package Text::Xslate;
# The Xslate engine class
use 5.008_001;
use strict;
use warnings;

use version; our $VERSION = version->declare('v3.5.9');

use Carp              ();
use File::Spec        ();
use Exporter          ();
use Data::MessagePack ();
use Scalar::Util      ();

use Text::Xslate::Util ();
    # all the exportable functions are defined in ::Util
    our @EXPORT_OK = qw(

our @ISA = qw(Text::Xslate::Engine);


# $bytecode_version + $fullpath + $compiler_and_parser_options
my $XSLATE_MAGIC   = qq{xslate;$BYTECODE_VERSION;%s;%s;};


# load backend (XS or PP)
my $use_xs = 0;
if(!exists $INC{'Text/Xslate/'}) {
    my $pp = ($Text::Xslate::Util::DEBUG =~ /\b pp \b/xms or $ENV{PERL_ONLY});
    if(!$pp) {
        eval {
            require XSLoader;
            XSLoader::load(__PACKAGE__, $VERSION);
            $use_xs = 1;
        die $@ if $@ && $Text::Xslate::Util::DEBUG =~ /\b xs \b/xms; # force XS
    if(!__PACKAGE__->can('render')) {
        require 'Text/Xslate/';
sub USE_XS() { $use_xs }

# for error messages (see T::X::Util)
sub input_layer { ref($_[0]) ? $_[0]->{input_layer} : ':utf8' }

package Text::Xslate::Engine; # XS/PP common base class

use version; our $VERSION = version->declare('v3.5.9');

use Text::Xslate::Util qw(

# constants
    our @ISA = qw(Exporter);

    my $dump_load = scalar($Text::Xslate::Util::DEBUG =~ /\b dump=load \b/xms);
    *_DUMP_LOAD = sub(){ $dump_load };

    my $save_src = scalar($Text::Xslate::Util::DEBUG =~ /\b save_src \b/xms);
    *_SAVE_SRC  = sub() { $save_src };

    *_ST_MTIME = sub() { 9 }; # see perldoc -f stat

unless(defined $DEFAULT_CACHE_DIR) {
    my $cache_dir = '.xslate_cache';
    foreach my $d($ENV{HOME}, File::Spec->tmpdir) {
        if(defined($d) and -d $d and -w _) {
            $cache_dir = File::Spec->catfile($d, '.xslate_cache');

    $DEFAULT_CACHE_DIR = $cache_dir;

# the real defaults are defined in the parser
my %parser_option = (
    line_start => undef,
    tag_start  => undef,
    tag_end    => undef,

# the real defaults are defined in the compiler
my %compiler_option = (
    syntax     => undef,
    type       => undef,
    header     => undef, # template augment
    footer     => undef, # template augment
    macro      => undef, # template augment

my %builtin = (
    html_escape  => \&Text::Xslate::Util::html_escape,
    mark_raw     => \&Text::Xslate::Util::mark_raw,
    unmark_raw   => \&Text::Xslate::Util::unmark_raw,
    uri_escape   => \&Text::Xslate::Util::uri_escape,

    is_array_ref => \&Text::Xslate::Util::is_array_ref,
    is_hash_ref  => \&Text::Xslate::Util::is_hash_ref,

    dump         => \&Text::Xslate::Util::dump,

    # aliases
    raw          => 'mark_raw',
    html         => 'html_escape',
    uri          => 'uri_escape',

sub default_functions { +{} } # overridable

sub parser_option { # overridable
    return \%parser_option;

sub compiler_option { # overridable
    return \%compiler_option;

sub replace_option_value_for_magic_token { # overridable
    #my($self, $name, $value) = @_;
    return $_[2];

sub options { # overridable
    my($self) = @_;
    return {
        # name       => default
        suffix       => '.tx',
        path         => ['.'],
        input_layer  => $self->input_layer,
        cache        => 1, # 0: not cached, 1: checks mtime, 2: always cached
        cache_dir    => $DEFAULT_CACHE_DIR,
        module       => undef,
        function     => undef,
        html_builder_module => undef,
        compiler     => 'Text::Xslate::Compiler',

        verbose      => 1,
        warn_handler => undef,
        die_handler  => undef,
        pre_process_handler => undef,

