Note that a Template::Document cannot be supplied as a value (TT
supports this).  However, it is possible to supply a value that is
equal to the hashref returned by the load_template method.

=item CACHE_SIZE

Number of compiled templates to keep in memory.  Default undef.
Undefined means to allow all templates to cache.  A value of 0 will
force no caching.  The cache mechanism will clear templates that have
not been used recently.

=item CACHE_STR_REFS

Default 1.  If set, any string refs will have an MD5 sum taken that
will then be used for caching the document - both in memory and on
the file system (if configured).  This will give a significant speed
boost.  Note that this affects strings passed to the EVALUATE directive
or eval filters as well.  It may be set using the CONFIG directive.

=item CALL_CONTEXT (Not in TT)

Can be one of 'item', 'list', or 'smart'.  The default type is
'smart'.  The CALL_CONTEXT configuration specifies in what Perl
context coderefs and methods used in the processed templates will be
called.  TT historically has avoided the distinction of item (scalar)
vs list context.  To avoid worrying about this, TT introduced 'smart'
context.  The C<@()> and C<$()> context specifiers make it easier to use
CALL_CONTEXT in some situations.

The following table shows the relationship between the various contexts:

       return values      smart context   list context    item context
       -------------      -------------   ------------    ------------
    A   'foo'              'foo'           ['foo']         'foo'
    B   undef              undef           [undef]         undef
    C   (no return value)  undef           []              undef
    D   (7)                7               [7]             7
    E   (7,8,9)            [7,8,9]         [7,8,9]         9
    F   @a = (7)           7               [7]             1
    G   @a = (7,8,9)       [7,8,9]         [7,8,9]         3
    H   ({b=>"c"})         {b=>"c"}        [{b=>"c"}]      {b=>"c"}
    I   ([1])              [1]             [[1]]           [1]
    J   ([1],[2])          [[1],[2]]       [[1],[2]]       [2]
    K   [7,8,9]            [7,8,9]         [[7,8,9]]       [7,8,9]
    L   (undef, "foo")     die "foo"       [undef, "foo"]  "foo"
    M   wantarray?1:0      1               [1]             0

Cases F, H, I and M are common sticking points of the smart context
in TT2.  Note that list context always returns an arrayref from a method
or function call.  Smart context can give confusing results sometimes,
especially the I and J cases.  Case L for smart match is very surprising.

The list and item context provide another feature for method calls.  In
smart context, TT will look for a hash key in the object by the same name
as the method, if a method by that name doesn't exist.  In item and list
context Alloy will die if a method by that name cannot be found.

The CALL_CONTEXT configuration item can be passed to new or it may
also be set during runtime using the CONFIG directive.  The following
method call would be in list context:

    [% CONFIG CALL_CONTEXT => 'list';
       results = my_obj.get_results;
       CONFIG CALL_CONTEXT => 'smart'
    %]

Note that we needed to restore CALL_CONTEXT to the default 'smart' value.
Template::Alloy has added the C<@()> (list) and the C<$()> (item) context
specifiers.  The previous example could be written as:

    [% results = @( my_obj.get_results ) %]

To call that same method in item (scalar) context you would do the following:

    [% results = $( my_obj.get_results ) %]

The C<@()> and C<$()> operators are based on the Perl 6 counterpart.

=item COMPILE_DIR

Base directory to store compiled templates.  Default undef. Compiled
templates will only be stored if one of COMPILE_DIR and COMPILE_EXT is
set.

If set, the AST of parsed documents will be cached.  If COMPILE_PERL is
set, the compiled perl code will also be stored.

=item COMPILE_EXT

Extension to add to stored compiled template filenames.  Default undef.

If set, the AST of parsed documents will be cached.  If COMPILE_PERL is
set, the compiled perl code will also be stored.

=item COMPILE_JS

Default false.

Requires installation of L<Template::Alloy::JS>.  When enabled, the parsed
templates will be translated into Javascript and executed using the
V8 javascript engine.  If compile_dir is also set, this compiled javascript
will be cached to disk.

If your templates are short, there is little benefit to using this other
than you can then use the JS directive.  If your templates are long or
you are running in a cached environment, this will speed up your templates.

Certain limitations exist when COMPILE_JS is set, most notably the USE
and VIEW directives are not supported, and method calls on objects
passed to the template do not work (code refs passed in do work
however).  These limitations are due to the nature of JavaScript::V8
bind and Perl/JavaScript OO differences.

