The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

NAME

HTML::YaTmpl - Yet Another Template Processor

SYNOPSIS

 use HTML::YaTmpl;

 my $t=HTML::YaTmpl->new( file=>'template.tmpl' );
 $t->evaluate( key1=>$value1,
               key2=>[$val21, $val22, ...]
               ... );
 $t->evaluate_to_file( $outputfilename,
                       key1=>$value1,
                       key2=>[$val21, $val22, ...]
                       ... );

ABSTRACT

HTML::YaTmpl aims mainly to provide a HTML template processor that saves the template writer typing.

There are general template processors like Text::Template and tools to embed perl in HTML like HTML::Embperl or HTML template processors like HTML::Template. Why have I decided to start yet another? Well, Text::Template is not really convenient when it comes to process repeating data records like HTML tables. With HTML::Embperl no professional "WEB Designer" will be able to "enhance" the pages. And HTML::Template enforces a strict division of design and programming. Thus, it enforces changes to the programming logic even if you only want to exchange a long number like 2835067264068365493 with a more human readable 2,835,067,264,068,365,493.

HTML::YaTmpl attempts to make simple things easy but complexity feasible.

DESCRIPTION

HTML::YaTmpl follows the object oriented paradigm, i.e. you have to create a template processor prior to using it. A template processor is not bound to any particular template. You can use one processor to evaluate various templates. But other properties like error handling are bound to the processor.

Constructor

new( attrname=>attrval, ... )

creates a new HTML::YaTmpl object. These attributes can be set:

template [optional]

is the actual template text. It is used for programmatically generated templates. It is overridden by template() and open() methods.

file [optional]

sets a file to read the template from. Setting this attribute via new() causes the open() method to be called. If open() fails new() will return undef instead of an object.

path [optional]

is an ARRAY (passed as reference) to be used as template search path. If omitted the environment variable HTML_TMPL_SEARCH_PATH is split by your path separator (see the Config module documentation) (on UNIX a ':' on Windows ';'). The path attribute is used by open().

package [optional]

specify a package to evaluate template code fragments. If omitted the caller package of the constructor is used. Thus, the package can be used to define convenience functions to be called from within a template.

no_eval_cache [optional]

HTML::YaTmpl wraps all code fragments from templates into subroutines and calls then these subroutines. Normally the subroutines are cached to avoid multiple calls of the perl interpreter. If no_eval_cache is set the cache is turned off. This can be useful for long running applications that process user provided templates. On the other hand there is the method clear_cache() to clear that cache from time to time.

onerror [optional]
eprefix [optional]
errors [optional]

see ERROR HANDLING below

compress [optional]

This attribute let you choose a compression method given by a file name suffix. By now only gz is defined.

evaluate() returns compressed output. evaluate_to_file() writes compressed output. The output file name is extented with a suffix indicating the compression method.

Only the tailing characters of the given string are evaluated to select the compression method but the complete string is used as file name suffix. Thus:

 $t->compress='.gz' or
 $t->compress='-gz' or
 $t->compress='gz'

all select the gzip compression method. But $t->evaluate_to_file( "out.html" ) will write out.html.gz or out.html-gz or out.htmlgz respectivly.

If the constructor is called as instance method rather than as class method the new object inherits all attributes from the old instance. Naturally, explicitely named attributes are overridden.

Attribute Access Methods

These methods have got the :lvalue attribute. They can be assigned using $self->method=$new_value syntax. Called without parameter they return the actual attribute value. Calling with parameter sets the attribute value.

template
file
path
package
no_eval_cache
compress

see new() above

onerror
eprefix
errors

see ERROR HANDLING below

Template Evaluation Methods

evaluate($private_data, key1=>$val1, key2=>$val2, ...)

This function evaluates the current template (set via template() or open()) and returns the result. Note, simply setting the current file via $self->file=$newfile will not change the current template. Only after a call to the open() method the template is changed.

$private_data is an optional argument that can be used to pass additional data to and from the template. From within the template it is accessible as $p.

All other parameters are key=>value pairs that provide variables to the template. Internally they are gathered in a hash that is accessible from within the template as $h.

evaluate_to_file($file, \%private_data, key1=>$val1, ...)

This function calls evaluate() and writes the result to $file. If $file is a GLOB reference the evaluation result is written to that file handle. If it is a CODE reference the referenced function is called with the evaluation result as $_[0]. The return value is returned. If $file is an object and the object UNIVERSAL::can print() then that print method is called. Otherwise $file is interpreted as the name of a file where the evaluation result is to be written. If the template evaluation throws an exception $file remains untouched. evaluate_to_file() returns false if something went wrong, e.g. no space left on device or no permission to write the file.

