CGI::Ex 2.10

[chaz/p5-CGI-Ex] / lib / CGI / Ex / Template.pod

=head1 NAME

CGI::Ex::Template - Fast and lightweight TT2/3 template engine

=head1 SYNOPSIS

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

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

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

    ### Anything in the Template::Toolkit SYNOPSIS would fit here also

=head1 DESCRIPTION

CGI::Ex::Template happened by accident (accidentally on purpose).  The
CGI::Ex::Template (CET 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 CGI::Ex::Template did.  This was fine and dandy for a couple
of years.  In winter of 2005-2006 CET was revamped to add a few
features.  One thing led to another and soon CET provided for most of
the features of TT2 as well as some from TT3.  CGI::Ex::Template is a
full-featured implementation of the Template::Toolkit language.

CGI::Ex::Template (CET hereafter) is smaller, faster, uses less memory
and less CPU than TT2.  However, it is most likely less portable, less
extendable, and probably has many of the bugs that TT2 has already massaged
out from years of bug reports and patches from a very active community
and mailing list.  CET does not have a vibrant community behind it.  Fixes
applied to TT2 will take longer to get into CET, should they get in at all.
An attempt will be made to follow updates made to TT2 to keep the two
in sync at a language level.  There already has been, and it is expected that
there will continue to be code sharing between the two projects.  (Acutally
I will try and keep applicable fixes in sync with TT).

CGI::Ex::Template uses a recursive regex based grammar (early versions
before the 2.10 release 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.

Most of the standard Template::Toolkit documentation covering directives,
variables, configuration, plugins, filters, syntax, and vmethods should
apply to CET just fine (This pod tries to explain everything - but there is
too much).  The section on differences between CET and TT will explain
what too look out for.

Note: A clarification on "faster".  All templates are going to take
different amounts of time to process.  Different types of DIRECTIVES
parse and play more quickly than others.  The test script
samples/benchmark/bench_template.pl was used to obtain sample numbers.
In general the following statements are true:

    If you load a new Template object each time and pass a filename, CET
    is around 3.5 times faster.

    If you load a new Template object and pass a string ref, CET
    is around 3 times faster.

    If you load a new Template object and use CACHE_EXT, CET
    is around 1.5 times faster.

    If you use a cached object with a cached in memory template,
    then CET is 50% faster.

    If you use Template::Stash::XS with a cached in memory template,
    then CET is about as fast.  But if you use CGI::Ex::Template::XS,
    the CETX is faster still (about twice as fast as CET).

It is pretty hard to beat the speed of XS stash with compiled in
memory templates.  Many systems don't have access to those so
CET may make more sense.  Hopefully as TT is revised, many of the CET
speed advantages can be incorporated so that the core TT is just as
fast or faster.  This was last updated at version 2.10 of CET and
2.18 of TT.

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


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

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

    There are currently no errors during CGI::Ex::Template object creation.

=item C<process>

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

Process takes three arguments.

    $t->process($in, $swap, $out)
        || die $t->error;

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.

=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<define_vmethod>

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

=back

=head1 TODO

    Add WRAPPER configuration item (the WRAPPER directive is supported).

    Add ERROR config item

=head1 HOW IS CGI::Ex::Template DIFFERENT

CET uses the same base template syntax and configuration items as TT2,
but the internals of CET were written from scratch.  Additionally much
of the planned TT3 syntax is supported.  The following is a list of
some of the ways that the configuration and syntax of CET are
different from that of TT2.  Note: items that are planned to work in
TT3 are marked with (TT3).

=over 4

=item Numerical hash keys work

    [% a = {1 => 2} %]

=item Quoted hash key interpolation is fine

    [% a = {"$foo" => 1} %]

=item Multiple ranges in same constructor

    [% a = [1..10, 21..30] %]

=item Constructor types can call virtual methods. (TT3)

    [% a = [1..10].reverse %]

    [% "$foo".length %]

    [% 123.length %]   # = 3

    [% 123.4.length %]  # = 5

    [% -123.4.length %] # = -5 ("." binds more tightly than "-")

    [% (a ~ b).length %]

    [% "hi".repeat(3) %] # = hihihi

    [% {a => b}.size %] # = 1

=item The "${" and "}" variable interpolators can contain expressions,
not just variables.

    [% [0..10].${ 1 + 2 } %] # = 4

    [% {ab => 'AB'}.${ 'a' ~ 'b' } %] # = AB

    [% color = qw/Red Blue/; FOR [1..4] ; color.${ loop.index % color.size } ; END %]
      # = RedBlueRedBlue

=item Tags can be nested.

    [% f = "[% (1 + 2) %]" %][% f|eval %] # = 3

=item Arrays can be accessed with non-integer numbers.