=item COMPILE_PERL

Default false.

If set to 1 or 2, will translate the normal AST into a perl 5 code document.
This document can then be executed directly, cached in memory, or cached
on the file system depending upon the configuration items set.

If set to 1, a perl code document will always be generated.

If set to 2, a perl code document will only be generated if an AST has
already been cached for the document.  This should give a speed benefit and
avoid extra compilation unless the document has been used more than once.

If Alloy is running in a cached environment such as mod_perl, then using
compile_perl can offer some speed benefit and makes Alloy faster than Text::Tmpl
and as fast as HTML::Template::Compiled (but Alloy has more features).

If you are not running in a cached environment, such as from commandline, or from
CGI, it is generally faster to only run from the AST (with COMPILE_PERL => 0).

=item CONSTANTS

Hashref.  Used to define variables that will be "folded" into the
compiled template.  Variables defined here cannot be overridden.

    CONSTANTS => {my_constant => 42},

    A template containing:

    [% constants.my_constant %]

    Will have the value 42 compiled in.

Constants defined in this way can be chained as in [%
constant.foo.bar.baz %].

=item CONSTANT_NAMESPACE

Allow for setting the top level of values passed in CONSTANTS.  Default
value is 'constants'.

=item DEBUG

Takes a list of constants |'ed together which enables different
debugging modes.  Alternately the lowercase names may be used
(multiple values joined by a ",").

    The only supported TT values are:
    DEBUG_UNDEF (2)    - debug when an undefined value is used (now easier to use STRICT)
    DEBUG_DIRS  (8)    - debug when a directive is used.
    DEBUG_ALL   (2047) - turn on all debugging.

    Either of the following would turn on undef and directive debugging:

    DEBUG => 'undef, dirs',            # preferred
    DEBUG => 2 | 8,
    DEBUG => DEBUG_UNDEF | DEBUG_DIRS, # constants from Template::Constants

=item DEBUG_FORMAT

Change the format of messages inserted when DEBUG has DEBUG_DIRS set on.
This essentially the same thing as setting the format using the DEBUG
directive.

=item DEFAULT

The name of a default template file to use if the passed one is not found.

=item DELIMITER

String to use to split INCLUDE_PATH with.  Default is :.  It is more
straight forward to just send INCLUDE_PATH an arrayref of paths.

=item DUMP

Configures the behavior of the DUMP tag.  May be set to 0, a hashref,
or another true value.  Default is true.

If set to 0, all DUMP directives will do nothing.  This is useful if
you would like to turn off the DUMP directives under some environments.

IF set to a true value (or undefined) then DUMP directives will operate.

If set to a hashref, the values of the hash can be used to configure
the operation of the DUMP directives.  The following are the values
that can be set in this hash.

=over 4

=item EntireStash

Default 1.  If set to 0, then the DUMP directive will not print the
entire contents of the stash when a DUMP directive is called without
arguments.

=item handler

Defaults to an internal coderef.  If set to a coderef, the DUMP directive will pass the
arguments to be dumped and expects a string with the dumped data.  This
gives complete control over the dump process.

Note 1: The default handler makes sure that values matching the
private variable regex are not included.  If you install your own handler,
you will need to take care of these variables if you intend for them
to not be shown.

Note 2: If you would like the name of the variable to be dumped, include
the string '$VAR1' and the DUMP directive will interpolate the value.  For
example, to dump all output as YAML - you could do the following:

    DUMP => {
       handler => sub {
           require YAML;
           return "\$VAR1 =\n".YAML::Dump(shift);
       },
    }

=item header

Default 1.  Controls whether a header is printed for each DUMP directive.
The header contains the file and line number the DUMP directive was
called from.  If set to 0 the headers are disabled.

=item html

Defaults to 1 if $ENV{'REQUEST_METHOD'} is set - 0 otherwise.  If set to
1, then the output of the DUMP directive is passed to the html filter
and encased in "pre" tags.  If set to 0 no html encoding takes place.

=item Sortkeys, Useqq, Ident, Pad, etc

Any of the Data::Dumper configuration items may be passed.

=back

=item ENCODING

Default undef.  If set, and if Perl version is greater than or equal to
5.7.3 (when Encode.pm was first included), then Encode::decode will
be called every time a template file is processed and will be passed
the value of ENCODING and text from the template.

