LayoutProcessor

PHP class for template processing with extensible scripting language.

This class can process templates conforming to a simple syntax based on indented blocks of lines.

The first character of a line is used to determine what kind of block it is, this first character is called the prefix. If it is a space (or TAB) character, it is an indented line and it belongs to the same block as the previous line. Blank lines are ignored.

A set of prefix characters has builtin support, you can override these and/or add your own prefixes.

The main building blocks of a script is the "layout", which is similar to procedures in other languages. You define layouts and call them by their name, optionally with parameters. Layout names may contain spaces and other special characters, except colon, which is used to separate the parameter from the layout name.

Builtin prefixes

The following prefixes are builtin:

• # Comments
• $ Variable assignments
• " String output (variables resolved)
• ' Literal output (variables not resolved)
• = Layout definitions
• < Markup
• ! Commands

You can add your own prefix or override any existing prefix with the define_prefix($prefix,$callback) method.

Builtin commands

The following commands are builtin:

• !php Embedded PHP code
• !if/!elseif/!else Conditional statements
• !loop/!while/!break/!continue Loops
• !scope Variable scopes
• !param Parameter handling
• !return Exit current layout

You can add custom commands with the define_command($cmd,$callback) method. You can not directly override builtin commands, but you can add aliases with the define_command_alias($alias,$aliased_command) method, this way you could define a builtin command to be an alias for your custom version of that command. Overriding loop commands or conditonal statements will probably not work well because of how these are handled internally.

There are two predefined aliases: !elif is an alias for !elseif and !foreach is an alias for !loop.

Basic examples

# Hello world
echo LayoutProcessor::run_script('"Hello world\n');

# Hello world using layout
$script = <<<'EOD'
=Hello:"Hello $$\n
Hello:world
EOD;
echo LayoutProcessor::run_script($script);

# Hello world using layout and variable
$script = <<<'EOD'
=Hello:
  !param string:$who
  "Hello $who\n
$who = 'world'
Hello:$who
EOD;
echo LayoutProcessor::run_script($script);

# Because 'Hello' is now defined we can call it directly:
echo LayoutProcessor::run_layout('Hello','world');

Extending the class

In general you would not use the LayoutProcessor class directly like in the basic examples above. Instead you would write a subclass and extend it to fit more closely with your own application. There are some parts you most likely would want to override.

Class constants

The following class constants can be overridden.

ERR_MSG_INTRO

This is a constant which is used to prefix all error messages. The default value is 'Layout processing error: ', you can change it to include your application name, like 'MyApp ERROR: ' or similar. Note that this constant is used both for ERR_TEXT and ERR_HTML output (see error handling), it should not contain any HTML. It is not used for the message sent to the logger callback.

PARAM_PLACEHOLDER

The placeholder for literal parameters is defined in this constant, default value is '$$'. There is usually no reason to override this, but you can.

MAX_RECURSION_DEPTH

The default maximum recursion depth is set to 255, this should be plenty in most cases, but you can change it if you need to. You will get a 'Recursion error' message when this limit is exceeded. It usually means there is a logical error in the code resulting in a loop: some layout A is calling B, which again is calling A.

Methods

Most methods can be overridden, but the following methods are the most usefull to override.

load($layout_name)

This is the most important method to override.

This method is executed when an undefined layout is called. It should be aware of the current context of the application and load the layout from a repository (file system or database) depending on this context. For instance if a web application has current URL /foo/bar/ you should first look for the layout in /foo/bar, if not found you should look in /foo and if it is not found there either you should look in / (root). This way any layout can be overridden for different parts of the application.

If the layout is not found this method must return false, an error message is generated from the run_script() method. When the layout is found this method must return an associative array with the following keys:

• content - the actual layout as a string (required)
• name - usually the layout name, could be a file name or other name (optional)
• parent - context information, file path or DB reference (optional)
• id - DB identifier, numeric or string (optional)

The optional parts of this array is only used for context for error messages. You can add more meta information in this layout if you need to, it is stored in the static::$layouts array with $layout_name as key.

For layouts which are defined within other layouts parent gets the name of the layout in which it is defined and id gets the line number.

get($layout_name)

This method is responsible for finding the layout when it is called. It calls load() if it can not find the layout in memory. If load() fails (returns false) then get() also must return false, otherwise it must return the layout as a string, this will normally be the content of the array returned from load().

You can override this if you want to insert debugging (see example below) or if you want to handle some layouts differently from others. If this method returns false an error will be triggered in run_script().