Configuration File Evaluation Methods

evaluate_as_config($private_data, key1=>$val1, key2=>$val2, ...)

This function is similar to evaluate(). However, it treats the template as some kind of configuration file. Here is an example how it works. Given the template:

 <=key1>Value1</=key1>
 <=key2>Value2</=key2>

evaluate_as_config() will return an hash reference:

 { key1=>'Value1', key2=>'Value2' }

Note, this function is experimental.

Cache Management Methods

HTML::YaTmpl uses a cache to avoid multiple compilation of the same code fragment or multiple parsing of the same template. Simply hashing each compiled code or template fragment can lead to memory leaks for long running processes that use user provided templates. To prevent that the cache is assigned a high and a low water mark. If the number of cache elements reaches the high water mark elements are deleted on a LRU basis so that it drops to the low water mark. The cache is not bound to a HTML::YaTmpl object but is shared by all objects.

The cache is implemented as 2 independent HASHes for compiled code and template fragments respectively. The high and low water marks are use for both of them.

clear_cache

deletes all cache entries.

cache_highwatermark

returns and sets the high water mark. It is set by default to 10000. This function is assigned the :lvalue attribute. It can thus be called HTML::YaTmpl-cache_highwatermark=$newvalue>.

cache_lowwatermark

returns and sets the low water mark. It is set by default to 5000. This function is assigned the :lvalue attribute. It can thus be called HTML::YaTmpl-cache_lowwatermark=$newvalue>.

cache_sizes

returns the actual number of cached code fragments.

Other Methods

open

opens the current file and sets it's content as the current template. You can construct a HTML::YaTmpl object without any template and later set the file and possibly the path attribute and call open() to set the template from the file's content.

open() can be called with arguments: $self->open( file=>$filename, path=>$path ); to set the current file and template path in one call. Both parameters are optional.

If the current file cannot be read for any reason open() returns undef. $! will indicate the reason. If a template file was read successfully $self is returned.

clear_errors

clears the objects error list returning it's content.

TEMPLATE SYNTAX AND EVALUATION

As for other templare processors a template is a file containing normal content and some special template sequences that will be exchanged with computed values during template evaluation. Since this is mainly a HTML template engine these sequences are chosen similar to HTML tags. There are 3 kinds of sequences:

Sequences starting with <=

are used for variable expansion.

Sequences starting with <:

are used to control further evaluation, e.g. including other template files, evaluate only parts of the template depending on some variables, ...

Sequences starting with <#

are used as comments

Template sequences as other HTML/XML sequences are opened with an opening tag and closed with a closing tag, e.g.

 <=variable_name parameter_list> body </=variable_name>

But if body is empty this sequence can be abbreviated to

 <=variable_name parameter_list/>

where the trailing /> instead of the simple > is essential. Thus, here come some valid template sequences:

<=name/>

this is probably the most used form. It simply inserts the value of the key name provided to evaluate() instead of <=name/>.

<=name type="array"/>

now the parameter type is set. It specifies that this sequence should be expanded only if the value provided to evaluate as name is a reference to a non-empty array. In that case the string consisting of all elements of the array concatenated is inserted. Otherwise the template sequence is simply deleted from the output.

<:include other_template.tmpl inherit a=b x=y/>

this is an example of a control sequence. It reads and evaluates the template file other_template.tmpl. This new evaluation inherits all variables from the current evaluation and adds 2 new variables a and x with the values b and y.

<:include other_template.tmpl><:set a>b</:set></:include>

just the same but with closing tags. The only variable other_template.tmpl sees is a.

Simple Variable Substitution

Basics

The basic form of variable substitution is <=name/> or <=name></=name>. Given the template:

 some text '<=var/>' other text

If evaluate is called as

 $t->evaluate( var=>'computed text' );

it will return

 some text 'computed text' other text

But you can provide also an ARRAY reference to evaluate instead of the scalar to achieve the same result.

 $t->evaluate( var=>['com', 'pu', 'ted', ' ', 'text'] );

Scalar, Array and Empty Processing

How does this work? If evaluate receives an ARRAY ref as a variable's value it sort of evaluates <=var/> for each array element and concatenates the results. But you can specify that a variable substitution should be done only if the provided value is of a particular type. 4 such types are available

scalar

only if the provided value is a non-empty scalar (ref() does not return ARRAY and length() returns something not equal zero) it is substituted. Otherwise an empty string is substitued.

array

only if the provided value is a non-empty array (ref() returns ARRAY and it consists of at least one element) it is substituted. Otherwise an empty string is substitued.

empty

a value is empty if it is not a scalar nor an array as described above, i.e. if it is either an empty string or an array without any element.

given

a value is given if it is a non-empty array or string. If this type is specified special list handling is turned off. It is usefull if you want for example generate a link list to array items that are processed in detail elsewhere. For example a table should be printed somewhere on a page and on top of that page a shortcut link to that table should be printed but only if the table is not empty.