This item can also be set using [% CONFIG ENCODING => encoding %] before
calling INCLUDE or PROCESS directives to change encodings on the fly.

=item END_TAG

Set a string to use as the closing delimiter for TT.  Default is "%]".

=item ERROR

Used as a fall back when the processing of a template fails.  May either
be a single filename that will be used in all cases, or may be a hashref
of options where the keynames represent error types that will be handled
by the filename in their value.  A key named default will be used if no
other matching keyname can be found.  The selection process is similar
to that of the TRY/CATCH/THROW directives (see those directives for more
information).

    my $t = Template::Alloy->new({
        ERROR => 'general/catch_all_errors.html',
    });

    my $t = Template::Alloy->new({
        ERROR => {
            default   => 'general/catch_all_errors.html',
            foo       => 'catch_all_general_foo_errors.html',
            'foo.bar' => 'catch_foo_bar_errors.html',
        },
    });

Note that the ERROR handler will only be used for errors during the
processing of the main document.  It will not catch errors that
occur in templates found in the PRE_PROCESS, POST_PROCESS, and WRAPPER
configuration items.

=item ERRORS

Same as the ERROR configuration item.  Both may be used interchangeably.

=item EVAL_PERL

Boolean.  Default false.  If set to a true value, PERL and RAWPERL blocks
will be allowed to run.  This is a potential security hole, as arbitrary
perl can be included in the template.  If Template::Toolkit is installed,
a true EVAL_PERL value also allows the perl and evalperl filters to be used.

=item FILTERS

Allow for passing in TT style filters.

    my $filters = {
        filter1 =>  sub { my $str = shift; $s =~ s/./1/gs; $s },
        filter2 => [sub { my $str = shift; $s =~ s/./2/gs; $s }, 0],
        filter3 => [sub { my ($context, @args) = @_; return sub { my $s = shift; $s =~ s/./3/gs; $s } }, 1],
    };

    my $str = q{
        [% a = "Hello" %]
        1 ([% a | filter1 %])
        2 ([% a | filter2 %])
        3 ([% a | filter3 %])
    };

    my $obj = Template::Alloy->new(FILTERS => $filters);
    $obj->process(\$str) || die $obj->error;

Would print:

        1 (11111)
        2 (22222)
        3 (33333)

Filters passed in as an arrayref should contain a coderef and a value
indicating if they are dynamic or static (true meaning dynamic).  The
dynamic filters are passed the pseudo context object and any arguments
and should return a coderef that will be called as the filter.  The filter
coderef is then passed the string.

=item GLOBAL_CACHE

Default 0.  If true, documents will be cached in $Template::Alloy::GLOBAL_CACHE.
It may also be passed a hashref, in which case the documents will be
cached in the passed hashref.

The TT, Tmpl, and velocity will automatically cache documents in the object.  The
HTML::Template interface uses a new object each time.  Setting the HTML::Template's
CACHE configuration is the same as setting GLOBAL_CACHE.

=item INCLUDE_PATH

A string or an arrayref or coderef that returns an arrayref that
contains directories to look for files included by processed
templates.  Defaults to "." (the current directory).

=item INCLUDE_PATHS

Non-TT item.  Same as INCLUDE_PATH but only takes an arrayref.  If not specified
then INCLUDE_PATH is turned into an arrayref and stored in INCLUDE_PATHS.
Overrides INCLUDE_PATH.

=item INTERPOLATE

Boolean.  Specifies whether variables in text portions of the template will be
interpolated.  For example, the $variable and ${var.value} would be substituted
with the appropriate values from the variable cache (if INTERPOLATE is on).

    [% IF 1 %]The variable $variable had a value ${var.value}[% END %]

=item LOAD_PERL

Indicates if the USE directive can fall back and try and load a perl module
if the indicated module was not found in the PLUGIN_BASE path.  See the
USE directive.  This configuration has no bearing on the COMPILE_PERL
directive used to indicate using compiled perl documents.

=item MAX_EVAL_RECURSE (Alloy only)

Will use $Template::Alloy::MAX_EVAL_RECURSE if not present.  Default is 50.
Prevents runaway on the following:

    [% f = "[% f|eval %]" %][% f|eval %]

=item MAX_MACRO_RECURSE (Alloy only)

Will use $Template::Alloy::MAX_MACRO_RECURSE if not present.  Default is 50.
Prevents runaway on the following:

    [% MACRO f BLOCK %][% f %][% END %][% f %]

=item NAMESPACE

No Template::Namespace::Constants support.  Hashref of hashrefs representing
constants that will be folded into the template at compile time.