    [% [0..10].${ 2.3 } %] # = 3

=item Reserved names are less reserved. (TT3)

    [% GET GET %] # gets the variable named "GET"

    [% GET $GET %] # gets the variable who's name is stored in "GET"

=item Filters and SCALAR_OPS are interchangeable. (TT3)

    [% a | length %]

    [% b . lower %]

=item Pipe "|" can be used anywhere dot "." can be and means to call
the virtual method. (TT3)

    [% a = {size => "foo"} %][% a.size %] # = foo

    [% a = {size => "foo"} %][% a|size %] # = 1 (size of hash)

=item Pipe "|" and "." can be mixed. (TT3)

    [% "aa" | repeat(2) . length %] # = 4

=item Added Virtual Object Namespaces. (TT3)

The Text, List, and Hash types give direct access
to virtual methods.

    [% a = "foobar" %][% Text.length(a) %] # = 6

    [% a = [1 .. 10] %][% List.size(a) %] # = 10

    [% a = {a=>"A", b=>"B"} ; Hash.size(a) %] = 2

    [% foo = {a => 1, b => 2}
       | Hash.keys
       | List.join(", ") %] # = a, b

=item Added "fmt" scalar, list, and hash virtual methods.

    [% list.fmt("%s", ", ") %]

    [% hash.fmt("%s => %s", "\n") %]

=item Whitespace is less meaningful. (TT3)

    [% 2-1 %] # = 1 (fails in TT2)

=item Added pow operator.

    [% 2 ** 3 %] [% 2 pow 3 %] # = 8 8

=item Added self modifiers (+=, -=, *=, /=, %=, **=, ~=). (TT3)

    [% a = 2;  a *= 3  ; a %] # = 6
    [% a = 2; (a *= 3) ; a %] # = 66

=item Added pre and post increment and decrement (++ --). (TT3)

    [% ++a ; ++a %] # = 12
    [% a-- ; a-- %] # = 0-1

=item Added qw// contructor. (TT3)

    [% a = qw(a b c); a.1 %] # = b

    [% qw/a b c/.2 %] # = c

=item Allow for scientific notation. (TT3)

    [% a = 1.2e-20 %]

    [% 123.fmt('%.3e') %] # = 1.230e+02

=item Allow for hexidecimal input. (TT3)

    [% a = 0xff0000 %][% a %] # = 16711680

    [% a = 0xff2 / 0xd; a.fmt('%x') %] # = 13a

=item FOREACH variables can be nested.

    [% FOREACH f.b = [1..10] ; f.b ; END %]

Note that nested variables are subject to scoping issues.
f.b will not be reset to its value before the FOREACH.

=item Post operative directives can be nested. (TT3)

Andy Wardley calls this side-by-side effect notation.

    [% one IF two IF three %]

    same as

    [% IF three %][% IF two %][% one %][% END %][% END %]


    [% a = [[1..3], [5..7]] %][% i FOREACH i = j FOREACH j = a %] # = 123567

=item Semi-colons on directives in the same tag are optional. (TT3)

    [% SET a = 1
       GET a
     %]

    [% FOREACH i = [1 .. 10]
         i
       END %]

Note: a semi-colon is still required in front of any block directive
that can be used as a post-operative directive.

    [% 1 IF 0
       2 %]   # prints 2

    [% 1; IF 0
       2
       END %] # prints 1

=item CATCH blocks can be empty.

TT2 requires them to contain something.

=item Added a DUMP directive.

Used for Data::Dumpering the passed variable or expression.

   [% DUMP a.a %]

=item CET does not generate Perl code.

It generates an "opcode" tree.  The opcode tree is an arrayref
of scalars and array refs nested as deeply as possible.  This "simple"
structure could be shared TT implementations in other languages
via JSON or YAML.

=item CET uses storable for its compiled templates.

If EVAL_PERL is off, CET will not eval_string on ANY piece of information.

=item There is eval_filter and MACRO recursion protection

You can control the nested nature of eval_filter and MACRO
recursion using the MAX_EVAL_RECURSE and MAX_MACRO_RECURSE
configuration items.

=item There is no context.

CET provides a context object that mimics the Template::Context
interface for use by some TT filters, eval perl blocks, and plugins.

=item There is no stash.

Well there is but it isn't an object.

CET only supports the variables passed in VARIABLES, PRE_DEFINE, and
those passed to the process method.  CET provides a stash object that
mimics the Template::Stash interface for use by some TT filters, eval
perl blocks, and plugins.

=item There is no provider.

CET uses the load_parsed_tree method to get and cache templates.

=item There is no grammar.

CET has its own built-in recursive regex based grammar system.

=item There is no VIEW directive.

=item The DEBUG directive is more limited.

It only understands DEBUG_DIRS (8) and DEBUG_UNDEF (2).

=item When debug dirs is on, directives on different lines separated by colons show the line they
are on rather than a general line range.

=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 CGI::Ex::Templates 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.

    ### 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       => CGI::Ex::Template->new,
    );

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

    ### pass the variables during object creation (will be available to every process call)
    my $cet = CGI::Ex::Template->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:

    CGI::Ex::Template=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 CET 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 CET.  They can be
used as arguments to functions, in place of variables in directives,
and in place of variables in expressions.  In CET 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

Hexidecimal 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 CET, not in TT.

You may also embed the current tags in strings (CET 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 CET, not in TT.