Extension example

define('DEBUG_MODE',0);

class MyApp extends LayoutProcessor {
  const ERR_MSG_INTRO = 'MyApp error: ';
  static function current_context() {
    # return current application context object
  }
  static function load($layout_name) {
    $context = static::current_context();
    $layout_item = $context->get_layout($layout_name);
    while(!$layout_item && $context = $context->parent_context()) 
      $layout_item = $context->get_layout($layout_name);
    if(!$layout_item) return false;
    return array(
      'content' => $layout_item->content,
      'name' => $layout_item->name,
      'parent' => $layout_item->parent,
      'id' => $layout_item->id);
  }
  static function get($layout_name) {
    if(DEBUG_MODE)
      MyLoggerClass::log('DEBUG',str_repeat('  ',count(static::$scope)).$layout_name);
    return parent::get($layout_name);
  }
}

$error_mode = MyApp::ERR_LOG;

if(DEBUG_MODE) # enable HTML errors
  $error_mode |= MyApp::ERR_HTML;

MyApp::on_error($error_mode);
MyApp::set_logger(array('MyLoggerClass','log'));
MyApp::run_layout('_init'); # setup, no output
echo MyApp::run_layout('_main'); # main application start 

In this example it is presumed that you have a MyLoggerClass with a static log() method and a context object with get_layout() and parent_context() methods. What exactly a context is depends on the application, for instance it can be based on a file path or from a hierarchy stored in a database. Other criteria can also be used to define a context, like if the user is logged in or not, if it is an admin user or not, what time of day it is, the IP address of the user and so on.

The '_init' and '_main' layouts are also just examples, you can call them anything and it does not have to be two separate startup layouts like this, it could be three or one. The point of having more than one is that it gives you more flexibility when it comes to overriding defaults. For instance '_init' could be used to load required PHP libraries and to set global variables and similar, and '_main' could be used to define the page template. You would allways use the same '_init', but '_main' could be overridden for different types of pages (web/mobile/ajax etc). Both could contain additional calls to layouts which may be overridden for different pages or different parts of the application.

Standard prefixes

Comments

Comments are prefixed with a # character. They produce no output, they are just used to document the script/template.

# This is a comment.
  Comments can span multiple lines if the lines are indented.

In some cases you can have single line comments on the same line as other statements, for instance after assignments if you use a semicolon after the expression:

$foo = 'bar';  # this is a valid comment

Assignments

The $ prefix is used for variable assignment. Any valid PHP syntax can be used, the semicolon to end the statement is optional unless it is followed by a comment. Big expressions can span multiple lines, just make sure they are indented.

$x = 100
$y = 200;  # comment allowed here
$z = SOME_CONSTANT
$foo = functionCall()
$bar = $obj->method()
$str = "x=$x".
  ($foo ? ' foo='.$foo : '').
  '<br>'
$coord = array($x,$y)
$coord = [$x,$y];  # requires PHP 5.4+
$avg = ($x+$y) / 2
$obj->prop = 1
$arr[] = 'new item'
$arr[2] = 'changed'
$str[0] = 'X'

Some statements which are not "normal" assignments are also allowed, for instance:

$x++
$x *= 2
$str .= 'more'

The following two special cases are also allowed, they execute but any return value is ignored:

$func()
$obj->method()

If you need the return values you must assign them to variables, like this:

$res = $func()
"$res
$res = $obj->method()
!if $res:
  DoSomething

NOTE: Variable names follow standard PHP rules for variables, the name must start with a letter or underscore and can contain letters, underscores and digits. In this context letters also include any character in the range 127-255, meaning national special characters like ζψειόρ and so on are also accepted.

String output

The " prefix is used for string output. The block is output after resolving variables and escape sequences. This works very similar to PHP double quoted strings, except you do not have to escape double quotes inside the string, and there is no double quote at the end.

"Hello world\n
"<div style="width:$width">
  $content
  </div>
"$var

Literal output

The ' prefix is used for literal output. The block is output as is, without resolving variables or escape sequences. It can contain anything including unescaped ' characters.

'This is literal output, '$foo' is just '$foo', variables are not resolved
'The output can span multiple 
  lines if they are indented.
  The indentation is kept in the output.

Layout definitions

The = prefix is used to define new or to override existing layouts. A layout can be very simple, defined on a singe line, or it can be quite complex and large. There is no defined limit to the size. You can define layouts within other layouts, but they are all global, just like PHP functions.

