I'm designing a class that defines a highly complex object with a ton (50+) of mostly optional parameters, many of which would have defaults (eg: $type = 'foo'; $width = '300'; $interactive = false;
). I'm trying to determine the best way to set up the constructor and instance/class variables in order to be able to:
In light of the above, I don't want to be passing the constructor a ton of arguments. I will be passing it a single hash which contains the initialization values, eg: $foo = new Foo(array('type'=>'bar', 'width'=>300, 'interactive'=>false));
In terms of coding the class, I still feel like I would rather have...
class Foo {
private $_type = 'default_type';
private $_width = 100;
private $_interactive = true;
...
}
...because I believe this would facilitate documentation generation (you get the list of the class' properties, which lets the API user know what 'options' they have to work with), and it "feels" like the right way to do it.
But then you run into the problem of mapping the incoming parameters in the constructor to the class variables, and without exploiting the symbol table, you get into a "brute force" approach which to me defeats the purpose (though I'm open to other opinions). E.g.:
function __construct($args){
if(isset($args['type'])) $_type = $args['type']; // yuck!
}
I've considered creating a single class variable that is itself an associative array. Initializing this would be really easy then, e.g.:
private $_instance_params = array(
'type' => 'default_type',
'width' => 100,
'interactive' => true
);
function __construct($args){
foreach($args as $key=>$value){
$_instance_params[$key] = $value;
}
}
But this seems like I'm not taking advantage of native features like private class variables, and it feels like documentation generation will not work with this approach.
Thanks for reading this far; I'm probably asking a lot here, but I'm new to PHP and am really just looking for the idiomatic / elegant way of doing this. What are your best practices?
Addendum (details about this particular Class)
It's quite likely that this class is trying to do too much, but it is a port of an old Perl library for creating and processing forms. There's probably a way of dividing the configuration options to take advantage of inheritance and polymorphism, but it may actually be counter-productive.
By request, here is a partial listing of some of the parameters (Perl code). You should see that these don't map very well to sub-classes.
The class certainly has getters and setters for many of these properties so the user can over-ride them; the objective of this post (and something the original code does nicely) is to provide a compact way of instantiating these Form objects with the required parameters already set. It actually makes for very readable code.
# Form Behaviour Parameters
# --------------------------
$self->{id}; # the id and the name of the <form> tag
$self->{name} = "webform"; # legacy - replaced by {id}
$self->{user_id} = $global->{user_id}; # used to make sure that all links have the user id encoded in them. Usually this gets returned as the {'i'} user input parameter
$self->{no_form}; # if set, the <form> tag will be omitted
$self->{readonly}; # if set, the entire form will be read-only
$self->{autosave} = ''; # when set to true, un-focusing a field causes the field data to be saved immediately
$self->{scrubbed}; # if set to "true" or non-null, places a "changed" radio button on far right of row-per-record forms that indicates that a record has been edited. Used to allow users to edit multiple records at the same time and save the results all at once. Very cool.
$self->{add_rowid}; # if set, each row in a form will have a hidden "rowid" input field with the row_id of that record (used primarily for scrubbable records). If the 'scrubbed' parameter is set, this parameter is also automatically set. Note that for this to work, the SELECT statement must pull out a unique row id.
$self->{row_id_prefix} = "row_"; # each row gets a unique id of the form id="row_##" where ## corresponds to the record's rowid. In the case of multiple forms, if we need to identify a specific row, we can change the "row_" prefix to something unique. By default it's "row_"
$self->{validate_form}; # parses user_input and validates required fields and the like on a form
$self->{target}; # adds a target window to the form tag if specified
$self->{focus_on_field}; # if supplied, this will add a <script> tag at the end of the form that will set the focus on the named field once the form loads.
$self->{on_submit}; # adds the onSubmit event handler to the form tag if supplied
$self->{ctrl_s_button_name}; # if supplied with the name of the savebutton, this will add an onKeypress handler to process CTRL-S as a way of saving the form
# Form Paging Parameters
# ----------------------
$self->{max_rows_per_page}; # when displaying a complete form using printForm() method, determines the number of rows shown on screen at a time. If this is blank or undef, then all rows in the query are shown and no header/footer is produced.
$self->{max_pages_in_nav} = 7; # when displaying the navbar above and below list forms, determines how many page links are shown. Should be an odd number
$self->{current_offset}; # the current page that we're displaying
$self->{total_records}; # the number of records returned by the query
$self->{hide_max_rows_selector} = ""; # hide the <select> tag allowing users to choose the max_rows_per_page
$self->{force_selected_row} = ""; # if this is set, calls to showPage() will also clear the rowid hidden field on the form, forcing the first record to be displayed if none were selected
$self->{paging_style} = "normal"; # Options: "compact"
We can, of course, allow ourselves to be drawn into a more lengthy debate around programming style. But I'm hoping to avoid it, for the sanity of all involved! Here (Perl code, again) is an example of instantiating this object with a pretty hefty set of parameters.
my $form = new Valz::Webform (
id => "dbForm",
form_name => "user_mailbox_recip_list_students",
user_input => \%params,
user_id => $params{i},
no_form => "no_form",
selectable => "checkbox",
selectable_row_prefix => "student",
selected_row => join (",", getRecipientIDsByType('student')),
this_page => $params{c},
paging_style => "compact",
hide_max_rows_selector => 'true',
max_pages_in_nav => 5
);
I can think of two ways of doing that. If you want to keep your instance variables you can just iterate through the array passed to the constructor and set the instance variable dynamically:
<?php
class Foo {
private $_type = 'default_type';
private $_width = 100;
private $_interactive = true;
function __construct($args){
foreach($args as $key => $val) {
$name = '_' . $key;
if(isset($this->{$name})) {
$this->{$name} = $val;
}
}
}
}
?>
When using the array approach you don't really have to abandon documentation. Just use the @property annotations in the class body:
<?php
/**
* @property string $type
* @property integer $width
* @property boolean $interactive
*/
class Foo {
private $_instance_params = array(
'type' => 'default_type',
'width' => 100,
'interactive' => true
);
function __construct($args){
$this->_instance_params = array_merge_recursive($this->_instance_params, $args);
}
public function __get($name)
{
return $this->_instance_params[$name];
}
public function __set($name, $value)
{
$this->_instance_params[$name] = $value;
}
}
?>
That said, a class with 50 member variables is either only used for configuration (which can be split up) or it is just doing too much and you might want to think about refactoring it.