You may also embed the current tags in strings (CET 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 CET, 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 CET 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 CET, not in TT.

=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.

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 %]
    [% y = 3 %]
    [% z = x * (y - 1) %] Prints 4

=head1 VIRTUAL METHODS

The following is the list of builtin virtual methods and filters that
can be called on each type of data.

In CGI::Ex::Template, the "|" operator can be used to call virtual
methods just the same way that the "." operator can.  The main
difference between the two is that on access to hashrefs or objects,
the "|" means to always call the virtual method or filter rather than
looking in the hashref for a key by that name, or trying to call that
method on the object.  This is similar to how TT3 will function.

Virtual methods are also made available via Virtual Objects which
are discussed in a later section.

=head2 SCALAR VIRTUAL METHODS AND FILTERS

The following is the list of builtin virtual methods and filters
that can be called on scalar data types.  In CET and TT3, filters and
virtual methods are more closely related than in TT2.  In general anywhere a
virtual method can be used a filter can be used also - and likewise all scalar
virtual methods can be used as filters.

In addition to the filters listed below, CET will automatically load
Template::Filters and use them if Template::Toolkit is installed.

In addition to the scalar virtual methods, any scalar will be
automatically converted to a single item list if a list virtual method
is called on it.

Scalar virtual methods are also available through the "Text" virtual
object (except for true filters such as eval and redirect).

=over 4

=item '0'

    [% item = 'foo' %][% item.0 %] Returns self.  Allows for scalars to mask as arrays (scalars
                                   already will, but this allows for more direct access).

=item chunk

    [% item.chunk(60).join("\n") %] Split string up into a list of chunks of text 60 chars wide.

=item collapse

    [% item.collapse %] Strip leading and trailing whitespace and collapse all other space to one space.

=item defined

    [% item.defined %] Always true - because the undef sub translates all undefs to ''.

=item indent

    [% item.indent(3) %] Indent that number of spaces.

    [% item.indent("Foo: ") %] Add the string "Foo: " to the beginning of every line.

=item eval

    [% item.eval %]

Process the string as though it was a template.  This will start the parsing
engine and will use the same configuration as the current process.  CET is several times
faster at doing this than TT is and is considered acceptable.

This is a filter and is not available via the Text virtual object.

=item evaltt

    Same as the eval filter.

=item file

    Same as the redirect filter.

=item fmt

    [% item.fmt('%d') %]
    [% item.fmt('%6s') %]
    [% item.fmt('%*s', 6) %]

Similar to format.  Returns a string formatted with the passed pattern.  Default pattern is %s.

=item format

    [% item.format('%d') %]
    [% item.format('%6s') %]
    [% item.format('%*s', 6) %]

Print the string out in the specified format.  It is similar to
the "fmt" virtual method, except that the item is split on newline and each line is
processed separately.

=item hash

    [% item.hash %] Returns a one item hash with a key of "value" and a value of the item.

=item html

    [% item.html %] Performs a very basic html encoding (swaps out &, <, > and " for the html entities)

=item int

    [% item.int %] Return the integer portion of the value (0 if none).

=item lcfirst

    [% item.lcfirst %] Capitalize the leading letter.

=item length

    [% item.length %] Return the length of the string.

=item list

    [% item.list %] Returns a list with a single value of the item.

=item lower

    [% item.lower %] Return a lower-casified string.

=item match

    [% item.match("(\w+) (\w+)") %] Return a list of items matching the pattern.

    [% item.match("(\w+) (\w+)", 1) %] Same as before - but match globally.

=item null

    [% item.null %] Do nothing.

=item rand

    [% item = 10; item.rand %] Returns a number greater or equal to 0 but less than 10.
    [% 1.rand %]

Note: This filter is not available as of TT2.15.

=item remove

    [% item.remove("\s+") %] Same as replace - but is global and replaces with nothing.

=item redirect

    [% item.redirect("output_file.html") %]

Writes the contents out to the specified file.  The filename
must be relative to the OUTPUT_PATH configuration variable and the OUTPUT_PATH variable must be set.

This is a filter and is not available via the Text virtual object.

=item repeat

    [% item.repeat(3) %] Repeat the item 3 times

    [% item.repeat(3, ' | ') %] Repeat the item 3 times separated with ' | '

=item replace

    [% item.replace("\s+", "&nbsp;") %] Globally replace all space with &nbsp;

    [% item.replace("foo", "bar", 0) %] Replace only the first instance of foo with bar.

    [% item.replace("(\w+)", "($1)") %] Surround all words with parenthesis.

=item search

    [% item.search("(\w+)") %] Tests if the given pattern is in the string.

=item size

    [% item.size %] Always returns 1.

=item split

    [% item.split %] Returns an arrayref from the item split on " "

    [% item.split("\s+") %] Returns an arrayref from the item split on /\s+/

    [% item.split("\s+", 3) %] Returns an arrayref from the item split on /\s+/ splitting until 3 elements are found.

=item stderr

    [% item.stderr %] Print the item to the current STDERR handle.

=item substr

    [% item.substr(i) %] Returns a substring of item starting at i and going to the end of the string.