The template

 some text '<=var type=scalar/>' other text

will show the result seen above only if evaluate() is called as

 $t->evaluate( var=>'computed text' );

The other case will produce

 some text '' other text

If a variable substitution is valid for multiple types they can be concatenated with a comma (,):

 <=var type=scalar,array/>

is evaluated if var is an array or a scalar but not if it's empty.

Further,

 <=var type=given>Link</=var>

evaluates to Link if var is a non-empty list or a non-empty string. Even if var is a list containing multiple values only one occurence of Link is substituted.

Modifying the substituted value on the fly

There are cases when it would be really useful to modify the substituted value a little from within the template. Imagine you want to make big numbers more readable (52345635476 should be displayed as 52,345,635,476) or you want to substitute a list (passed as array reference, see above) and don't want the elements simply be concatenated but displayed as HTML list elements (<li>element</li>). HTML::YaTmpl provides 2 ways for doing this.

Using <:/>

Given the template

 fruits comprise
 <ul>
 <=fruits><li><:/></li>
 </=fruits></ul>

the call

 $t->evaluate( fruits=>[qw{apples pears plums cherries}] );

will generate

 fruits comprise
 <ul>
 <li>apples</li>
 <li>pears</li>
 <li>plums</li>
 <li>cherries</li>
 </ul>

Now we are using the long form of variable substitution (<=var></=var>). The tag body describes the actual substitution. Within the tag body the control sequence <:/> stands for the actual value. Of course <:/> can be given several times within a substitution body:

 <=fruits><:/> and <:/> give <:/>
 </=fruits>but
 men and women give children

generates

 apples and apples give apples
 pears and pears give pears
 plums and plums give plums
 cherries and cherries give cherries
 but
 men and women give children

Using Perl code fragments

The same result can be achieved using Perl inside the <:/> control sequence:

 <=fruits><:"$v and $v give $v\n"/></=fruits>but
 men and women give children

Now we see the <:/> control sequence in action. It can contain perl code fragments that change the substituted value. The code fragment is called with $v set to the actual variable. If the evaluate() call was passed the optional $private_data parameter it is available as $p from within these code fragments. Otherwise $p points to a HASH that is created once for each evaluate() call. Thus, code fragments can communicate with each other using this hash:

 <=fruits><: ++$p->{fruitcounter} />. <:/>
 </=fruits>total: <:$p->{fruitcounter}/> fruits

produces

 1. apples
 2. pears
 3. plums
 4. cherries
 total: 4 fruits

Maybe you have noticed the total counter in the last line was generated without any surrounding <=var></=var> sequence. Yes, that works too. In this case $v is undef.

Supplying Perl fragments using code=...

By now the most examples have used the long variable substitution form (<=var>...</=var>). In most cases this is probably the *right* thing but you can use the short form even when modifying the substitution value:

 <=fruits code="<: ++$p->{fruitcounter} />. <:/>
 "/>total: <:$p->{fruitcounter}/> fruits

will produce exactly the same result as above but it's almost not readable. Whereas:

 <html><body>
 <table>
 <=fruits code="<tr><td><: ++$p->{counter} /></td><td><:/></td></tr>"/>
 <tr><td>total number of fruits</td><td><:$p->{counter}/></td></tr>
 </table>
 </body></html>

generate perfectly valid (but not easy human readable) HTML text.

You see, the code=... parameter to a <=var .../> sequence does the trick. It can comprehend any text but should in most cases be surrounded by double quote characters ("). Within this surrounding double quote and backslash characters must be quoted with backslashes, all other can. Thus:

 <=fruits code="<li><:\"\\u$v\"/></li>"/>

generates a HTML list of capitalized fruits:

 <li>Apples</li><li>Pears</li><li>Plums</li><li>Cherries</li>

However, I believe, this kind of code fragments is the wrong way. But it can be even worse. You can omit the surrounding double quotes but then you must quote almost anything except characters matching \w with backslashes. The previous example without surrounding quotes looks like:

 <=fruits code=\<li\>\<:\"\\u$v\"/\>\</li\>/>

Special list processings

When a list is to be substituted often special treatment is required for the first and the last list element or some text should be prepended or appended to the substitution result:

 <=fruits first="<: ucfirst $v/>" last=" and <:/>" code=", <:/>"/>
 are fruits.

generates

 Apples, pears, plums and cherries
 are fruits.