        %{ $self->parser_option },
        %{ $self->compiler_option },

sub new {
    my $class = shift;
    my %args  = (@_ == 1 ? %{$_[0]} : @_);

    my $options = $class->options;
    my $used    = 0;
    my $nargs   = scalar keys %args;
    foreach my $key(keys %{$options}) {
        if(exists $args{$key}) {
        elsif(defined($options->{$key})) {
            $args{$key} = $options->{$key};

    if($used != $nargs) {
        my @unknowns = sort grep { !exists $options->{$_} } keys %args;
            => "$class: Unknown option(s): " . join ' ', @unknowns);

    $args{path} = [
        map { ref($_) ? $_ : File::Spec->rel2abs($_) }
            ref($args{path}) eq 'ARRAY' ? @{$args{path}} : $args{path}

    my $arg_function= $args{function};

    my %funcs;
    $args{function} = \%funcs;

    $args{template} = {}; # template structures

    my $self = bless \%args, $class;

    # definition of functions and methods

    # builtin functions
    %funcs = %builtin;

    # per-class functions
    $self->_merge_hash(\%funcs, $class->default_functions());

    # user-defined functions
    if(defined $args{module}) {

    # user-defined html builder functions
    if(defined $args{html_builder_module}) {
        my $raw = Text::Xslate::Util::import_from(@{$args{html_builder_module}});
        my $html_builders = +{
            map {
                ($_ => &Text::Xslate::Util::html_builder($raw->{$_}))
            } keys %$raw
        $self->_merge_hash(\%funcs, $html_builders);

    $self->_merge_hash(\%funcs, $arg_function);


    return $self;

sub _merge_hash {
    my($self, $base, $add) = @_;
    foreach my $name(keys %{$add}) {
        $base->{$name} = $add->{$name};

sub _resolve_function_aliases {
    my($self, $funcs) = @_;

    foreach my $f(values %{$funcs}) {
        my %seen; # to avoid infinite loops
        while(!( ref($f) or Scalar::Util::looks_like_number($f) )) {
            my $v = $funcs->{$f} or $self->_error(
               "Cannot resolve a function alias '$f',"
               . " which refers nothing",

            if( ref($v) or Scalar::Util::looks_like_number($v) ) {
                $f = $v;
            else {
                $seen{$v}++ and $self->_error(
                    "Cannot resolve a function alias '$f',"
                    . " which makes circular references",


sub load_string { # called in render_string()
    my($self, $string) = @_;
    if(not defined $string) {
        $self->_error("LoadError: Template string is not given");
    $self->note('  _load_string: %s', join '\n', split /\n/, $string)
        if _DUMP_LOAD;
    $self->{source}{'<string>'} = $string if _SAVE_SRC;
    $self->{string_buffer} = $string;
    my $asm = $self->compile($string);
    $self->_assemble($asm, '<string>', \$string, undef, undef);
    return $asm;

my $updir = File::Spec->updir;
sub find_file {
    my($self, $file) = @_;

    if($file =~ /\Q$updir\E/xmso) {
        $self->_error("LoadError: Forbidden component (updir: '$updir') found in file name '$file'");

    my $fullpath;
    my $cachepath;
    my $orig_mtime;
    my $cache_mtime;
    foreach my $p(@{$self->{path}}) {
        $self->note("  find_file: %s in  %s ...\n", $file, $p) if _DUMP_LOAD;

        my $cache_prefix;
        if(ref $p eq 'HASH') { # virtual path
            defined(my $content = $p->{$file}) or next;
            $fullpath = \$content;

            # NOTE:
            # Because contents of virtual paths include their digest,
            # time-dependent cache verifier makes no sense.
            $orig_mtime   = 0;
            $cache_mtime  = 0;
            $cache_prefix = 'HASH';
        else {
            $fullpath = File::Spec->catfile($p, $file);
            defined($orig_mtime = (stat($fullpath))[_ST_MTIME])
                or next;
            $cache_prefix = Text::Xslate::uri_escape($p);
            if (length $cache_prefix > 127) {
                # some filesystems refuse a path part with length > 127
                $cache_prefix = $self->_digest($cache_prefix);

        # $file is found
        $cachepath = File::Spec->catfile(
            $file . 'c',
        # stat() will be failed if the cache doesn't exist
        $cache_mtime = (stat($cachepath))[_ST_MTIME];

    if(not defined $orig_mtime) {
        $self->_error("LoadError: Cannot find '$file' (path: @{$self->{path}})");

