I am creating breadcrumbs on my simple site.
I have some helper classes. I use them like this (just example):
$Breadcrumbs = new Breadcrumbs();
$Breadcrumbs->add(new Breadcrumb("/", "page1"));
$Breadcrumbs->add(new Breadcrumb("/", "page2"));
$Breadcrumbs->add(new Breadcrumb("/", "page3"));
$breadcrumb->show(); returns this:
<ol class="breadcrumb">
<li>page1</li>
<li>page2</li>
<li class="active">page3</li>
</ol>
So, in my project I have some switch-case constructions in which I include some files.
In this files I am using $breadcrumbs->add(...). This code:
<div class="container body">
<? $Breadcrumbs->show();?>
<?
$page = isset($_GET['page']) ? $_GET['page'] : null;
switch($page):
case "suppliers":
require_once($DOCUMENT_ROOT."/modules/suppliers.php");
break;
default:
require_once($DOCUMENT_ROOT."/modules/default.php");
break;
endswitch;
?>
<? $Breadcrumbs->show();?>
</div>
gives me this result:
Well, it works like it must work. I am using $breadcrumbs->add(...) in require files after I called $breadcrumb->show() first time thats why 1st call returns blank result. 2nd call of show() is after all breadcrumbs are added, so it returns fine result.
The questions is how to output breadcrumbs before switch blocks but with right content. Maybe I need a buffer or idk?
This is a good example of why it is such a good idea to separate out logic from presentation: you have a nice abstraction for crumb links, but can't use it properly because your other code is outputting as it goes along, rather than working with abstract data.
Obviously, you could throw away your current structure and port both logic and display directly into a new framework, but assuming you want to migrate from where you are now, here's one approach:
Create an object or array that represents the "result" of whatever module is called. Replace all current use of echo or ?> with concatenation to a string called something like $results['generic_output']. This is effectively like buffering your output, and is enough to let you use your existing abstractions like $breadcrumbs at any time. At this stage, your "template" would consist mostly of echo $results['generic_output'], plus the boilerplate header and footer which is probably already gathered in one place.
Start breaking down the output into sections. Particularly look for sections which are similar on multiple pages. For instance, if you have a "sidebar" with different content on each page but similar styling, make a $results['sidebar_content'] with just the content of that sidebar; the boilerplate to lay it out can then go into your template, and you've reduced the amount of code duplication.
Make the data you pass to the template increasingly abstract, with the goal of eventually having no HTML outside of the template(s). For instance, maybe the sidebar is made up of panels; you might start with an array of HTML blocks, one for each panel, but then turn it into an array of objects based on the actual data being displayed (say, a special offer, or the customer's current basket), with a set of templates for handling different kinds of panel. Eventually, it should be theoretically possible to build a plain-text version of your site with no HTML, just by changing the template layer, and none of the original modules.
The final step is to separate decisions about what to show from decisions about what to do. Continuing with my imaginary sidebar, your template could always receive the current basket as a general variable for use somewhere on the page, rather than as "sidebar item 1". This allows you to completely separate the actions that led into a page from the output that eventually results.
I would like to stress that this is not the way to a perfect framework, or the definitive solution to your situation, but it's one way of organising existing code (and existing thinking) in the right direction.
In the above, the "templates" could just be a set of PHP files using ?> or echo to produce the output, or it could be a dedicated templating system such as Smarty or Twig. Indeed, the point of the separation is that you could change your mind on that front later, because the result of the code modules would be an array of data to be displayed, which is just what Smarty or Twig would need as input.
Related
I have been designing websites for a while now, but there is one thing that I have never been quite sure of when using PHP and HTML. Is it better to have the whole document in PHP and echo HTML like so:
<?php
doSomething();
echo "<div id=\"some_div\">Content</div>";
?>
Or have a HTML file like so and just add in the PHP:
<html>
<body>
<?php doSomething(); ?>
<div id="some_div">Content</div>
</body>
</html>
It seems tidier to echo HTML, especially if lots of PHP gets used throughout the page, but doing so loses all formatting of the HTML i.e. colors in the IDE etc.
There are varying opinions on this. I think there are two good ways:
Use a templating engine like Smarty that completely separates code and presentation.
Use your second example, but when mixing PHP into HTML, only output variables. Do all the code logic in one block before outputting anything, or a separate file. Like so:
<?php $content = doSomething();
// complex calculations
?>
<html>
<body>
<?php echo $content; ?>
<div id="some_div">Content</div>
</body>
</html>
Most full-fledged application frameworks bring their own styles of doing this; in that case, it's usually best to follow the style provided.