    Template::Alloy->new(NAMESPACE => {constants => {
         foo => 'bar',
    }});

Is the same as

    Template::Alloy->new(CONSTANTS => {
         foo => 'bar',
    });

Any number of hashes can be added to the NAMESPACE hash.

=item NEGATIVE_STAT_TTL (Not in TT)

Defaults to STAT_TTL which defaults to $STAT_TTL which defaults to 1.

Similar to STAT_TTL - but represents the time-to-live
seconds until a document that was not found is checked again against
the system for modifications.  Setting this number higher will allow for
fewer file system accesses.  Setting it to a negative number will allow
for the file system to be checked every hit.

=item NO_INCLUDES

Default false.  If true, calls to INCLUDE, PROCESS, WRAPPER and INSERT
will fail.  This option is also available when using the process method.

=item OUTPUT

Alternate way of passing in the output location for processed templates.
If process is not passed an output argument, it will look for this value.

See the process method for a listing of possible values.

=item OUTPUT_PATH

Base path for files written out via the process method or via the redirect
and file filters.  See the redirect virtual method and the process method
for more information.

=item PLUGINS

A hashref of mappings of plugin modules.

   PLUGINS => {
      Iterator => 'Template::Plugin::Iterator',
      DBI      => 'MyDBI',
   },

See the USE directive for more information.

=item PLUGIN_BASE

Default value is Template::Plugin.  The base module namespace
that template plugins will be looked for.  See the USE directive
for more information.  May be either a single namespace, or an arrayref
of namespaces.

=item POST_CHOMP

Set the type of chomping at the ending of a tag.
See the section on chomping for more information.

=item POST_PROCESS

Only available via when using the process interface.

A list of templates to be processed and appended to the content
after the main template.  During this processing the "template"
namespace will contain the name of the main file being processed.

This is useful for adding a global footer to all templates.

=item PRE_CHOMP

Set the type of chomping at the beginning of a tag.
See the section on chomping for more information.

=item PRE_DEFINE

Same as the VARIABLES configuration item.

=item PRE_PROCESS

Only available via when using the process interface.

A list of templates to be processed before and pre-pended to the content
before the main template.  During this processing the "template"
namespace will contain the name of the main file being processed.

This is useful for adding a global header to all templates.

=item PROCESS

Only available via when using the process interface.

Specify a file to use as the template rather than the one passed in
to the ->process method.

=item RECURSION

Boolean.  Default false.  Indicates that INCLUDED or PROCESSED files
can refer to each other in a circular manner.  Be careful about recursion.

=item RELATIVE

Boolean.  Default false.  If true, allows filenames to be specified
that are relative to the currently running process.

=item SEMICOLONS

Boolean.  Default false.  If true, then the syntax will require that
semi-colons separate multiple directives in the same tag.  This is
useful for keeping the syntax a little more clean as well as trouble
shooting some errors.

=item SHOW_UNDEFINED_INTERP (Not in TT)

Default false (default true in Velocity).  If INTERPOLATE is true,
interpolated dollar variables that return undef will be removed.  With
SHOW_UNDEFINED_INTERP set, undef values will leave the variable there.

    [% CONFIG INTERPOLATE => 1 %]
    [% SET foo = 1 %][% SET bar %]
    ($foo)($bar) ($!foo)($!bar)

Would print:

    (1)() (1)()

But the following:

    [% CONFIG INTERPOLATE => 1, SHOW_UNDEFINED_INTERP => 1 %]
    [% SET foo = 1 %][% SET bar %]
    ($foo)($bar) ($!foo)($!bar)

Would print:

    (1)($bar) (1)()

Note that you can use an exclamation point directly after the dollar
to make the variable silent.  This is similar to how Velocity works.

=item START_TAG

Set a string or regular expression to use as the opening delimiter
for TT.  Default is "[%".
You should be sure that the tag does not include grouping parens or
INTERPOLATE will not function properly.

=item STASH

Template::Alloy manages its own stash of variables.  You can pass a
Template::Stash or Template::Stash::XS object, but Template::Alloy
will copy all of values out of the object into its own stash.
Template::Alloy won't use any of the methods of the passed STASH
object.  The STASH option is only available when using the process
method.