    $self->note("  find_file: %s (mtime=%d)\n",
        $fullpath, $cache_mtime || 0) if _DUMP_LOAD;

    return {
        name        => ref($fullpath) ? $file : $fullpath,
        fullpath    => $fullpath,
        cachepath   => $cachepath,

        orig_mtime  => $orig_mtime,
        cache_mtime => $cache_mtime,

sub load_file {
    my($self, $file, $mtime, $omit_augment) = @_;

    local $self->{omit_augment} = $omit_augment;

    $self->note("%s->load_file(%s)\n", $self, $file) if _DUMP_LOAD;

    if($file eq '<string>') { # simply reload it
        return $self->load_string($self->{string_buffer});

    my $fi = $self->find_file($file);

    my $asm = $self->_load_compiled($fi, $mtime) || $self->_load_source($fi, $mtime);

    # $cache_mtime is undef : uses caches without any checks
    # $cache_mtime > 0      : uses caches with mtime checks
    # $cache_mtime == 0     : doesn't use caches
    my $cache_mtime;
    if($self->{cache} < 2) {
        $cache_mtime = $fi->{cache_mtime} || 0;

    $self->_assemble($asm, $file, $fi->{fullpath}, $fi->{cachepath}, $cache_mtime);
    return $asm;

sub slurp_template {
    my($self, $input_layer, $fullpath) = @_;

    if (ref $fullpath eq 'SCALAR') {
        return $$fullpath;
    } else {
        open my($source), '<' . $input_layer, $fullpath
            or $self->_error("LoadError: Cannot open $fullpath for reading: $!");
        local $/;
        return scalar <$source>;

sub _load_source {
    my($self, $fi) = @_;
    my $fullpath  = $fi->{fullpath};
    my $cachepath = $fi->{cachepath};

    $self->note("  _load_source: try %s ...\n", $fullpath) if _DUMP_LOAD;

    # This routine is called when the cache is no longer valid (or not created yet)
    # so it should be ensured that the cache, if exists, does not exist
    if(-e $cachepath) {
        unlink $cachepath
            or Carp::carp("Xslate: cannot unlink $cachepath (ignored): $!");

    my $source = $self->slurp_template($self->input_layer, $fullpath);
    $source = $self->{pre_process_handler}->($source) if $self->{pre_process_handler};
    $self->{source}{$fi->{name}} = $source if _SAVE_SRC;

    my $asm = $self->compile($source,
        file => $fullpath,
        name => $fi->{name},

    if($self->{cache} >= 1) {
        my($volume, $dir) = File::Spec->splitpath($fi->{cachepath});
        my $cachedir      = File::Spec->catpath($volume, $dir, '');
        if(not -e $cachedir) {
            require File::Path;
            if (! -e $cachedir) {
                Carp::croak("Xslate: Cannot prepare cache directory $cachepath (ignored): $@");

        my $tmpfile = sprintf('%s.%d.%d', $cachepath, $$, Scalar::Util::refaddr($self));

        if (open my($out), ">:raw", $tmpfile) {
            my $mtime = $self->_save_compiled($out, $asm, $fullpath, utf8::is_utf8($source));

            if(!close $out) {
                 Carp::carp("Xslate: Cannot close $cachepath (ignored): $!");
                 unlink $tmpfile;
            elsif (rename($tmpfile => $cachepath)) {
                # set the newest mtime of all the related files to cache mtime
                if (not ref $fullpath) {
                    my $main_mtime = (stat $fullpath)[_ST_MTIME];
                    if (defined($main_mtime) && $main_mtime > $mtime) {
                        $mtime = $main_mtime;
                    utime $mtime, $mtime, $cachepath;
                    $fi->{cache_mtime} = $mtime;
                else {
                    $fi->{cache_mtime} = (stat $cachepath)[_ST_MTIME];
            else {
                Carp::carp("Xslate: Cannot rename cache file $cachepath (ignored): $!");
                unlink $tmpfile;
        else {
            Carp::carp("Xslate: Cannot open $cachepath for writing (ignored): $!");
    if(_DUMP_LOAD) {
        $self->note("  _load_source: cache(mtime=%s)\n",
            defined $fi->{cache_mtime} ? $fi->{cache_mtime} : 'undef');

    return $asm;