You see, if there are first=... and last=... parameters to a variable substitution they affect the first respectively last list element.

On the other side there are pre=... and post=... parameters. They are evaluated before the first respectivly after the last list element:

 <=fruits pre="<select name=\"fruit\"><:\"\\n\"/>"
          code="<option><:/></option><:\"\\n\"/>"
          post="</select><:\"\\n\"/>"/>

produces

 <select name="fruit">
 <option>apples</option>
 <option>pears</option>
 <option>plums</option>
 <option>cherries</option>
 </select>

Using the long substitution form these examples would look like:

 <=fruits>
 <:first><:"\u$v"/></:first>
 <:last> and <:/></:last>
 <:code>, <:/></:code>
 </=fruits>
 are fruits.

respectively:

 <=fruits>
 <:pre><select name="fruit">
 </:pre>
 <:post></select>
 </:post>
 <:code><option><:/></option>
 </:code>
 </=fruits>

I think, the <:first>, <:last>, <:prev> and <:post> semantics are intuitively clear. <:code> needs some explanation. In the previous examples using the long form the text between <=var> and </=var> has described what to substitute. That will remain to work. If no <:code> section is found all text between <=var> and </=var> save the control sequences will be used. But this results in many unnecessary newlines since the previous example without <:prev> and <:post> look like:

 <=fruits>
                                               # empty line
                                               # empty line
 <option><:/></option>
 </=fruits>

Thus, <:code> can be used for convenience.

More special list processings

Perl knows list operations such as map, grep and sort. These are also useful in templates. Why the program logic should know about the order in which a list is displayed? It must supply the list. That's it.

Just to arrange our fruits alphabetically we can write:

 <=fruits sort="$a cmp $b">
 <:code><:/>, </:code>
 <:last><:/></:last>
 </=fruits>

and get

 apples, cherries, pears, plums

But can we order reverse fruits, i.e. selppa instead of apples?

 <=fruits map="scalar reverse $_" sort="$a cmp $b">
 <:code><:/>, </:code>
 <:last><:/></:last>
 </=fruits>

give

 seirrehc, selppa, smulp, sraep

But I want:

 cherries, apples, plums, pears

ok, here comes the template:

 <=fruits map="scalar reverse $_"
          sort="$a cmp $b"
          map="scalar reverse $_">
 <:code><:/>, </:code>
 <:last><:/></:last>
 </=fruits>

But I don't like plums! Ok then:

 <=fruits grep="!/plum/i"
          map="scalar reverse $_"
          sort="$a cmp $b"
          map="scalar reverse $_">
 <:code><:/>, </:code>
 <:last><:/></:last>
 </=fruits>

results in

 cherries, apples, pears

Of course these parameters can also be written in long form and even mixed:

 <=fruits grep="!/plum/i">
 <:map>scalar reverse $_</:map>
 <:sort>$a cmp $b</:sort>
 <:code><:/>, </:code>
 <:last><:/></:last>
 </=fruits>

produces

 seirrehc, selppa, sraep

There can be as many as you like grep/sort/map fragments. The source list is first processed by the fragments passed as parameters (<=var grep=... map=... sort=... ...>) in left to right order and then by the fragments given in long form (<:sort>...</:sort>) in top down order.

Real World Example

Suppose you want to create a WEB application with an input field to put the name of a town. The user would first see a simple <input type="text"> field. He types in some characters and submits the form. Now your program matches the user input with a database. There can be 3 results. The user input can non-ambiguously match a database record or there are several matches or no match at all. In the first case your program should answer the user with a page simply showing the matching record. In the second case it should display a select field and in the 3rd the <input type="text"> field. Your template would look like:

 <=town type=empty><input type="text" name="town"></=town>
 <=town type=array pre="<select name=\"town\">"
                   post="</select>">
 <option><:/></option>
 </=town>
 <=town type=scalar><b><:/></b></=town>

and you get a text field town is passed as undef, an empty string or an empty array. A selection field is generated if town is passed as a non-empty array and the town set in bold is produced if it is passed as non-empty string.

Control Statements

Many control sequences have been shown in the previous chapter. Here the are listed again:

<:/> or <:></:>

is used to execute perl code. <: some perl code /> and <:> some perl code </:> are equivalent. Within the perl code the variables $v, $p and $h can be used. $v contains the current variable to be substituted. $p holds the private_data parameter. $h points at a HASH containing all parameters passed to the current scope (see <:eval> for an example of using it).