=greeting:"Hello!\n
=alert:!param string: $msg
  <div class="alert alert-danger">
  <span class="glyphicon glyphicon-exclamation-sign" style="font-size:150%;color:red"></span>
  " $msg
  </div>

A layout with a single statement can be defined on one line, but care must be taken when it is a single multiline statement. This will fail:

=foo:<p>This paragraph 
        spans two lines</p>

The parser can not distinguish between a single multiline statement and multiple statements. It will treat the second line as a statement and fail with the message Undefined layout "spans two lines</p>"

The solution is to start the multiline statement on a new indented line:

=foo:
  <p>This paragraph 
     spans two lines</p>

When there are multiple statements this is not a problem. This works:

=foo:<p>This paragraph 
        spans two lines</p>
     <p>Another paragraph</p>

It works because the second line is indented more than the third. This would have failed:

=foo:<p>This paragraph 
     spans two lines</p>
     <p>Another paragraph</p>

Layout names must start with a letter (a-z) or an underscore and can contain any characters except colon.

Markup

The < prefix is used to output markup. This is similar to ' (literal output), it does not resolve variables or escape sequences. Though it is quite primitive, it is very useful when writing scripts which produce HTML or XML output.

=MainMenu:
  <ul class="menu">
  MenuItems
  </ul>
<div class="MenuContainer">
MainMenu
</div>
<!-- This comment will be visible in the HTML source -->
# This comment will not be visible
<p class="example">
  This is a paragraph.
  It can be split into multiple indented lines, 
  the browser will format it as a normal paragraph
  depending on the available width on screen.
</p>

See also the =alert example above.

Commands

The ! prefix is used for commands. Some commands are builtin, and you can add your own using the define_command($cmd,$callback) method. The name of the command must start with a letter or an underscore and can contain letters, digits, underscores and dashes.

!php Embedded PHP code

This command is used to embed native PHP code in your script. It can be a single statement or a larger block of code, for instance a function call or a class or function definition.

!php var_dump($var);
!php include($php_file);
!php function foo() { return 'bar'; }
!php
  class FooBar {
    function paragraph($param) {
      return "<p>$param</p>";
    }
  }

Variables in the current layout are automatically made available in the !php block, but a variable created inside the block is not automatically exported to the layout.

The PHP code is evaluated within a method of the class, this means you have access to the internal functions and variables trough the static keyword.

!php static::add_transform('utf8_encode','utf8_encode');
!php
  $pscope = static::parent_scope();
  $caller_layout = $pscope['layout_name'];

By default any output generated by the embedded PHP is output in the layout, this can be cancelled if you return false from the block or if you return anything other than NULL, then the return value is output instead of any output generated from the PHP code. This return value must be possible to convert to a string, which means it can not be an array and it can only be an object if the object has a __toString() method.

!if/!elseif/!else Conditional statements

These commands are used for conditional execution. !elseif and !else can only be used immediatly after an !if or an !elseif. You can have multiple !elseif but only one !else. You can have nested !if inside another !if. How deep you can nest is limited only by MAX_RECURSION_DEPTH (default 255). Unlike PHP the expression to evaluate does not need to be in parentheses, but it must be a valid PHP expression. Long expressions can be broken on multiple lines, but they must of course be indented.

After the expression a colon and a new line is required. Even for short single statement blocks you can not put the statement on the same line as the condition. Indentation is (as always) also required.

The !else statement has no condition, the provided code block is executed only if the corresponding !if conditon and all !elseif conditions are false. Unlike !if and !elseif it allows a statement on the same line, see examples of this below.

!if $foo == 'bar':
  FooBar
  
!if $obj->method():
  "Ok!
!else: "Failed!

!if $height > 400:
  !if $width > 800:
    HighVeryWideOutput
  !elseif $width > 400:
    HighWideOutput
  !else: HighNarrowOutput
!else:
  !if $width > 400:
    WideOutput
  !else: SmallOutput

NOTE: !elif is defined as an alias for !elseif. You can use either, but if there are errors the error messages will always report it as error in !elseif.

!loop/!while/!break/!continue Loops

These statements are used for making loops. The !loop is similar to the PHP foreach statement, it takes an array expression followed by the keyword as and a variable assignment or a key/value assignment. Like !if and !elseif the expression must end with colon and the code block must start on a new line.

# count to 5
!loop [1,2,3,4,5] as $i:
  " $i
# count to 5
!loop range(1,5) as $i:
  " $i
# list posted key/value pairs
!loop $_POST as $k => $v:
  "$k = $v<br>
# output a table
<table>
!loop $rows as $row:
  <tr>
  !loop $row as $col:
    "<td>$col</td>
  </tr>
</table>

NOTE: !foreach is defined as an alias for !loop. You can use either, but if there are errors the error messages will always report it as error in !loop.