I think this would depend on your group's or your own decided convention. And it can and should vary depending on what type of file you're working in. If you follow the MVC pattern then your views should be the latter. If you're writing a class or some non-output script/code then you should use the former.
Try to keep a separation of display or formatting of output and the logic that provides the data. For instance let's say you need to make a quick page that runs a simple query and outputs some data. In this case (where there is no other existing infrastructure or framework) you could place the logic in an include or in the top or the bottom of the file. Example:
<?php
# define some functions here that provide data in a raw format
?>
<html>
<body>
<?php foreach($foo = data_function($some_parameter) as $key => $value): ?>
<p>
<?=$value;?>
</p>
<?php endforeach; ?>
</body>
</html>
Or you could place the logic and function definitions in an include file or at the bottom of the file.
Now if you're producing some sort of class that has output (it really shouldn't) then you would echo the HTML or return it from the method being called. Preferably return it so that it can be output whenever and however the implementer would like.
The syntax highlighting is an important benefit of the second method, as you said. But also, if you're following good practices where logic and presentation are separated, you will naturally find that your files that contain HTML are almost entirely HTML, which then, naturally, leads to your second method again. This is the standard for MVC frameworks and the like. You'll have a bunch of files that are all PHP, doing logic, and then when that's done they'll include a presentation file which is mostly HTML with a sprinkling of PHP.
Simple:
More PHP - close HTML in PHP. When you generate HTML code in PHP, when you are doing something like a class, so it is better to make it in echo.
Less PHP - close PHP in HTML. This is stuff like just putting vars into fields of HTML stuff, like forms... And such.
The best approach is to separate the HTML from the PHP using template system or at least some kind of HTML skeleton like:
<main>
<header/>
<top-nav/>
<left-col>
<body />
</left-col>
<right-col />
<footer/>
</main>
Each node represents a template file e.g. main.php, hrader.php and so on. Than you have to separate the PHP code from the templates as something like functions.php and fineally use your second approach for template files and keeping functions clean of "echos" and HTML.
If you can, use a template engine instead.
Although it is slightly easier at first to mix your HTML and PHP, separating them makes things much easier to maintain later on.
I would recommend checking out TemplateLite which is based on Smarty but is a little more light weight.
I've reached a conclusion that using views in MVC framework e.g. Laravel, Yii, CodeIgniter is the best approach even if you are not displaying the html straight away.
Inside the view do all the echoing and looping of prepared variables, don't create or call functions there, unless formatting existing data e.g. date to specific format date('Y-m-d', strtodate(123456789)). It should be used only for creating HTML, not processing it. That's what frameworks have controllers for.
If using plain PHP, create you own view function to pass 3 variables to - html file, array of variables, and if you want to get output as string or print it straight away for the browser. I don't find a need for it as using frameworks is pretty much a standard. (I might improve the answer in the future by creating the function to get view generated HTML) Please see added edit below as a sample.
Frameworks allow you to get the HTML of the view instead of displaying it. So if you need to generate separate tables or other elements, pass the variables to a view, and return HTML.
Different fremeworks may use various type of templating languages e.g. blade. They help formatting the data, and essentially make templates easier to work with. It's also not necessary to use them for displaying data, or if forced to use it by the framework, just do required processing before posting the variables, and just "print" it using something like {{ yourVariable }} or {{ yourVariable.someProperty }}
Edit: here's a plain PHP (not framework PHP) - simple-php-view repository as a sample view library that allows to generate HTML using variables. Could be suitable for school/university projects or such where frameworks may not be allowed.
The repository allows to generate HTML at any time by calling a function and passing required variables to it, similar to frameworks. Separately generated HTML can then be combined by another view.
It depends on the context. If you are outputting a lot of HTML with attributes, you're going to get sick of escaping the quotation marks in PHP strings. However, there is no need to use ?><p><? instead of echo "<p>"; either. It's really just a matter of personal taste.
The second method is what I usually use. And it was the default method for me too. It is just to handy to get php to work inside html rather than echo out the html code. But I had to modify the httpd.conf file as my server just commented out the php code.
I'm new to PHP and want to apply a specific class to the title of my page depending on what part of the site the viewer is browsing.