=item STAT_TTL

Defaults to $STAT_TTL which defaults to 1.  Represents time-to-live
seconds until a cached in memory document is compared to the file
system for modifications.  Setting this number higher will allow for
fewer file system accesses.  Setting it to a negative number will allow
for the file system to be checked every hit.

=item STREAM

Defaults to false.  If set to true, generated template content will be
printed to the currently selected filehandle (default is STDOUT) as
soon as it is ready - there will be no buffering of the output.

The Stream role uses the Play role's directives (non-compiled_perl).

All directives and configuration work, except for the following
exceptions:

=over 4

=item CLEAR directive

Because the output is not buffered - the CLEAR directive would have no
effect.  The CLEAR directive will throw an error when STREAM is on.

=item TRIM configuration

Because the output is not buffered - trim operations cannot be played
on the output buffers.

=item WRAPPER configuration/directive

The WRAPPER configuration and directive items effectively turn off
STREAM since the WRAPPERS are generated in reverse order and because
the content is inserted into the middle of the WRAPPERS.  WRAPPERS
will still work, they just won't stream.

=item VARIOUS errors

Because the template is streaming, items that cause errors my result
in partially printed pages - since the error would occur part way
through the print.

=back

All output is printed directly to the currently selected filehandle
(defaults to STDOUT) via the CORE::print function.  Any output
parameter passed to process or process_simple will be ignored.

If you would like the output to go to another handle, you will need to
select that handle, process the template, and re-select STDOUT.

=item STRICT

Defaults to false.  If set to true, any undefined variable that is
encountered will cause the processing of the template to abort.  This
can be caught with a TRY block.  This can be useful for making sure
that the template only attempts to use variables that were correctly
initialized similar in spirit to Perl's "use strict."

When this occurs the strict_throw method is called.

See the STRICT_THROW configuration for additional options.

Similar functionality could be implemented using UNDEFINED_ANY.

The STRICT configuration item can be passed to new or it may also be
set during runtime using the CONFIG directive.  Once set though it
cannot be disabled for the duration of the current template and sub
components.  For example you could call [% CONFIG STRICT => 1 %] in
header.tt and strict mode would be enabled for the header.tt and any
sub templates processed by header.tt.

=item STRICT_THROW (not in TT)

Default undef.  Can be set to a subroutine which will be called when
STRICT is set and an undefined variable is processed.  It will be
passed the error type, error message, and a hashref of template
information containing the current component being processed, the
current outer template being processed, the identity reference for the
variable, and the stringified name of the identity.  This override can
be used for filtering allowable elements.

    my $ta = Template::Alloy->new({
        STRICT => 1,
        STRICT_THROW => sub {
            my ($ta, $err_type, $msg, $args) = @_;

            return if $args->{'component'} eq 'header.tt'
                      && $args->{'template'} eq 'main.html'
                      && $args->{'name'} eq 'foo.bar(1)'; # stringified identity name

            $ta->throw($err_type, $msg); # all other undefined variables die
        },
    });

=item SYNTAX (not in TT)

Defaults to "cet".  Indicates the syntax that will be used for parsing
included templates or eval'ed strings.  You can use the CONFIG
directive to change the SYNTAX on the fly (it will not affect
the syntax of the document currently being parsed).

The syntax may be passed in upper or lower case.

The available choices are:

    alloy - Template::Alloy style - the same as TT3
    tt3   - Template::Toolkit ver3 - same as Alloy
    tt2   - Template::Toolkit ver2 - almost the same as TT3
    tt1   - Template::Toolkit ver1 - almost the same as TT2
    ht    - HTML::Template - same as HTML::Template::Expr without EXPR
    hte   - HTML::Template::Expr
    js    - JavaScript style - requires compile_js to be set.
    jsr   - JavaScript Raw style - requires compile_js to be set.

Passing in a different syntax allows for the process method
to use a non-TT syntax and for the output method to use a non-HT
syntax.

The following is a sample of HTML::Template interface usage parsing
a Template::Toolkit style document.

    my $obj = Template::Alloy->new(filename => 'my/template.tt'
                                     syntax   => 'cet');
    $obj->param(\%swap);
    print $obj->output;

The following is a sample of Template::Toolkit interface usage parsing
a HTML::Template::Expr style document.

    my $obj = Template::Alloy->new(SYNTAX => 'hte');
    $obj->process('my/template.ht', \%swap);

You can use the define_syntax method to add another custom syntax to
the list of available options.