    [% item.substr(i, n) %] Returns a substring of item starting at i and going n characters.

=item trim

    [% item.trim %] Strips leading and trailing whitespace.

=item ucfirst

    [% item.ucfirst %] Lower-case the leading letter.

=item upper

    [% item.upper %] Return a upper-casified string.

=item uri

    [% item.uri %] Perform a very basic URI encoding.

=back

=head2 LIST VIRTUAL METHODS

The following methods can be called on an arrayref type data structures (scalar
types will automatically promote to a single element list and call these methods
if needed):

Additionally, list virtual methods can be accessed via the List
Virtual Object.

=over 4

=item fmt

    [% mylist.fmt('%s', ', ') %]
    [% mylist.fmt('%6s', ', ') %]
    [% mylist.fmt('%*s', ', ', 6) %]

Passed a pattern and an string to join on.  Returns a string of the values of the list
formatted with the passed pattern and joined with the passed string.
Default pattern is %s and the default join string is a space.

=item first

    [% mylist.first(3) %]  Returns a list of the first 3 items in the list.

=item grep

    [% mylist.grep("^\w+\.\w+$") %] Returns a list of all items matching the pattern.

=item hash

    [% mylist.hash %] Returns a hashref with the array indexes as keys and the values as values.

=item join

    [% mylist.join %] Joins on space.
    [% mylist.join(", ") Joins on the passed argument.

=item last

    [% mylist.last(3) %]  Returns a list of the last 3 items in the list.

=item list

    [% mylist.list %] Returns a reference to the list.

=item max

    [% mylist.max %] Returns the last item in the array.

=item merge

    [% mylist.merge(list2) %] Returns a new list with all defined items from list2 added.

=item nsort

    [% mylist.nsort %] Returns the numerically sorted items of the list.  If the items are
    hashrefs, a key containing the field to sort on can be passed.

=item pop

    [% mylist.pop %] Removes and returns the last element from the arrayref (the stash is modified).

=item push

    [% mylist.push(23) %] Adds an element to the end of the arrayref (the stash is modified).

=item random

    [% mylist.random %] Returns a random item from the list.
    [% ['a' .. 'z'].random %]

Note: This filter is not available as of TT2.15.

=item reverse

    [% mylist.reverse %] Returns the list in reverse order.

=item shift

    [% mylist.shift %] Removes and returns the first element of the arrayref (the stash is modified).

=item size

    [% mylist.size %] Returns the number of elements in the array.

=item slice

    [% mylist.slice(i, n) %] Returns a list from the arrayref beginning at index i and continuing for n items.

=item sort

    [% mylist.sort %] Returns the alphabetically sorted items of the list.  If the items are
    hashrefs, a key containing the field to sort on can be passed.

=item splice

    [% mylist.splice(i, n) %] Removes items from array beginning at i and continuing for n items.

    [% mylist.splice(i, n, list2) %] Same as before, but replaces removed items with the items
    from list2.

=item unique

    [% mylist.unique %] Return a list of the unique items in the array.

=item unshift

    [% mylist.unshift(23) %] Adds an item to the beginning of the arrayref.

=back 4

=head2 HASH VIRTUAL METHODS

The following methods can be called on hash type data structures:

Additionally, list virtual methods can be accessed via the Hash
Virtual Object.

=over 4

=item fmt

    [% myhash.fmt('%s => %s', "\n") %]
    [% myhash.fmt('%4s => %5s', "\n") %]
    [% myhash.fmt('%*s => %*s', "\n", 4, 5) %]

Passed a pattern and an string to join on.  Returns a string of the key/value pairs
of the hash formatted with the passed pattern and joined with the passed string.
Default pattern is "%s\t%s" and the default join string is a newline.

=item defined

    [% myhash.defined('a') %]  Checks if a is defined in the hash.

=item delete

    [% myhash.delete('a') %]  Deletes the item from the hash.

=item each

    [% myhash.each.join(", ") %]  Turns the contents of the hash into a list - subject
    to change as TT is changing the operations of each and list.

=item exists

    [% myhash.exists('a') %]  Checks if a is in the hash.

=item hash

    [% myhash.hash %]  Returns a reference to the hash.

=item import

    [% myhash.import(hash2) %]  Overlays the keys of hash2 over the keys of myhash.

=item item

    [% myhash.item(key) %] Returns the hashes value for that key.

=item items

    [% myhash.items %] Returns a list of the key and values (flattened hash)

=item keys

    [% myhash.keys.join(', ') %] Returns an arrayref of the keys of the hash.

=item list

    [% myhash.list %] Returns an arrayref with the hash as a single value (subject to change).

=item pairs

    [% myhash.pairs %] Returns an arrayref of hashrefs where each hash contains {key => $key, value => $value}
    for each value of the hash.

=item nsort

    [% myhash.nsort.join(", ") %] Returns a numerically sorted list of the keys.

=item size

    [% myhash.size %] Returns the number of key/value pairs in the hash.

=item sort

    [% myhash.sort.join(", ") Returns an alphabetically sorted list.