# load compiled templates if they are fresh enough
sub _load_compiled {
    my($self, $fi, $threshold) = @_;

    if($self->{cache} >= 2) {
        # threshold is the most latest modified time of all the related caches,
        # so if the cache level >= 2, they seems always fresh.
        $threshold = 9**9**9; # force to purge the cache
    else {
        $threshold ||= $fi->{cache_mtime};
    # see also tx_load_template() in xs/Text-Xslate.xs
    if(!( defined($fi->{cache_mtime}) and $self->{cache} >= 1
            and $threshold >= $fi->{orig_mtime} )) {
        $self->note( "  _load_compiled: no fresh cache: %s, %s",
            $threshold || 0, Text::Xslate::Util::p($fi) ) if _DUMP_LOAD;
        $fi->{cache_mtime} = undef;
        return undef;

    my $cachepath = $fi->{cachepath};
    open my($in), '<:raw', $cachepath
        or $self->_error("LoadError: Cannot open $cachepath for reading: $!");

    my $magic = $self->_magic_token($fi->{fullpath});
    my $data;
    read $in, $data, length($magic);
    if($data ne $magic) {
        return undef;
    else {
        local $/;
        $data = <$in>;
        close $in;
    my $unpacker = Data::MessagePack::Unpacker->new();
    my $offset  = $unpacker->execute($data);
    my $is_utf8 = $unpacker->data();


    my @asm;
    if($is_utf8) { # TODO: move to XS?
        my $seed = "";
        push @asm, ['print_raw_s', $seed, __LINE__, __FILE__];
    while($offset < length($data)) {
        $offset = $unpacker->execute($data, $offset);
        my $c = $unpacker->data();

        # my($name, $arg, $line, $file, $symbol) = @{$c};
        if($c->[0] eq 'depend') {
            my $dep_mtime = (stat $c->[1])[_ST_MTIME];
            if(!defined $dep_mtime) {
                Carp::carp("Xslate: Failed to stat $c->[1] (ignored): $!");
                return undef; # purge the cache
            if($dep_mtime > $threshold){
                $self->note("  _load_compiled: %s(%s) is newer than %s(%s)\n",
                    $c->[1],    scalar localtime($dep_mtime),
                    $cachepath, scalar localtime($threshold) )
                        if _DUMP_LOAD;
                return undef; # purge the cache
        elsif($c->[0] eq 'literal') {
            # force upgrade to avoid UTF-8 key issues
            utf8::upgrade($c->[1]) if($is_utf8);
        push @asm, $c;

    if(_DUMP_LOAD) {
        $self->note("  _load_compiled: cache(mtime=%s)\n",
            defined $fi->{cache_mtime} ? $fi->{cache_mtime} : 'undef');

    return \@asm;

sub _save_compiled {
    my($self, $out, $asm, $fullpath, $is_utf8) = @_;
    my $mp = Data::MessagePack->new();
    local $\;
    print $out $self->_magic_token($fullpath);
    print $out $mp->pack($is_utf8 ? 1 : 0);

    my $newest_mtime = 0;
    foreach my $c(@{$asm}) {
        print $out $mp->pack($c);

        if ($c->[0] eq 'depend') {
            my $dep_mtime = (stat $c->[1])[_ST_MTIME];
            if ($newest_mtime < $dep_mtime) {
                $newest_mtime = $dep_mtime;
    return $newest_mtime;

sub _magic_token {
    my($self, $fullpath) = @_;

    $self->{serial_opt} ||= Data::MessagePack->pack([
        ref($self->{compiler}) || $self->{compiler},
        [sort keys %{ $self->{function} }],

    if(ref $fullpath) { # ref to content string
        $fullpath = join ':', ref($fullpath),
    return sprintf $XSLATE_MAGIC, $fullpath, $self->{serial_opt};

sub _digest {
    my($self, $content) = @_;
    require 'Digest/'; # we don't want to create its namespace
    my $md5 = Digest::MD5->new();
    return $md5->hexdigest();

sub _extract_options {
    my($self, $opt_ref) = @_;
    my @options;
    foreach my $name(sort keys %{$opt_ref}) {
        if(exists $self->{$name}) {
            push @options, $name => $self->{$name};
    return @options;

sub _filter_options_for_magic_token {
    my($self, @options) = @_;
    my @filterd_options;
    while (@options) {
        my $name  = shift @options;
        my $value = $self->replace_option_value_for_magic_token($name, shift @options);
        push(@filterd_options, $name => $value);

sub _compiler {
    my($self) = @_;
    my $compiler = $self->{compiler};

    if(!ref $compiler){
        require Mouse;

        my $input_layer = $self->input_layer;
        $compiler = $compiler->new(
            engine      => $self,
            input_layer => $input_layer,
            parser_option => {
                input_layer => $input_layer,