=item TAG_STYLE

Allow for setting the type of tag delimiters to use for parsing the TT.
See the TAGS directive for a listing of the available types.

=item TRIM

Remove leading and trailing whitespace from blocks and templates.
This operation is performed after all enclosed template tags have
been executed.

=item UNDEFINED_ANY

This is not a TT configuration option.  This option expects to be a code
ref that will be called if a variable is undefined during a call to play_expr.
It is passed the variable identity array as a single argument.  This
is most similar to the "undefined" method of Template::Stash.  It allows
for the "auto-defining" of a variable for use in the template.  It is
suggested that UNDEFINED_GET be used instead as UNDEFINED_ANY is a little
to general in defining variables.

You can also sub class the module and override the undefined_any method.

=item UNDEFINED_GET

This is not a TT configuration option.  This option expects to be a code
ref that will be called if a variable is undefined during a call to GET.
It is passed the variable identity array as a single argument.  This is more useful
than UNDEFINED_ANY in that it is only called during a GET directive
rather than in embedded expressions (such as [% a || b || c %]).

You can also sub class the module and override the undefined_get method.

=item V1DOLLAR

This allows for some compatibility with TT1 templates.  The only real
behavior change is that [% $foo %] becomes the same as [% foo %].  The
following is a basic table of changes invoked by using V1DOLLAR.

   With V1DOLLAR        Equivalent Without V1DOLLAR (Normal default)
   "[% foo %]"          "[% foo %]"
   "[% $foo %]"         "[% foo %]"
   "[% ${foo} %]"       "[% ${foo} %]"
   "[% foo.$bar %]"     "[% foo.bar %]"
   "[% ${foo.bar} %]"   "[% ${foo.bar} %]"
   "[% ${foo.$bar} %]"  "[% ${foo.bar} %]"
   "Text: $foo"         "Text: $foo"
   "Text: ${foo}"       "Text: ${foo}"
   "Text: ${$foo}"      "Text: ${foo}"

=item V2EQUALS

Default 1 in the TT syntax, defaults to 0 in the HTML::Template syntax.

If set to 1 then "==" is an alias for "eq" and "!= is an alias for
"ne".

    [% CONFIG V2EQUALS => 1 %][% ('7' == '7.0') || 0 %]
    [% CONFIG V2EQUALS => 0 %][% ('7' == '7.0') || 0 %]

    Prints

    0
    1

=item V2PIPE

Restores the behavior of the pipe operator to be compatible with TT2.

With V2PIPE = 1

    [%- BLOCK a %]b is [% b %]
    [% END %]
    [%- PROCESS a b => 237 | repeat(2) %]

    # output of block "a" with b set to 237 is passed to the repeat(2) filter

    b is 237
    b is 237

With V2PIPE = 0 (default)

    [%- BLOCK a %]b is [% b %]
    [% END %]
    [% PROCESS a b => 237 | repeat(2) %]

    # b set to 237 repeated twice, and b passed to block "a"

    b is 237237

=item VARIABLES

A hashref of variables to initialize the template stash with.  These
variables are available for use in any of the executed templates.
See the section on VARIABLES for the types of information that can be passed in.

=item VMETHOD_FUNCTIONS

Defaults to 1.  All scalar virtual methods are available as top level functions as well.
This is not true of TT2.  In Template::Alloy the following are equivalent:

    [% "abc".length %]
    [% length("abc") %]

You may set VMETHOD_FUNCTIONS to 0 to disable this behavior.

=item WRAPPER

Only available via when using the process interface.

Operates similar to the WRAPPER directive.  The option can be given a
single filename, or an arrayref of filenames that will be used to wrap
the processed content.  If an arrayref is passed the filenames are
processed in reverse order, so that the first filename specified will
end up being on the outside (surrounding all other wrappers).

   my $t = Template::Alloy->new(
       WRAPPER => ['my/wrappers/outer.html', 'my/wrappers/inner.html'],
   );

Content generated by the PRE_PROCESS and POST_PROCESS will come before
and after (respectively) the content generated by the WRAPPER
configuration item.

See the WRAPPER directive for more examples of how wrappers are constructed.

=back

=head1 CONFIGURATION (HTML::Template STYLE)

The following HTML::Template and HTML::Template::Expr
configuration variables are supported (in HTML::Template documentation order).
Note: for further discussion you can refer to the HT documentation.
Many of the variables mentioned in the TT CONFIGURATION section
apply here as well.  Unless noted, these items only apply when
using the output method.