=item values

    [% myhash.values.join(', ') %] Returns an arrayref of the values of the hash.

=back

=head1 VIRTUAL OBJECTS

TT3 has a concept of Text, List, and Hash virtual objects which provide
direct access to the scalar, list, and hash virtual methods.  In the TT3
engine this will allow for more concise generated code.  Because CET does
not generated perl code to be executed later, CET provides for these virtual
objects but does so as more of a namespace (using the methods does not
provide a speed optimization in your template - just may help clarify things).

    [% a = "foo"; a.length %] => 3

    [% a = "foo"; Text.length(a) %] => 3

    [% a = Text.new("foo"); a.length %] => 3


    [% a = [1 .. 30]; a.size %] => 30

    [% a = [1 .. 30]; List.size(a) %] => 30

    [% a = List.new(1 .. 30); a.size %] => 30


    [% a = {a => 1, b => 2}; a.size %] => 2

    [% a = {a => 1, b => 2}; Hash.size(a) %] => 2

    [% a = Hash.new({a => 1, b => 2}); a.size %] => 2

    [% a = Hash.new(a => 1, b => 2); a.size %] => 2

    [% a = Hash.new(a = 1, b = 2); a.size %] => 2

    [% a = Hash.new('a', 1, 'b', 2); a.size %] => 2

One limitation is that if you pass a key named "Text",
"List", or "Hash" in your variable stash - the corresponding
virtual object will be hidden.

Additionally, you can use all of the Virtual object methods with
the pipe operator.

    [% {a => 1, b => 2}
       | Hash.keys
       | List.join(", ") %] => a, b

Again, there aren't any speed optimizations to using the virtual
objects in CET, but it can help clarify the intent in some cases.

Note: these aren't really objects.  All of the "virtual objects" are
references to the $SCALAR_OPS, $LIST_OPS, and $HASH_OPS hashes
found in the $VOBJS hash of CGI::Ex::Template.

=head1 DIRECTIVES

This section contains the alphabetical list of DIRECTIVES available in
the TT language.  DIRECTIVES are the "functions" and control
structures of the Template Toolkit mini-language.  For further
discussion and examples beyond what is listed below, please refer to
the TT directives documentation.

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

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

Multiple directives can be inside the same set of '[%' and '%]' tags
as long as they are separated by space or semi-colons (;).  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 CET 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<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>

This is not provided in TT.  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.

If the template is being processed in a web request, DUMP will html
encode the DUMP automatically.

    [% 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<FILTER>

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

=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 CGI::Ex::Template::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

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 %]

=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.

=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.

=item C<LAST>

Used to exit out of a WHILE or FOREACH loop.

=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)

=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 CGI::Ex::Template 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.

=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.

=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
    template1 => ['[\[%]%', '%[%\]]'], # allow TT1 style

If custom tags are supplied, by default they are escaped using
quotemeta.  If a third argument is given and is equal to "unquoted",
then no quoting takes place on the new tags.

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

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

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

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

=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, CET 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 rethrew.
        [% THROW $error %]
      [% END %]
    [% CATCH foo %]
      Caught foo.
    [% CATCH foo.bar %]
      Caught foo.bar.
    [% CATCH %]
      Caught anything else.
    [% END %]

Would print:

        Caught default - but rethrew.

      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 CET,
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
CGI::Ex::Template::Iterator if Template::Toolkit is not installed.  No
other plugins come installed with CGI::Ex::Template.

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
CET 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 cet = CGI::Ex::Template %]
    [%~ cet.dump_parse_expr('2 * 3').replace('\s+', ' ') %]

Would print:

    $VAR1 = [ [ undef, '*', '2', '3' ], 0 ];

See the PLUGIN_BASE, and PLUGINS configuration items.

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

=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 CGI::Ex::Template.

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 %]
    My content to be processed.[% a = 2 %]
    [% END %]

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

This would print.

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

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

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

Would print

    (foobar)');

=back



=head1 OPERATORS

The following operators are available in CGI::Ex::Template.  Except
where noted these are the same operators available in TT.  They are
listed in the order of their precedence (the higher the precedence the
tighter it binds).

=over 4

=item C<.>

The dot operator.  Allows for accessing sub-members, methods, or
virtual methods of nested data structures.

    my $obj->process(\$content, {a => {b => [0, {c => [34, 57]}]}}, \$output);

    [% a.b.1.c.0 %] => 34

Note: on access to hashrefs, any hash keys that match the sub key name
will be used before a virtual method of the same name.  For example if
a passed hash contained pair with a keyname "defined" and a value of
"2", then any calls to hash.defined(another_keyname) would always
return 2 rather than using the vmethod named "defined."  To get around
this limitation use the "|" operator (listed next).  Also - on objects
the "." will always try and call the method by that name.  To always
call the vmethod - use "|".

=item C<|>

The pipe operator.  Similar to the dot operator.  Allows for
explicit calling of virtual methods and filters (filters are "merged"
with virtual methods in CGI::Ex::Template and TT3) when accessing
hashrefs and objects.  See the note for the "." operator.