<:code></:code>

is used to mark the actual template code within variable substitution or <:eval> or <:for> blocks. This can be used for convenience.

<:pre></:pre>
<:post></:post>
<:first></:first>
<:last></:last>
<:map></:map>
<:grep></:grep>
<:sort></:sort>

see previous chapter.

<:for ...>...</:for>

The <:for> statement is used to evaluate a part of the template with changed parameters. For example you want to make up a HTML table that is passed as list of lists:

 $t->evaluate( fruit_colors=>[[qw{apples red/green}],
                              [qw{pears green}],
                              [qw{plums blue/yellow}],
                              [qw{cherries red}]] );

with the template

 <table>
 <=fruit_colors><:for f="<:/>">
 <:code><tr><=f><td><:/></td></=f></tr>
 </:code>
 </:for></=fruit_colors></table>

generates

 <table>
 <tr><td>apples</td><td>red/green</td></tr>
 <tr><td>pears</td><td>green</td></tr>
 <tr><td>plums</td><td>blue/yellow</td></tr>
 <tr><td>cherries</td><td>red</td></tr>
 </table>

Here the opening <=fruit_colors> begins a scope of substituting fruit_colors. As fruit_colors is an array the tag body is evaluated for each element. The tag body contains a single <:for> statement used to assign f temporarily the current value of fruit_colors, i.e. f is assigned in turn each element of the fruit_colors list. Within to <:for> statement we see the evaluation of <=f> surrounded by <tr>. As f is also an array it generated <td>'s for each element.

The <:for> statement features a template within a template. The inner template can be given as a <:code> statement or if omitted the body of the tag is used. Thus, the following template is almost equivalent to the previous one:

 <table>
 <=fruit_colors><:for f="<:/>">
 <tr><=f><td><:/></td></=f></tr>
 </:for></=fruit_colors></table>

In fact it generates an empty line before each table row.

The parameter list for the inner template (the one that is passed to the evaluate() function) is completely made anew. In our example the inner template receives only one variable: f. But there is a way to bequeath all current variables to the inner template. Just put the reserved word :inherit or :inheritparms in the parameter list of the <:for> statement. With

 <:for f="<:/>" :inherit>

the inner template would see also fruit_colors and all other outer variables.

All parameters containing an unquoted equal sign (=) are used to set up the parameter list for the include template. The simplest form of a parameter is a string like

 <:for some_fruit="plum">

Slightly more complex is passing an outer variable with an other name:

 <:for inner_fruits="<=fruits/>">

This evaluates fruits and stores the result as inner_fruits. If fruits is an array so does inner_fruits. If the substitution statement of fruits contains pre or post components inner_fruits will contain them as first / last list element. You also can surround the variable substitution with plain text. In this case each element of inner_fruits gets surrounded by this text. And you can include in the definition of one inner variable multiple outer variables. In this case each outer list variable is expanded and the resulting number of list elements is the mathematical product of the numbers of elements of all outer lists. The template (Note: The sequence <#.../> is a comment and used in this example to hide the newline in the template. For more details on comments see below.):

 <:for inner_fruits="PRE <=fruits pre=pre1
                                  post=post1
                                  grep=\"/l/\"/> <#
                   />BETWEEN <=fruits pre=pre2
                                      post=post2
                                      grep=\"/r/\"/> POST">
 <:code><=inner_fruits last="(<:/>)">(<:/>)
 </=inner_fruits></:code>

generates a total of 16 lines:

 (PRE pre1 BETWEEN pre2 POST)
 (PRE pre1 BETWEEN pears POST)
 (PRE pre1 BETWEEN cherries POST)
 (PRE pre1 BETWEEN post2 POST)
 (PRE apples BETWEEN pre2 POST)
 (PRE apples BETWEEN pears POST)
 (PRE apples BETWEEN cherries POST)
 (PRE apples BETWEEN post2 POST)
 (PRE plums BETWEEN pre2 POST)
 (PRE plums BETWEEN pears POST)
 (PRE plums BETWEEN cherries POST)
 (PRE plums BETWEEN post2 POST)
 (PRE post1 BETWEEN pre2 POST)
 (PRE post1 BETWEEN pears POST)
 (PRE post1 BETWEEN cherries POST)
 (PRE post1 BETWEEN post2 POST)

Due to the grep statements the first expansion of fruits contains only elements with an l letter, i.e. apples and plums, whereas the second expansion consists of elements with an r letter, i.e pears and cherries. As shown each fruits list is expanded with a pre and post elements. Thus, the total number of elements of the inner_fruits list is 4*4=16.