        $compiler->define_function(keys %{ $self->{function} });

        $self->{compiler} = $compiler;

    return $compiler;

sub compile {
    my $self = shift;
    return $self->_compiler->compile(@_,
        omit_augment => $self->{omit_augment});

sub _error {
    die make_error(@_);

sub note {
    my($self, @args) = @_;
    printf STDERR @args;

package Text::Xslate;

=for stopwords sandboxing vs metacharacters name-coderef html xml cb tx TTerse HTML xslate HTMLTemplate gardejo jjn1056 clouder chiba turugina cho45 shmorimo ueda punytan

=head1 NAME

Text::Xslate - Scalable template engine for Perl5

=head1 VERSION

This document describes Text::Xslate version v3.5.9.


    use Text::Xslate qw(mark_raw);

    my $tx = Text::Xslate->new();

    my %vars = (
        title => 'A list of books',
        books => [
            { title => 'Islands in the stream' },
            { title => 'Programming Perl'      },
            # ...

        # mark HTML components as raw not to escape its HTML tags
        gadget => mark_raw('<div class="gadget">...</div>'),

    # for files
    print $tx->render('hello.tx', \%vars);

    # for strings (easy but slow)
    my $template = q{
        <h1><: $title :></h1>
        : for $books -> $book {
            <li><: $book.title :></li>
        : } # for

    print $tx->render_string($template, \%vars);


B<Xslate> is a template engine, tuned for persistent applications,
safe as an HTML generator, and with rich features.

There are a lot of template engines in CPAN, for example Template-Toolkit,
Text::MicroTemplate, HTML::Template, and so on, but all of them have
some weak points: a full-featured template engine may be slow,
while a fast template engine may be too simple to use. This is why Xslate is
developed, which is the best template engine for web applications.

The concept of Xslate is strongly influenced by Text::MicroTemplate
and Template-Toolkit 2, but the central philosophy of Xslate is different
from them. That is, the philosophy is B<sandboxing> that the template logic
should not have no access outside the template beyond your permission.

Other remarkable features are as follows:

=head2 Features

=head3 High performance

This engine introduces the virtual machine paradigm. Templates are
compiled into intermediate code, and then executed by the virtual machine,
which is highly optimized for rendering templates. Thus, Xslate is
much faster than any other template engines.

The template roundup project by Sam Graham shows Text::Xslate got
amazingly high scores in I<instance_reuse> condition
(i.e. for persistent applications).


=item The template roundup project


=item Perl Template Roundup October 2010 Performance vs Variant Report: instance_reuse



There are also benchmarks in F<benchmark/> directory in the Xslate distribution.

=head3 Smart escaping for HTML metacharacters

Xslate employs the B<smart escaping strategy>, where a template engine
escapes all the HTML metacharacters in template expressions unless users
mark values as B<raw>.
That is, the output is unlikely to prone to XSS.

=head3 Template cascading

Xslate supports the B<template cascading>, which allows you to extend
templates with block modifiers. It is like a traditional template inclusion,
but is more powerful.

This mechanism is also called as template inheritance.

=head3 Easiness to enhance

Xslate is ready to enhance. You can add functions and methods to the template
engine and even add a new syntax via extending the parser.


=head2 Methods

=head3 B<< Text::Xslate->new(%options) >>

Creates a new Xslate template engine with options. You can reuse this instance
for multiple calls to C<render()>.

Possible options are:


=item C<< path => \@path // ['.'] >>

Specifies the include paths, which may be directory names or virtual paths,
i.e. HASH references which contain C<< $file_name => $content >> pairs.

Note that if you use taint mode (C<-T>), you have to give absolute paths
to C<path> and C<cache_dir>. Otherwise you'll get errors because they
depend on the current working directory which might not be secure.

=item C<< cache => $level // 1 >>

Sets the cache level.

If C<< $level == 1 >> (default), Xslate caches compiled templates on the disk, and
checks the freshness of the original templates every time.

If C<< $level >= 2 >>, caches will be created but the freshness
will not be checked.

C<< $level == 0 >> uses no caches, which is provided for testing.

=item C<< cache_dir => $dir // "$ENV{HOME}/.xslate_cache" >>

Specifies the directory used for caches. If C<$ENV{HOME}> doesn't exist,
C<< File::Spec->tmpdir >> will be used.