Items may be passed in upper or lower case.  All passed items are resolved
to upper case.

These variables should be passed to the "new" constructor.

    my $obj = Template::Alloy->new(
        type   => 'filename',
        source => 'my/template.ht',
        die_on_bad_params => 1,
        loop_context_vars => 1,
        global_vars       => 1
        post_chomp => "=",
        pre_chomp  => "-",
    );

=over 4

=item TYPE

Can be one of filename, filehandle, arrayref, or scalarref.  Indicates what type
of input is in the "source" configuration item.

=item SOURCE

Stores where to read the input file.  The type is specified in the "type"
configuration item.

=item FILENAME

Indicates a filename to read the template from.  Same as putting the
filename in the "source" item and setting "type" to "filename".

Must be set to enable caching.

=item FILEHANDLE

Should contain an open filehandle to read the template from.  Same as
putting the filehandle in the "source" item and setting "type" to "filehandle".

Will not be cached.

=item ARRAYREF

Should contain an arrayref whose values are the lines of the template.  Same as
putting the arrayref in the "source" item and setting "type" to "arrayref".

Will not be cached.

=item SCALARREF

Should contain an reference to a scalar that contains the template.  Same as
putting the scalar ref in the "source" item and setting "type" to "scalarref".

Will not be cached.

=item CACHE

If set to one, then Alloy will use a global, in-memory document cache
to store compiled templates in between calls.  This is generally only
useful in a mod_perl environment.  The document is checked for a different
modification time at each request.

=item BLIND_CACHE

Same as with cache enabled, but will not check if the document has
been modified.

=item FILE_CACHE

If set to 1, will cache the compiled document on the file system.  If
true, file_cache_dir must be set.

=item FILE_CACHE_DIR

The directory where to store cached documents when file_cache is true.
This is similar to the TT compile_dir option.

=item DOUBLE_FILE_CACHE

Uses a combination of file_cache and cache.

=item PATH

Same as INCLUDE_PATH when using the process method.

=item ASSOCIATE

May be a single CGI object or an arrayref of objects.  The params
from these objects will be added to the params during the
output call.

=item CASE_SENSITIVE

Allow passed variables set through the param method, or the
associate configuration to be used case sensitively.  Default is
off.  It is highly suggested that this be set to 1.

=item LOOP_CONTEXT_VARS

Default false.  When true, calls to the loop directive will
create the following variables that give information about the
current iteration of the loop:

   __first__   - True on first iteration only
   __last__    - True on last iteration only
   __inner__   - True on any iteration that isn't first or last
   __odd__     - True on odd iterations
   __counter__ - The iteration count

These variables are also available to LOOPs run under
TT syntax if loop_context_vars is set and if QR_PRIVATE is set to 0.

=item GLOBAL_VARS.

Default true in HTE mode.  Default false in HT.  Allows top level
variables to be used in LOOPs.  When false, only variables defined
in the current LOOP iteration hashref will be available.

=item DEFAULT_ESCAPE

Controls the type of escape used on named variables in TMPL_VAR
directives.  Can be one of HTML, URL, or JS.  The values of
TMPL_VAR directives will be encoded with this type unless
they specify their own type via an ESCAPE attribute.

You may alternately use the AUTO_FILTER directive which can be
any of the item vmethod filters (you must use lower case when
specifying the AUTO_FILTER directive).  The AUTO_FILTER directive
will also be applied to TMPL_VAR EXPR and TMPL_GET items while
DEFAULT_ESCAPE only applies to TMPL_VAR NAME items.

=item NO_TT

Default false in 'hte' syntax.  Default true in 'ht' syntax.  If true,
no extended TT directives will be allowed.

The output method uses 'hte' syntax by default.

=back

=head1 SEMI PUBLIC METHODS

The following list of methods are other interesting methods of Alloy that
may be re-implemented by subclasses of Alloy.

=over 4

=item C<exception>

Creates an exception object blessed into the package listed in
Template::Alloy::Exception.

=item C<execute_tree>

Executes a parsed tree (returned from parse_tree)

=item C<play_expr>

Play the parsed expression.  Turns a variable identity array into the
parsed variable.  This method is also responsible for playing
operators and running virtual methods and filters.  The variable
identity array may also contain literal values, or operator identity
arrays.