By now we have seen in this chapter assigning of simple strings and expanded arrays to inner variables. But what happens if <:for> is placed within a variable substitution scope like in the prefacing example to this chapter? The template (provided with line numbers)

  1  <:>
  2  sub fac {
  3    use Math::BigInt;
  4    my $x=shift;
  5    my $res=Math::BigInt->new(1);
  6    for( my $i=1; $i<=$x; $i++ ) {
  7      $res*=$i;
  8    }
  9    return $res;
 10  }
 11  </:><html><body>
 12  <table>
 13  <:for x="<:[map {[$_, $_**2, $_**3, fac $_]} 1..30]/>"
 14        h="<:[qw{n n^2 n^3 n!}]/>">
 15  <:code>
 16      <tr>
 17        <=h><th><:/></th> </=h>
 18      </tr>
 19    <=x>
 20      <:code>
 21        <tr>
 22          <:for y="<:/>">
 23            <:code><=y><td><:/></td> </=y></:code>
 24          </:for>
 25        </tr>
 26      </:code>
 27    </=x></:code>
 28  </:for></table>
 29  </body></html>

generates a HTML page containing a table of the first 30 square and cube numbers and factorials. The first control sequence (lines 1-11) is substituted with nothing as the value of this code fragment is undef. It just defines a fac function. Lines 13 and 14 opens a <:for> scope that is closed at line 28. Within this scope the variables x and h are valid. Both are list variables since the <:/> control sequences return ARRAY references. Within that scope at line 17 h is evaluated to a list of <th> statements. That is more or less what we have seen before. But now in line 19 a variable substitution scope is opened. That means for each element of the x list lines 21 to 25 are evaluated repeatedly. Since x consists of arrays the evaluation of <:/> produces an array. That is used in a nested <:for> scope at lines 22 to 24 to subsequently assign to a variable y each of them. Now the body of our nested <:for> can evaluate y and create a list of <td> statements.

By now we have seen parameter lists formed like

 <:for a="b" c="d" ... />

This can lead to a lot of quoting backslashes within strings. A little foreboding give the last inner_fruits example. There we had to write grep=\"/l/\" to quote the double quotes. This can be avoided using the <:set> control sequence.

 <:for>
 <:set inner_fruits>PRE <=fruits pre=pre1
                                 post=post1
                                 grep="/l/"/> <#
                  />BETWEEN <=fruits pre=pre2
                                     post=post2
                                     grep="/r/"/> POST</:set>
 <:code><=inner_fruits last="(<:/>)">(<:/>)
 </=inner_fruits></:code>
 </:for>

generates exactly the same result without quoting any character.

Well, I believe, these examples are enough to show what can be done with <:for>. BTW, all these examples are contained in the file t/5_fruits.t of the distribution. The module is tested against them.

<:eval ...>...</:eval>

works exactly the same as the <:for> statement but resulting text is evaluated again. This can be useful in some rare cases, e.g.:

  1  <:for fruits="<:[qw/apple pear/]/>"
  2        books="<:['The Silmarillion',
  3                  'The Lord of the Rings',
  4                  'The Hobbit or There And Back Again']/>"><#
  5  /><:eval what="<:[qw/book fruit/]/>"><#
  6  /><=what><#
  7  /><=<:/>s>
  8  <<#/>:pre>Select a <: ucfirst $v />:
  9  <select name="<:/>"><<#/>/:pre>
 10  <<#/>:post></select>
 11  <<#/>/:post>
 12  <<#/>:code><option><<#/>:/></option>
 13  <<#/>/:code>
 14  </=<:/>s><#
 15  /></=what><#
 16  /></:eval><#
 17  /></:for>

Consider you get a lot of specialized variables that need to be treated all the same way. In the example above the outer <:for> scope built with lines 1-4 and 17 creates 2 variables. For both of them a <select> box should be generated. Of course one could write

 <=books>...</=books><=fruits>...</=fruits>

but it's also possible to let the template system generate the actual template and then evaluate it. That's what <:eval> is for. Within the <:eval> scope between lines 5 and 16 the variable what contains a list of 2 elements book and fruit. Variable substitution then occurs at lines 7, 14, 8 and 9. Lines 7 and 14 are evaluated to <=books> and <=fruits> and </=books> and </=fruits> respectively thus creating something looking very similar to a template. Further, all comments are eliminated. Here the <#/> comments are important. They prevent constructs like <<#/>:/> to be evaluated as <:/> in the first run. They sort of quote template sequences. In fact the generated template is:

 <=books>
 <:pre>Select a Book:
 <select name="book">
 </:pre>
 <:post></select>
 </:post>
 <:code><option><:/></option>
 </:code>
 </=books><=fruits>
 <:pre>Select a Fruit:
 <select name="fruit">
 </:pre>
 <:post></select>
 </:post>
 <:code><option><:/></option>
 </:code>
 </=fruits>

This template is then evaluated with the parameter list of the <:eval>'s outer scope, i.e. it sees fruits and books but not what.