You B<should> specify this option for productions to avoid conflicts of
template names.

=item C<< function => \%functions >>

Specifies a function map which contains name-coderef pairs.
A function C<f> may be called as C<f($arg)> or C<$arg | f> in templates.

Note that these registered functions have to return a B<text string>,
not a binary string unless you want to handle bytes in whole templates.
Make sure what you want to use returns either a text string or a binary

For example, some methods of C<Time::Piece> might return a binary string
which is encoded in UTF-8, so you'll want to decode their values.

    # under LANG=ja_JP.UTF-8 on MacOSX (Darwin 11.2.0)
    use Time::Piece;
    use Encode qw(decode);

    sub ctime {
        my $ctime = Time::Piece->new->strftime; # UTF-8 encoded bytes
        return decode "UTF-8", $ctime;

    my $tx = Text::Xslate->new(
        function => {
            ctime => \&ctime,

Built-in functions are described in L<Text::Xslate::Manual::Builtin>.

=item C<< module => [$module => ?\@import_args, ...] >>

Imports functions from I<$module>, which may be a function-based or bridge module.
Optional I<@import_args> are passed to C<import> as C<< $module->import(@import_args) >>.

For example:

    # for function-based modules
    my $tx = Text::Xslate->new(
        module => ['Digest::SHA1' => [qw(sha1_hex)]],
    print $tx->render_string(
        '<: sha1_hex($x).substr(0, 6) :>',
        { x => foo() },
    ); # => 0beec7

    # for bridge modules
    my $tx = Text::Xslate->new(
        module => ['Text::Xslate::Bridge::Star'],
    print $tx->render_string(
        '<: $x.uc() :>',
        { x => 'foo' },
    ); # => 'FOO'

Because you can use function-based modules with the C<module> option, and
also can invoke any object methods in templates, Xslate doesn't require
specific namespaces for plugins.

=item C<< html_builder_module => [$module => ?\@import_args, ...] >>

Imports functions from I<$module>, wrapping each function with C<html_builder()>.

=item C<< input_layer => $perliolayers // ':utf8' >>

Specifies PerlIO layers to open template files.

=item C<< verbose => $level // 1 >>

Specifies the verbose level.

If C<< $level == 0 >>, all the possible errors will be ignored.

If C<< $level >= 1 >> (default), trivial errors (e.g. to print nil) will be ignored,
but severe errors (e.g. for a method to throw the error) will be warned.

If C<< $level >= 2 >>, all the possible errors will be warned.

=item C<< suffix => $ext // '.tx' >>

Specify the template suffix, which is used for C<cascade> and C<include>
in Kolon.

Note that this is used for static name resolution. That is, the compiler
uses it but the runtime engine doesn't.

=item C<< syntax => $name // 'Kolon' >>

Specifies the template syntax you want to use.

I<$name> may be a short name (e.g. C<Kolon>), or a fully qualified name
(e.g. C<Text::Xslate::Syntax::Kolon>).

This option is passed to the compiler directly.

=item C<< type => $type // 'html' >>

Specifies the output content type. If I<$type> is C<html> or C<xml>,
smart escaping is applied to template expressions. That is,
they are interpolated via the C<html_escape> filter.
If I<$type> is C<text> smart escaping is not applied so that it is
suitable for plain texts like e-mails.

I<$type> may be B<html>, B<xml> (identical to C<html>), and B<text>.

This option is passed to the compiler directly.

=item C<< line_start => $token // $parser_defined_str >>

Specify the token to start line code as a string, which C<quotemeta> will be applied to. If you give C<undef>, the line code style is disabled.

This option is passed to the parser via the compiler.

=item C<< tag_start => $str // $parser_defined_str >>

Specify the token to start inline code as a string, which C<quotemeta> will be applied to.

This option is passed to the parser via the compiler.

=item C<< tag_end => $str // $parser_defined_str >>

Specify the token to end inline code as a string, which C<quotemeta> will be applied to.

This option is passed to the parser via the compiler.

=item C<< header => \@template_files >>

Specify the header template files, which are inserted to the head of each template.

This option is passed to the compiler.

=item C<< footer => \@template_files >>

Specify the footer template files, which are inserted to the foot of each template.

This option is passed to the compiler.

=item C<< warn_handler => \&cb >>

Specify the callback I<&cb> which is called on warnings.

=item C<< die_handler => \&cb >>

Specify the callback I<&cb> which is called on fatal errors.