The !while command takes an expression as first argument and continues to execute the code block until the expression is false. Like !loop, !if and !elseif the expression must end with colon and the code block must start on a new line and be indented.

# count down from 5
$x = 5
!while $x:
  " $x
  $x--

The !break command is used to exit a loop, for nested loops you can provide a number to exit more than one loop. !continue skips back to the start of the current loop and continues with the next iteration.

# output: ab123  
!loop ['a','b','c'] as $x:
  "$x
  !if $x == 'a':
    !continue
  !loop range(1,5) as $y:
    "$y
    !if $x == 'b' && $y == 3:
      !break 2
  

!scope Variable scopes

This command is used to import variables from other variable scopes. Each layout has a separate variable scope, which means variables are by default local to the current layout. With !scope you can access variables which belongs to a different layout.

There are four variants of the !scope command.

• !scope global Access global variables
• !scope caller Access variables from the calling layout
• !scope parent Access variables from the defining layout
• !scope from ... Access variables from any active layout

Example using !scope caller:

=Greeting:
  !scope caller: $who
  "Hello $who\n
$who = 'world'
Greeting
=GreetJane:
  $who = 'Jane'
  Greeting
GreetJane

Output of the above would be:

Hello world
Hello Jane

Example using !scope from ...:

=Start:Step1
=Step1: 
  $x = 42
  $y = 1
  Step2
  "end of Step1: y=$y\n
=Step2:
  !scope caller: $y
  $y++
  Step3
=Step3:
  !scope from Step1: $x,$y
  "Step3: x=$x y=$y\n
  $y++
Start

Output:

Step3: x=42 y=2
end of Step1: y=3

Note that y is 2 in Step3 because Step2 modified it before Step3 was executed, and so did Step3 so it is 3 at the end of Step1.

You can only fetch variables from active layouts, for instance in the example above Step1 could not fetch anything from Step3 because it is not active while Step1 is running: it has not started before Step2 is called, and it is finished when Step2 returns.

The global scope is the same as PHP global scope, which means you can also access variables defined as global by the PHP script which called the LayoutProcessor run_script() or run_layout() methods.

!param Parameter handling

This command can split and transform a parameter into one or more variables.

For simple layouts with only one parameter which is always handled literally (no variables) you can use the placeholder $$ to indicate where the parameter goes. The first =Hello example above shows this usage. When using this no !param command is needed for that layout.

In addition to the $$ placeholder all layouts have a "magic" variable named $_param which holds the parameter used when the layout was called. In many cases you can just use this variable, but sometimes you need to use the !param command to manipulate the parameter. A common and simple usage is !param string which is used to resolve variables in the parameter, you can see a couple of examples of that above. (The second =Hello at the start and the =alert example for layout definitions.)

string is one of many predefined transformation types. You can provide multiple transformation types in the same !param call, each is executed in order.

The formal syntax for the !param command is like this:

!param [<transformations>] [<separator> [(<count>|<min>-<max>)] ] [ : <variables>]

There can be zero or more transformations. If there is only one parameter, there is no separator. For multiple parameters there can be only one main separator, but there can be an additional separator for each part separated by the main separator, defined in the <variables> part of the command. When a <separator> is provided you can provide a <count> or a range in parentheses. If it is a count it defines how many parameters are required. If it is a range it means there are some optional parameters; the layout must be called with at least <min> and at most <max> parameters.

The <variables> part has two formats: either it is a simple comma separated list of variable names, or it is a line separated list of variable names with separate parameter definitions.

Some examples might clarify:

# Variables are resolved in $_param
!param string

# Variables are resolved and stored in local variable $str
!param string:$str

# Split parameter on colon, $_param becomes an array
!param colon

# Split parameter on colon, accept 0-2 parameters stored in 
  local variables $a and $b. 
  Use isset() to check if parameters are provided
!param colon(0-2):$a,$b

# Split on colon into 3 parts. 
  The first part is split on the | character and stored in an array named $color, 
    the first item is stored in $fg, 
    the second item is optional and stored in $bg if provided.
  The second part can be a variable (using 'string') 
    the resolved result is stored in $msg.
  The third part is split on comma and stored in an array named $dim, 
    the two members are stored in variables $width and $height
  Example parameter: red|black:$msg:200,30 
!param colon(3):
  $color: pipe(1-2):$fg,$bg
  $msg: string
  $dim: comma(2):$width,$height

Predefined transformations:

• raw - No transformation (default)
• string - Resolve variables
• expr - Resolve variables and PHP expressions
• php - Execute the parameter as PHP code
• layout - Execute the parameter as a layout

You can add custom transformations using the add_transform($name,$callback) method. Transformation names must start with a letter or an underscore and can contain letters, digits, underscores and dashes.