For instance, I want to apply the class "blog" to the if the viewer is at domain.com/blog OR domain.com/blog/post-1 so on and so forth BUT apply the class "pics" if they're viewing domain.com/pics or domain.com/pics/gallery-1 etc etc.
I found something that could be modified to serve my needs using javascript here
but I figured seeing as I'm using PHP already, it'd make more sense to keep this sort of thing server side.
As I say, I'm new to PHP. I've experimented with some regular expressions, but to no avail.
EDIT: Sorry not to be more specific in my first post, I am using wordpress as my CMS - this is my first stackoverflow post, I trust you can forgive me :)
<?php
if (substr($_SERVER['REQUEST_URI'], 0,5)=='/pics') {
$h1class='someclass';
}
The it depends on how you're putting the class in the tag, might be like this
?><h1 class="<?php echo $h1class; ?>">...
Rather than putting a class in based on the page, I would suggest having seperate CSS files for home.css, blog.css, whateverelse.css. Of course, these would be in addition to some sort of default.css or site.css or whatever that would contain the styles used across the site.
You can then build a function/method to create the CSS calls in the HTML header. I usually have a "Page" object that builds the actual HTML page, and a "get_css" method that spits out the CSS calls.
It's hard to get more specific without knowing how you currently build pages.
Here is the solution I ended up crafting. Credit to #m.buettner for pointing me towards explode().
<h1 id="title-text" class="
<?php #returns the category as a class
$url = array();
$url = explode('/', get_permalink());
echo $url[3];
?>
mono in-case-404">
SITE
</h1>
First steps are done, some pages are coded. And after looking to the result I've got the question... how much code is there should be in templates (View)?
For example, here is a template file:
<?php $this->load->view('header'); ?>
<?php $this->load->view('banner'); ?>
<div id="items">
<?php
for($i=0; $i<count($main); $i++) {
echo '<div class="item">
<div class="name">'.$main[$i]['name'].'</div>';
if($main[$i]['icq']=='') { }
else { echo '<div class="phone">'.$main[$i]['phone'].'</div>'; }
echo '</div>';
}
?>
</div>
<?php $this->load->view('footer'); ?>
Do you think there is too much code in this template or it is normal?
The first answer was actually spot on, but the user deleted it (probably due to peer pressure). Basically, you do not want any logic in your template. In an ideal world, you had a tag for all the model data, but since we are in an HTML world, there isn't, so you either have to use XSLT or use ViewHelpers.
Let's focus on the ViewHelper approach.
This is hard to maintain:
<div id="items">
<?php
for($i=0; $i<count($main); $i++) {
echo '<div class="item">
<div class="name">'.$main[$i]['name'].'</div>';
if($main[$i]['icq']=='') { }
else { echo '<div class="phone">'.$main[$i]['phone'].'</div>'; }
echo '</div>';
}
?>
</div>
And it won't get any better if you replace the PHP with Smarty. This is easy to maintain:
<div id="items">
<?php echo itemlist($items, 'template.htm') ?>;
</div>
Below the now deleted question, there was a commentor objecting this "code isn't easy to maintain for non-coders because now they don't know where itemlist is defined and what it does." but that's complete gibberish. Think about it for a second.
On the one hand, they claim non-coders will have issues with a simple function call but on the other hand they expect them to understand the mishmash of PHP code mixed with HTML. A designer does not care about the presentation logic, but just the actual presentation. The output.
A single function call clearly says: "here be an item list", which is much easier to grasp than "here is a for that echoes a div and maybe something else if icq is given". A function call is as good as a tag. It has clearly defined input and output. How it achieves that output is irrelevant to anyone but the developer.
Your ViewHelper encapsulates the presentation logic. It's a snippet you can reuse across all of your Views. That is much more maintainable than copy and pasting all that logic over and over again when needed. In the example above, there is two arguments to the helper:
$items is an array or other traversable type holding your actual item data
'template.htm' is the filename of the template used to render the data
I'll make the second one optional, because I'd assume it's always the same template anyway. But since the commentor complained the non-coder wouldn't know where to look, I felt it necessary to show how easy it is to tell the non-coder where to look.
function itemlist($items, $template = './templates/itemlist.htm') {
ob_start();
foreach($items as $item) {
include $template;
}
return ob_get_flush();
}
There might be more efficient approaches to solve the inclusion of the template. The main idea should be clear though. Hide the presentation logic from the actual template. Your "template.htm" would then look like this:
<div class="item">
<div class="name"><?php echo $item['name'] ?></div>
<?php echo contact($item, 'icq' 'phone'); ?>
</div>
No if and elses. No string concatenations or curly braces. The logic to decide how to the user can be contacted is hidden in a ViewHelper as well. All the non-coder has to know now is the arguments to the ViewHelpers and that's as easy as knowing which attribute to write to a tag. Give them a cheat sheet if necessary.