=item C<< pre_process_handler => \&cb >>

Specify the callback I<&cb> which is called after templates are loaded from the disk
in order to pre-process template.

For example:

    # Remove whitespace from templates
    my $tx = Text::Xslate->new(
        pre_process_handler => sub {
            my $text = shift;
            return $text;

The first argument is the template text string, which can be both B<text strings> and C<byte strings>.

This filter is applied only to files, not a string template for C<render_string>.


=head3 B<< $tx->render($file, \%vars) :Str >>

Renders a template file with given variables, and returns the result.
I<\%vars> is optional.

Note that I<$file> may be cached according to the cache level.

=head3 B<< $tx->render_string($string, \%vars) :Str >>

Renders a template string with given variables, and returns the result.
I<\%vars> is optional.

Note that I<$string> is never cached, so this method should be avoided in
production environment. If you want in-memory templates, consider the I<path>
option for HASH references which are cached as you expect:

    my %vpath = (
        'hello.tx' => 'Hello, <: $lang :> world!',

    my $tx = Text::Xslate->new( path => \%vpath );
    print $tx->render('hello.tx', { lang => 'Xslate' });

Note that I<$string> must be a text string, not a binary string.

=head3 B<< $tx->load_file($file) :Void >>

Loads I<$file> into memory for following C<render()>.
Compiles and saves it as disk caches if needed.

=head3 B<< Text::Xslate->current_engine :XslateEngine >>

Returns the current Xslate engine while executing. Otherwise returns C<undef>.
This method is significant when it is called by template functions and methods.

=head3 B<< Text::Xslate->current_vars :HashRef >>

Returns the current variable table, namely the second argument of
C<render()> while executing. Otherwise returns C<undef>.

=head3 B<< Text::Xslate->current_file :Str >>

Returns the current file name while executing. Otherwise returns C<undef>.
This method is significant when it is called by template functions and methods.

=head3 B<< Text::Xslate->current_line :Int >>

Returns the current line number while executing. Otherwise returns C<undef>.
This method is significant when it is called by template functions and methods.

=head3 B<< Text::Xslate->print(...) :Void >>

Adds the argument into the output buffer. This method is available on executing.

=head3 B<< $tx->validate($file) :Void >>

Checks whether the syntax of I<$file> is valid or invalid as Xslate.
If it detects the invalid factor, this method throws the exception.

=head2 Exportable functions

=head3 C<< mark_raw($str :Str) :RawStr >>

Marks I<$str> as raw, so that the content of I<$str> will be rendered as is,
so you have to escape these strings by yourself.

For example:

    use Text::Xslate qw( mark_raw );

    my $tx   = Text::Xslate->new();
    my $tmpl = 'Mailaddress: <: $email :>';
    my %vars = (
        email => mark_raw('Foo &lt;foo at;'),
    print $tx->render_string($tmpl, \%email);
    # => Mailaddress: Foo &lt;;

This function is available in templates as the C<mark_raw> filter, although
the use of it is strongly discouraged.

=head3 C<< unmark_raw($str :Str) :Str >>

Clears the raw marker from I<$str>, so that the content of I<$str> will
be escaped before rendered.

This function is available in templates as the C<unmark_raw> filter.

=head3 C<< html_escape($str :Str) :RawStr >>

Escapes HTML meta characters in I<$str>, and returns it as a raw string (see above).
If I<$str> is already a raw string, it returns I<$str> as is.

By default, this function will automatically be applied to all template

This function is available in templates as the C<html> filter, but you're better
off using C<unmark_raw> to ensure that expressions are html-escaped.

=head3 C<< uri_escape($str :Str) :Str >>

Escapes URI unsafe characters in I<$str>, and returns it.

This function is available in templates as the C<uri> filter.

=head3 C<< html_builder { block } | \&function :CodeRef >>

Wraps a block or I<&function> with C<mark_raw> so that the new subroutine
will return a raw string.

This function is used to tell the xslate engine that I<&function> is an
HTML builder that returns HTML sources. For example:

    sub some_html_builder {
        my @args = @_;
        my $html;
        # build HTML ...
        return $html;

    my $tx = Text::Xslate->new(
        function => {
            some_html_builder => html_builder(\&some_html_builder),

See also L<Text::Xslate::Manual::Cookbook>.