The pipe character is similar to TT2 in that it can be used in place
of a directive as an alias for FILTER.  It similar to TT3 in that it
can be used for virtual method access.  This duality is one source of
difference between CGI::Ex::Template and TT2 compatibility.  Templates
that have directives that end with a variable name that then use the
"|" directive to apply a filter will be broken as the "|" will be
applied to the variable name.

The following two cases will do the same thing.

    [% foo | html %]

    [% foo FILTER html %]

Though they do the same thing, internally, foo|html is stored as a
single variable while "foo FILTER html" is stored as the variable foo
which is then passed to the FILTER html.

A TT2 sample that would break in CGI::Ex::Template or TT3 is:

    [% PROCESS foo a = b | html %]

Under TT2 the content returned by "PROCESS foo a = b" would all be
passed to the html filter.  Under CGI::Ex::Template and TT3, b would
be passed to the html filter before assigning it to the variable "a"
before the template foo was processed.

A simple fix is to do any of the following:

    [% PROCESS foo a = b FILTER html %]

    [% | html %][% PROCESS foo a = b %][% END %]

    [% FILTER html %][% PROCESS foo a = b %][% END %]

This shouldn't be too much hardship and offers the great return of disambiguating
virtual method access.

=item C<\>

Unary.  The reference operator.  Not well publicized in TT.  Stores a reference
to a variable for use later.  Can also be used to "alias" long names.

    [% f = 7 ; foo = \f ; f = 8 ; foo %] => 8

    [% foo = \f.g.h.i.j.k; f.g.h.i.j.k = 7; foo %] => 7

    [% f = "abcd"; foo = \f.replace("ab", "-AB-") ; foo %] => -AB-cd

    [% f = "abcd"; foo = \f.replace("bc") ; foo("-BC-") %] => a-BC-d

    [% f = "abcd"; foo = \f.replace ; foo("cd", "-CD-") %] => ab-CD-

=item C<++ -->

Pre and post increment and decrement.  My be used as either a prefix
or postfix operator.

    [% ++a %][% ++a %] => 12

    [% a++ %][% a++ %] => 01

    [% --a %][% --a %] => -1-2

    [% a-- %][% a-- %] => 0-1

=item C<**  ^  pow>

Right associative binary.  X raised to the Y power.  This isn't available in TT 2.15.

    [% 2 ** 3 %] => 8

=item C<!>

Prefix not.  Negation of the value.

=item C<->

Prefix minus.  Returns the value multiplied by -1.

    [% a = 1 ; b = -a ; b %] => -1

=item C<*>

Left associative binary. Multiplication.

=item C</  div  DIV>

Left associative binary. Division.  Note that / is floating point division, but div and
DIV are integer division.

   [% 10  /  4 %] => 2.5
   [% 10 div 4 %] => 2

=item C<%  mod  MOD>

Left associative binary. Modulus.

   [% 15 % 8 %] => 7

=item C<+>

Left associative binary.  Addition.

=item C<->

Left associative binary.  Minus.

=item C<_  ~>

Left associative binary.  String concatenation.

    [% "a" ~ "b" %] => ab

=item C<< <  >  <=  >= >>

Non associative binary.  Numerical comparators.

=item C<lt  gt  le  ge>

Non associative binary.  String comparators.

=item C<==  eq>

Non associative binary.  Equality test.  TT chose to use Perl's eq for both operators.
There is no test for numeric equality.

=item C<!= ne>

Non associative binary.  Non-equality test.  TT chose to use Perl's ne for both
operators.  There is no test for numeric non-equality.

=item C<&&>

Left associative binary.  And.  All values must be true.  If all values are true, the last
value is returned as the truth value.

    [% 2 && 3 && 4 %] => 4

=item C<||>

Right associative binary.  Or.  The first true value is returned.

    [% 0 || '' || 7 %] => 7

Note: perl is left associative on this operator - but it doesn't matter because
|| has its own precedence level.  Setting it to right allows for CET to short
circuit earlier in the expression optree (left is (((1,2), 3), 4) while right
is (1, (2, (3, 4))).

=item C<..>

Non associative binary.  Range creator.  Returns an arrayref containing the values
between and including the first and last arguments.

    [% t = [1 .. 5] %] => variable t contains an array with 1,2,3,4, and 5

It is possible to place multiple ranges in the same [] constructor.  This is not available in TT.

    [% t = [1..3, 6..8] %] => variable t contains an array with 1,2,3,6,7,8

The .. operator is the only operator that returns a list of items.

=item C<? :>

Ternary - right associative.  Can be nested with other ?: pairs.

    [% 1 ? 2 : 3 %] => 2
    [% 0 ? 2 : 3 %] => 3

=item C<*= += -= /= **= %= ~=>

Self-modifying assignment - right associative.  Sets the left hand side
to the operation of the left hand side and right (clear as mud).
In order to not conflict with SET, FOREACH and other operations, this
operator is only available in parenthesis.