One could argument that the same result can be achieved avoiding an intermediate template. In fact

  1  <:for fruits="<:[qw/apple pear/]/>"
  2        books="<:['The Silmarillion',
  3                  'The Lord of the Rings',
  4                  'The Hobbit or There And Back Again']/>"><#
  5  /><:for what="<:[qw/book fruit/]/>" :inherit><#
  6  /><=what><#
  7  />Select a <: ucfirst $v />:
  8  <select name="<:/>">
  9  <:for el="<:$h->{$v.'s'}/>" :inherit><#
 10  /><=el>
 11  <:code><option><:/></option>
 12  </:code>
 13  </=el></select>
 14  </:for><#
 15  /></=what><#
 16  /></:for><#
 17  /></:for>

produces exactly the same output. But it needs the special variable $h to do the trick. Anyhow, <:eval> is useful if parts of a template are fetched from a database for example.

<:include file .../> or <:include file>...</:include>

evaluates another template and inserts the result. The include directive expects a parameter list containing at least on item without an unquoted equal sign (=). The first such element is used as name of the template file. Templates are searched according to the given search path.

All other parameters are used to make up the parameter list for the include template, see <:for> above. The tag body is ignored save <:set>...</:set> statements which are also used to form the parameter list.

<:cond ...><:case ...>...</:case>...</:cond>

is used to allow conditionally evaluation of templates. The statement somehow resembles Lisp's cond statement or C's switch. The template

  1  <:for>
  2  <:set goods><:
  3  [
  4   [apple=>'300'],
  5   [pear=>'90'],
  6   [cherry=>'82'],
  7   [plum=>'120'],
  8  ]
  9  /></:set>
 10  <:code><=goods pre="<table>" post="
 11  </table>">
 12  <tr><:for x="<:/>"><=x><td><:/></td></=x><td></:for><:cond>
 13  <:case "$v->[1]>150"><b>very expensive</b></:case>
 14  <:case "$v->[1]<100"><b>bargain</b></:case>
 15  <:case 1>normal prize</:case>
 16  </:cond></td></tr></=goods></:code>
 17  </:for>

generates a table of goods marking a few of them as very expensive or bargain buy depending on their prizes. The <=goods> statement at line 10 opens a variable substitution scope. Within this scope at line 12 starts a <:cond> statement thus $v references to one list element. The <:case> statements at lines 13, 14 and 15 are then evaluated top down. The first <:case>'es body whose condition evaluates to true builds the value of the whole <:cond> statement.

The example above shows <:cond> without parameters. If it is called with parameters they name variables that are to be used in the <:case> conditions. Thus you can compare more than one variable in <:case>es.

Consider you often want to display thumbnails and only some of them should link to the originals. It would be useful to write an include template that can be called:

 <:include thumb.tmpl name=img321 link=yes/>

to create a thumbnail that links to it's original or

 <:include thumb.tmpl name=img322/>

to create a thumbnail without a link.

thumb.tmpl could look like:

  1  <:cond link>
  2  <:case "$link eq 'yes'">
  3    <a href="orig/<=name/>.jpg><img src="<=name/>.jpg"></a>
  4  </:case>
  5  <:case 1>
  6    <img src="<=name/>.jpg">
  7  </:case>
  8  </:cond>

Now the <:cond> at line 1 names link. Thus $link can be used at line 2 in the <:case: condition.

<:m name .../> or <:m file>...</:m> or =item <:macro name .../> or <:macro file>...</:macro> =item <:defmacro name>...</:defmacro>

<:defmacro> defines a macro. For example, one can define a macro as

 <:defmacro td><td align="center"><=val/></td></:defmacro>

Later on it can be invoked as

 <:m td val="..."/> or <:macro td val="..."/>

Our previous example then can be written as

  1  <:defmacro td><td align="center"><=val/></td></:defmacro><#
  2  /><:defmacro tr><tr><:for x="<:/>"><#
  3  /><=x><:m td val="<:/>"/></=x></:for><:macro td>
  4  <:set val><:cond>
  5  <:case "$v->[1]>150"><b>very expensive</b></:case>
  6  <:case "$v->[1]<100"><b>bargain</b></:case>
  7  <:case 1>normal prize</:case>
  8  </:cond></:set>
  9  </:macro></tr> </:defmacro><#
 10  /><:for>
 11  <:set goods><:
 12  [
 13   [apple=>'300'],
 14   [pear=>'90'],
 15   [cherry=>'82'],
 16   [plum=>'120'],
 17  ]
 18  /></:set>
 19  <:code><=goods pre="<table>" post="
 20  </table>">
 21  <:m tr/></=goods></:code>
 22  </:for>