Separators:

• colon - Split on the : character
• semicolon - Split on the ; character
• comma - Split on the , character
• dot - Split on the . character
• amp - Split on the & character
• space - Split on the space character
• line - Split on new line (\n)
• tab - Split on TAB (\t)
• pipe - Split on the | character
• dash - Split on the - character
• plus - Split on the + character
• slash - Split on the / character

Parameters are always trimmed when they are separated, spaces and linefeeds are removed, so for instance if comma is used as separator for the foobar layout then all of these calls are equivalent:

foobar:1,2,3
foobar: 1, 2 ,3
foobar:1,
  2 , 3
foobar:
  1, 2, 3
foobar:
  1,
  2, 
  3

!return Exit current layout

This command exits the current layout and resumes execution with the next statement in the caller layout. Unlike the PHP return statement it can not return a value.

=HandlePost:
  !if !isset($_POST['email']):
    !return
  $email = $_POST['email']
  NewsletterSubscribe:$email

Error handling

By default error messages are output where errors are encountered, but the script continues to run. You can configure the error handling using the on_error($mode) method.

The default mode is useful for a system under development. When it goes to production you should disable the visual error messages and instead use ERR_LOG to write the messages to a file or a database table.

The following error mode constants controls output of the error messages:

• ERR_SILENT No error message is output unless the logger returns a message
• ERR_TEXT Error message is output as plain text (default)
• ERR_HTML Error message is output as HTML (in a <p> element)
• ERR_LOG Error message is sent to a logger callback
• ERR_NO_LOG Error message is not sent to a logger callback (default)

Use either ERR_SILENT, ERR_TEXT or ERR_HTML. When more than one is used ERR_SILENT is ignored, if ERR_HTML is used ERR_TEXT is ignored and HTML messages are output. When none of them are used ERR_SILENT is the default and no error message is output, unless ERR_LOG is used and the logger returns a message, or if ERR_DIE is used (see below). Using ERR_SILENT combined with ERR_LOG is recommended for an application in production, you don't want to show errors to the users.

The following error mode constants controls program flow when an error is encountered:

• ERR_CONTINUE Error handled according to output/log settings, continues running the script (default)
• ERR_RESUME Exits the current layout, then continues running the script
• ERR_EXIT Stops execution of run_script(), already produced results is returned
• ERR_CANCEL Stops execution of run_script(), only the error message is returned
• ERR_DIE Stops execution of the PHP script, only the error message is output

Use one of ERR_CONTINUE, ERR_RESUME, ERR_EXIT, ERR_CANCEL, ERR_DIE or none of them. If they are combined the most severe action will be taken, for instance if ERR_DIE is enabled it will die, if any of the others are enabled ERR_CONTINUE is ignored, and so on.

ERR_DIE outputs a text error message regardless of ERR_TEXT or ERR_HTML settings. It also exits the PHP script. It is usually better to use ERR_CANCEL, which returns control to the PHP script which called the run_script() or run_layout() method. It can return the error message if ERR_TEXT or ERR_HTML is enabled, but you can also check the LayoutProcessor::$error_exit static variable, it will contain the name of the layout which failed. It will be false if there was no error.

When using ERR_LOG you must also define a callback for the logger using the set_logger($callback) method. The callback expects two parameters, the context for the error and the message.

You can combine multiple flags using the | operator, this example outputs HTML messages to screen, exits the layout with the error but resumes running the script, and also writes messages to a file named debug.log:

class LP extends LayoutProcessor {}
LP::on_error(LP::ERR_HTML | LP::ERR_RESUME | LP::ERR_LOG);
LP::set_logger(function($context,$msg) {
  error_log(date('Y-m-d H:i:s')." $context: $msg\n",3,'debug.log');
  });

Note that the logger callback does not need to write to a log file, it can do anything, for instance send an email and/or write a user friendly message on a designated area of the screen. You can also use it to override the default error message by not using ERR_TEXT or ERR_HTML but instead return the formatted error message from the logger callback.

Dependencies

• Requires PHP 5.3 or later
• Using Indentation for block parsing