   [% a = 2 %][%  a += 3  %] --- [% a %]    => --- 5   # is handled by SET
   [% a = 2 %][% (a += 3) %] --- [% a %]    => 5 --- 5

=item C<=>

Assignment - right associative.  Sets the left-hand side to the value of the righthand side.  In order
to not conflict with SET, FOREACH and other operations, this operator is only
available in parenthesis.  Returns the value of the righthand side.

   [%  a = 1  %] --- [% a %]    => --- 1   # is handled by SET
   [% (a = 1) %] --- [% a %]    => 1 --- 1

=item C<not  NOT>

Prefix. Lower precedence version of the '!' operator.

=item C<and  AND>

Left associative. Lower precedence version of the '&&' operator.

=item C<or OR>

Right associative. Lower precedence version of the '||' operator.

=item C<{}>

This operator is not used in TT.  It is used internally
by CGI::Ex::Template to delay the creation of a hash until the
execution of the compiled template.

=item C<[]>

This operator is not used in TT.  It is used internally
by CGI::Ex::Template to delay the creation of an array until the
execution of the compiled template.

=back


=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.

=head1 CONFIGURATION

The following TT2 configuration variables are supported (in
alphabetical order).  Note: for further discussion you can refer to
the TT config documentation.

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

    my $obj = CGI::Ex::Template->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 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 BLOCKS

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_parsed_tree 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 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.

=item COMPILE_EXT

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

=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.
    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 END_TAG

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

=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 = CGI::Ex::Template->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 INCLUDE_PATH

A string or an arrayref or coderef that returns an arrayref that
contains directories to look for files included by processed
templates.

=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.

=item MAX_EVAL_RECURSE (CET only)

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

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

=item MAX_MACRO_RECURSE (CET only)

Will use $CGI::Ex::Template::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.

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

Is the same as

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

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

=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

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

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

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 START_TAG

Set a string to use as the opening delimiter for TT.  Default is "[%".

=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 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.

=back



=head1 UNSUPPORTED TT CONFIGURATION

=over 4

=item WRAPPER

This will be supported - just not done yet.

=item ERROR

This will be supported - just not done yet.

=item LOAD_TEMPLATES

CGI::Ex::Template has its own mechanism for loading and storing
compiled templates.  TT would use a Template::Provider that would
return a Template::Document.  The closest thing in CGI::Ex::Template
is the load_parsed_template method.  There is no immediate plan to
support the TT behavior.

=item LOAD_PLUGINS

CGI::Ex::Template uses its own mechanism for loading plugins.  TT
would use a Template::Plugins object to load plugins requested via the
USE directive.  The functionality for doing this in CGI::Ex::Template
is contained in the list_plugins method and the play_USE method.  There
is no immediate plan to support the TT behavior.

Full support is offered for the PLUGINS and LOAD_PERL configuration items.

Also note that CGI::Ex::Template only natively supports the Iterator plugin.
Any of the other plugins requested will need to provided by installing
Template::Toolkit or the appropriate plugin module.

=item LOAD_FILTERS

CGI::Ex::Template uses its own mechanism for loading filters.  TT
would use the Template::Filters object to load filters requested via the
FILTER directive.  The functionality for doing this in CGI::Ex::Template
is contained in the list_filters method and the play_expr method.

Full support is offered for the FILTERS configuration item.

=item TOLERANT

This option is used by the LOAD_TEMPLATES and LOAD_PLUGINS options and
is not applicable in CGI::Ex::Template.

=item SERVICE

CGI::Ex::Template has no concept of service (theoretically the CGI::Ex::Template
is the "service").

=item CONTEXT

CGI::Ex::Template provides its own pseudo context object to plugins,
filters, and perl blocks.  The CGI::Ex::Template model doesn't really
allow for a separate context.  CGI::Ex::Template IS the context.

=item STASH

CGI::Ex::Template manages its own stash of variables.  A pseudo stash
object is available via the pseudo context object for use in plugins,
filters, and perl blocks.

=item PARSER

CGI::Ex::Template has its own built in parser.  The closest similarity is
the parse_tree method.  The output of parse_tree is an optree that is
later run by execute_tree.

=item GRAMMAR

CGI::Ex::Template maintains its own grammar.  The grammar is defined
in the parse_tree method and the callbacks listed in the global
$DIRECTIVES hashref.

=back


=head1 VARIABLE PARSE TREE

CGI::Ex::Template parses templates into an tree of operations.  Even
variable access is parsed into a tree.  This is done in a manner
somewhat similar to the way that TT operates except that nested
variables such as foo.bar|baz contain the '.' or '|' in between each
name level.  Operators are parsed and stored as part of the variable (it
may be more appropriate to say we are parsing a term or an expression).