=head2 Command line interface

The C<xslate(1)> command is provided as a CLI to the Text::Xslate module,
which is used to process directory trees or to evaluate one liners.
For example:

    $ xslate -Dname=value -o dest_path src_path

    $ xslate -e 'Hello, <: $ARGV[0] :> wolrd!' Xslate
    $ xslate -s TTerse -e 'Hello, [% ARGV.0 %] world!' TTerse

See L<xslate(1)> for details.


There are multiple template syntaxes available in Xslate.


=item Kolon

B<Kolon> is the default syntax, using C<< <: ... :> >> inline code and
C<< : ... >> line code, which is explained in L<Text::Xslate::Syntax::Kolon>.

=item Metakolon

B<Metakolon> is the same as Kolon except for using C<< [% ... %] >> inline code and
C<< %% ... >> line code, instead of C<< <: ... :> >> and C<< : ... >>.

=item TTerse

B<TTerse> is a syntax that is a subset of Template-Toolkit 2 (and partially TT3),
which is explained in L<Text::Xslate::Syntax::TTerse>.

=item HTMLTemplate

There's HTML::Template compatible layers in CPAN.

L<Text::Xslate::Syntax::HTMLTemplate> is a syntax for HTML::Template.

L<HTML::Template::Parser> is a converter from HTML::Template to Text::Xslate.


=head1 NOTES

There are common notes in Xslate.

=head2 Nil/undef handling

Note that nil (i.e. C<undef> in Perl) handling is different from Perl's.
Basically it does nothing, but C<< verbose => 2 >> will produce warnings on it.


=item to print

Prints nothing.

=item to access fields

Returns nil. That is, C<< >> produces nil.

=item to invoke methods

Returns nil. That is, C<< >> produces nil.

=item to iterate

Dealt as an empty array.

=item equality

C<< $var == nil >> returns true if and only if I<$var> is nil.



Perl 5.8.1 or later.

If you have a C compiler, the XS backend will be used. Otherwise the pure Perl
backend will be used.

=head1 TODO


=item *

Context controls. e.g. C<< <: [ $foo->bar @list ] :> >>.

=item *

Augment modifiers.

=item *

Default arguments and named arguments for macros.

=item *

External macros.

Just idea: in the new macro concept, macros and external templates will be
the same in internals:

    : macro foo($lang) { "Hello, " ~ $lang ~ " world!" }
    : include foo { lang => 'Xslate' }
    : # => 'Hello, Xslate world!'

    : extern bar 'my/bar.tx';     # 'extern bar $file' is ok
    : bar( value => 42 );         # calls an external template
    : include bar { value => 42 } # ditto

=item *

A "too-safe" HTML escaping filter which escape all the symbolic characters






=head1 BUGS

Please report issues at L<>.
Patches are always welcome.

=head1 SEE ALSO



Xslate template syntaxes:




Xslate command:


Other template modules that Xslate has been influenced by:




L<Template> (Template::Toolkit)








L<> -  Enforcing Strict Model-View Separation in Template Engines


Thanks to lestrrat for the suggestion to the interface of C<render()>,
the contribution of Text::Xslate::Runner (was App::Xslate), and a lot of

Thanks to tokuhirom for the ideas, feature requests, encouragement, and bug finding.

Thanks to gardejo for the proposal to the name B<template cascading>.

Thanks to makamaka for the contribution of Text::Xslate::PP.

Thanks to jjn1056 to the concept of template overlay (now implemented as C<cascade with ...>).

Thanks to typester for the various inspirations.

Thanks to clouder for the patch of adding C<AND> and C<OR> to TTerse.

Thanks to punytan for the documentation improvement.

Thanks to chiba for the bug reports and patches.

Thanks to turugina for the patch to fix Win32 problems

Thanks to Sam Graham for the bug reports.

Thanks to Mons Anderson for the bug reports and patches.

Thanks to hirose31 for the feature requests and bug reports.

Thanks to c9s for the contribution of the documents.

Thanks to shiba_yu36 for the bug reports.

Thanks to kane46taka for the bug reports.

Thanks to cho45 for the bug reports.

Thanks to shmorimo for the bug reports.

Thanks to ueda for the suggestions.

=head1 AUTHOR

Fuji, Goro (gfx) E<lt>gfuji@cpan.orgE<gt>.

Makamaka Hannyaharamitu (makamaka) (Text::Xslate::PP)

Maki, Daisuke (lestrrat) (Text::Xslate::Runner)


Copyright (c) 2010-2013, Fuji, Goro (gfx). All rights reserved.

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