Line 1 defines a macro named td that prints one HTML table cell. It uses the variable val. Lines 2 to 9 define a macro called tr that implements a table row. It is designed to be used in a substitution scope as it assigns the current value <:/> to a variable x to build a <:for> scope. In line 3 the macro defined in line 1 is used twice, once as <:m/> and once as <:macro>. Then in line 21 the tr macro is invoked and generates all table rows.

Macros are bound to a HTML::YaTmpl instance. That means you can define a set of macros in one template, evaluate it and then evaluate another template with the same instance using macros defined in the first template.

<:set></:set>

is used to make up the parameter list for <:for>, <:eval>, <:include> and <:macro> statements. Outside one of these scopes or inside a <:code> segment <:set> can be used to manipulate the current parameter list.

Thus, the example above can be rewritten as:

  1  <:set goods><:
  2  [
  3   [apple=>'300'],
  4   [pear=>'90'],
  5   [cherry=>'82'],
  6   [plum=>'120'],
  7  ]
  8  /></:set><=goods pre="<table>" post="
  9  </table>">
 10  <:m tr/></=goods>

reusing the macros as described above.

Commenting templates

Writing templates is very similar to writing programs. And as programs should contain comments to be maintainable so do templates. You can even put POD sections into a template and generate documentation using POD translators.

When thinking about comments within templates I first thought that using

 <:# any comment/>

or

 <:># any comment</:>

would be appropriate. It invokes the perl interpreter to evaluate just

 # any comment

which is a perl comment and evaluates to undef. Although it works it can slow down template eveluation. Hence I decided to let comments look like

 <# this is a comment />

or

 <#> this is a comment </#>

Thus, to put a longer comment (POD section) into your template surround it with <#> and </#>:

 <#>
 
 =head1 NAME
 
 My very cute template
 
 ...
 
 </#>

ERROR HANDLING

There are various error conditions that can occur while evaluating a template. Operating system errors like Template File not found or No space left on device can be recognized by the return codes of open() or evaluate_to_file(). But what to do if the evaluation of a template code fragment dies. Abort the whole process, insert nothing in place of the failed fragment, ...? Well, that's the subject of this chapter.

There are 3 object attributes to specify what to do in case of an error during template evaluation.

onerror

defines what to do if a template code fragment dies during compilation or at runtime. onerror can be assigned the following values:

"warn"

report the error via a warn statement, continue template evaluation and insert nothing instead of the failed fragment.

"die"

report the error via a die statement aborting further temlpate evaluation.

"output"

continue template evaluation and insert the error message instead of the code fragment.

a CODE reference

the specified function is called in scalar context receiving the actual error message as $_[0]. It's return value is inserted instead of the code fragment. If the function dies the whole template evaluation is aborted.

In either case the error message is appended to list of errors occured during template evaluation. This list can be retrieved after evaluate() has returned via the errors() method.

eprefix

That string will be prepended to each error message. This can be used to distinguish between errors from different object using all the same error list.

errors

contains the reference to the object's error list. This list is empty after a new object is created. If an error occur while evaluating a template with that object the error message is appended to that array. The error list is not cleared between calls to evaluate(). Thus, if an object is used to evaluate multiple templates or one templates several times it is recommended to clear this list between evaluate() using either clear_errors() or $self->errors=[].

TODO

  • Since our parse function can parse our templates why not use it to read configuration files or even files that contain data to be filled in other templates?

  • output options: filters, gzip compression, ...

  • plugins. fetch variable content directly from a database

  • more testing

  • mod_perl support. (maybe it is only a matter of testing)

  • optional using Safe compartments to evaluate template code fragments.

  • It would be nice if some native speaker could correct my english.

CAVEATS

The template parser is completely based on perl regular expressions. Though I have spent some effort to let the parsing process finish quickly even if the template is erroneous, it is certainly possible to construct malicious templates that will be parsed infinitely. Please let me know if you find such a template. I will try to fix it.

AUTHOR

Torsten Förtsch <Torsten.Foertsch@gmx.net>

COPYRIGHT

Copyright 2003 Torsten Förtsch.

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