The following table shows a variable or expression and the corresponding parsed tree
(this is what the parse_expr method would return).

    one                [ 'one',  0 ]
    one()              [ 'one',  [] ]
    one.two            [ 'one',  0, '.', 'two',  0 ]
    one|two            [ 'one',  0, '|', 'two',  0 ]
    one.$two           [ 'one',  0, '.', ['two', 0 ], 0 ]
    one(two)           [ 'one',  [ ['two', 0] ] ]
    one.${two().three} [ 'one',  0, '.', ['two', [], '.', 'three', 0], 0]
    2.34               2.34
    "one"              "one"
    "one"|length       [ [ undef, '~', "one" ], 0, '|', 'length', 0 ]
    "one $a two"       [ [ undef, '~', 'one ', ['a', 0], ' two' ], 0 ]
    [0, 1, 2]          [ [ undef, '[]', 0, 1, 2 ], 0 ]
    [0, 1, 2].size     [ [ undef, '[]', 0, 1, 2 ], 0, '.', 'size', 0 ]
    ['a', a, $a ]      [ [ undef, '[]', 'a', ['a', 0], [['a', 0], 0] ], 0]
    {a  => 'b'}        [ [ undef, '{}', 'a', 'b' ], 0 ]
    {a  => 'b'}.size   [ [ undef, '{}', 'a', 'b' ], 0, '.', 'size', 0 ]
    {$a => b}          [ [ undef, '{}', ['a', 0], ['b', 0] ], 0 ]
    1 + 2              [ [ undef, '+', 1, 2 ], 0]
    a + b              [ [ undef, '+', ['a', 0], ['b', 0] ], 0 ]
    a * (b + c)        [ [ undef, '*', ['a', 0], [ [undef, '+', ['b', 0], ['c', 0]], 0 ]], 0 ]
    (a + b)            [ [ undef, '+', ['a', 0], ['b', 0] ]], 0 ]
    (a + b) * c        [ [ undef, '*', [ [undef, '+', ['a', 0], ['b', 0] ], 0 ], ['c', 0] ], 0 ]
    a ? b : c          [ [ undef, '?', ['a', 0], ['b', 0], ['c', 0] ], 0 ]
    a || b || c        [ [ undef, '||', ['a', 0], [ [undef, '||', ['b', 0], ['c', 0] ], 0 ] ], 0 ]
    ! a                [ [ undef, '!', ['a', 0] ], 0 ]

Some notes on the parsing.

    Operators are parsed as part of the variable and become part of the variable tree.

    Operators are stored in the variable tree using an operator identity array which
    contains undef as the first value, the operator, and the operator arguments.  This
    allows for quickly descending the parsed variable tree and determining that the next
    node is an operator.

    Parenthesis () can be used at any point in an expression to disambiguate precedence.

    "Variables" that appear to be literal strings or literal numbers
    are returned as the literal (no operator tree).

The following perl can be typed at the command line to view the parsed variable tree:

    perl -e 'use CGI::Ex::Template; print CGI::Ex::Template::dump_parse_expr("foo.bar + 2")."\n"'

Also the following can be included in a template to view the output in a template:

    [% USE cet = CGI::Ex::Template %]
    [%~ cet.dump_parse_expr('foo.bar + 2').replace('\s+', ' ') %]


=head1 SEMI PUBLIC METHODS

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

=over 4

=item C<dump_parse>

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<exception>

Creates an exception object blessed into the package listed in
$CGI::Ex::Template::PACKAGE_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 is not natively implemented in CET.

=item C<list_plugins>

Returns an arrayref of modules that are under a base Namespace.

  my @modules = @{ $self->list_plugins({base => 'Template::Plugins'}) }:

=item C<load_parsed_tree>

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

  my $doc = $self->load_parsed_tree($file) || $self->throw('undef', "Zero length content");

=item C<parse_args>

Allow for the multitudinous ways that TT parses arguments.  This allows
for positional as well as named arguments.  Named arguments can be separated with a "=" or "=>",
and positional arguments should be separated by " " or ",".  This only returns an array
of parsed variables.   To get the actual values, you must call play_expr on each value.

=item C<parse_tree>

Used by load_parsed_tree.  This is the main grammar engine of the program.  It
uses method in the $DIRECTIVES hashref to parse different DIRECTIVE TYPES.

=item C<parse_expr>

Used to parse a variable, an expression, a literal string, or a number.  It
returns a parsed variable tree.  Samples of parsed variables can be found in the VARIABLE PARSE TREE
section.

=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<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 CET.  Generally, these
shouldn't be overwritten by subclasses.

=over 4

=item C<apply_precedence>

Allows for parsed operator array to be translated to a tree based
upon operator precedence.

=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.

=item C<DEBUG>

TT2 Holdover that is used once for binmode setting during a TT2 test.

=item C<debug_node>

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

=item C<filter_*>

Methods by these names implement filters that are more than one line.

=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<parse_*>

Methods by these names are used by parse_tree to parse the template.  These are the grammar.

=item C<play_*>

Methods by these names are used by execute_tree to execute the parsed tree.

=item C<play_operator>

Used to execute any found operators.  The single argument is
an operator identy returned by the parse_expr method (if the expression
contained an operator).  Normally you would just call play_expr
instead and it will call play_operator if the structure
contains an operator.

=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<_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.

=item C<vmethod_*>

Methods by these names implement virtual methods that are more than one line.

=back


=head1 AUTHOR

Paul Seamons <paul at seamons dot com>

=cut