=item C<include_filename>

Takes a file path, and resolves it into the full filename using
paths from INCLUDE_PATH or INCLUDE_PATHS.

=item C<_insert>

Resolves the file passed, and then returns its contents.

=item C<list_filters>

Dynamically loads the filters list from Template::Filters when a filter
is used that does not have a native implementation in Alloy.

=item C<load_template>

Given a filename or a string reference will return a "document" hashref
hash that contains the parsed tree.

    my $doc = $self->load_template($file); # errors die

This method handles the in-memory caching of the document.

=item C<load_tree>

Given the "document" hashref, will either load the parsed AST from
file (if configured to do so), or will load the content, parse the
content using the Parse role, and will return the tree.  File based
caching of the parsed AST happens here.

=item C<load_perl>

Only used if COMPILE_PERL is true (default is false).

Given the "document" hashref, will either load the compiled perl from
file (if configured to do so), or will load the AST using "load_tree",
will compile a new perl code document using the Compile role, and will
return the perl code.  File based caching of the compiled perl happens
here.

=item C<parse_tree>

Parses the passed string ref with the appropriate template syntax.

See L<Template::Alloy::Parse> for more details.

=item C<parse_expr>

Parses the passed string ref for a variable or expression.

See L<Template::Alloy::Parse> for more details.

=item C<parse_args>

See L<Template::Alloy::Parse> for more details.

=item C<set_variable>

Used to set a variable.  Expects a variable identity array and the
value to set.  It will autovifiy as necessary.

=item C<strict_throw>

Called during processing of template when STRICT configuration is
set and an uninitialized variable is met.  Arguments are the variable
identity reference.  Will call STRICT_THROW configuration item if set,
otherwise will call throw with a useful message.

=item C<throw>

Creates an exception object from the arguments and dies.

=item C<undefined_any>

Called during play_expr if a value is returned that is undefined.
This could be used to magically create variables on the fly.  This is
similar to Template::Stash::undefined.  It is suggested that
undefined_get be used instead.  Default behavior returns undef.  You
may also pass a coderef via the UNDEFINED_ANY configuration variable.
Also, you can try using the DEBUG => 'undef', configuration option
which will throw an error on undefined variables.

=item C<undefined_get>

Called when a variable is undefined during a GET directive.  This is
useful to see if a value that is about to get inserted into the text
is undefined.  undefined_any is a little too general for most cases.
Also, you may pass a coderef via the UNDEFINED_GET configuration
variable.

=back

=head1 OTHER UTILITY METHODS

The following is a brief list of other methods used by Alloy.  Generally, these
shouldn't be overwritten by subclasses.

=over 4

=item C<ast_string>

Returns perl code representation of a variable.

=item C<context>

Used to create a "pseudo" context object that allows for portability
of TT plugins, filters, and perl blocks that need a context object.
Uses the Template::Alloy::Context class.

=item C<debug_node>

Used to get debug info on a directive if DEBUG_DIRS is set.

=item C<get_line_number_by_index>

Used to turn string index position into line number

=item C<interpolate_node>

Used for parsing text nodes for dollar variables when interpolate is on.

=item C<play_operator>

Provided by the Operator role.  Allows for playing an operator AST.

See L<Template::Alloy::Operator> for more details.

=item C<apply_precedence>

Provided by the Parse role.  Allows for parsed operator array to be
translated to a tree based upon operator precedence.

=item C<_process>

Called by process and the PROCESS, INCLUDE and other directives.

=item C<slurp>

Reads contents of passed filename - throws file exception on error.

=item C<split_paths>

Used to split INCLUDE_PATH or other directives if an arrayref is not passed.

=item C<tt_var_string>

Returns a template toolkit representation of a variable.

=item C<_vars>

Return a reference to the current stash of variables.  This is currently only used
by the pseudo context object and may disappear at some point.

=back

=head1 THANKS

Thanks to Andy Wardley for creating Template::Toolkit.

Thanks to Sam Tregar for creating HTML::Template.

Thanks to David Lowe for creating Text::Tmpl.

Thanks to the Apache Velocity guys.

Thanks to Ben Grimm for a patch to allow passing a parsed document to the ->process method.

Thanks to David Warring for finding a parse error in HTE syntax.

Thanks to Carl Franks for adding the base ENCODING support.

=head1 AUTHOR

Paul Seamons <paul@seamons.com>

=head1 LICENSE

This module may be distributed under the same terms as Perl itself.

=cut