Related information:
http://martinfowler.com/eaaCatalog/templateView.html
EDIT
Due to the two comments below I decided to expand on this answer. The above is not abstraction for abstraction's sake. It is for reusability and maintainability. I've already pointed that out above but let me explain that here again.
Actually, I find it odd to object to using ViewHelpers because you'll have "have presentation in two places" but not complain about separating header, banner and footer. It's the same thing. You isolate parts that are reusable and put them into their own templates. Separating the logic from the template in that step is just the natural next step to achive even more maintainability.
A view template that contains logic is effectively a script and not a template. Any view template that contains the logic to assemble itself is doomed to repeat itself. This might not be an issue with small sites but if you are working on a site with a couple dozen or even hundreds of view and widgets, not abstracting these parts will lead to code duplication. Put all the logic into the template and it will quickly become a tangled mess of c&p'ed markup mixed with conditionals. For any duplication, you'll double the time it takes to change it. Add inline styles and obtrusive Javascript and you are in maintenance hell.¹
If you use OOP for the other parts of your application, then why would you go procedural in your view? If you understood that you should separate Javascript and CSS from your HTML, why would you mix PHP into your template? It doesn't make sense. The OP's code snippet is a script and not a template. As such, it is the domain of the developer. And because of that you apply all the good practise you apply to other parts of your application as well. And that includes isolating logic into functions and/or classes.
Granted, a designer looking into a script might not immediately know what's going on. But then again, she might not know that either once you start to add in native PHP functions. What's mb_strimwidth doing? And substr? What's that ?: construct? The more actual PHP you add, the harder it will be to read for non-developers.
If you want to have designers work on templates, don't give them scripts. Give them templates. And if you do, isolate the logic from it and replace it with easy to grasp function calls. Use function names that clearly communicate what the function does. So the designer only needs to know if "I use this with that input, I will always get that output. I don't care how that output comes to be. I'll leave that to the developer".
Isolating the logic into functions (or classes) also gives you the immediate advantage of being able to test the logic used to render specific parts on that page in isolation. You don't have to setup the entire environment required to create a page, but just pass in the required input and assert it's output (that's why the function buffers the string instead of outputting it btw).
¹For those of you who think this is not an issue, I strongly suggest to find yourself a legacy application that truly keeps all the logic in the view template. Try to change a few things. Once you have had the pleasure of eating through 2500 lines of spaghetti code with at least 500 chars each and containing a mix of repeated PHP, HTML, CSS and JavaScript you'll know what I am talking about.
I don't think what you're trying to achieve in the view is too much logic, but I think it should be cleaned up somewhat.
Something like:
<?php foreach($main as $item): ?>
<div class="item">
<div class="name"><?php echo $item['name']; ?></div>
<?php if($item['icq']): ?>
<div class="phone"><?php echo $item['phone']; ?></div>
<?php endif; ?>
</div>
<?php endforeach; ?>
I find that a lot easier to read, and it does the same job. I also don't think it would be hard for a non-coder to modify, which is ultimately the goal of a view.
When the templates need a lot of logic is could be useful to introduce the ViewModel pattern
The logic that determines when a certain block is shown is placed in the view model.
Example
A template snippet:
<?php if($item['show_phone_number']): ?>
<div class="phone"><?php echo $item['phone_number']; ?></div>
<?php endif; ?>
A view model snippet:
$item['show_phone_number'] = $item["icq"] == '';
short answer:
As little as is viably possible.
View is no a template.
Or at least - it should not be a template. Views are generally instance of classes, which are responsible for dealing with presentational logic. They juggle multiple templates and decide which one to use and what data pass into them.
But CI has none of this. You are basically stuck with a template, which pretends to be "View" , and and instances of ActiveRecord, which they tell you are "Models". Hell .. in proper MVC the Model isn't even a specific class but an application layer.
In long term, this situation will force both presentation logic and business logic into the controller, while at the same time, producing hard-to-maintain templates and activerecord classes.
I think as James, too much code, but you can add an array variable to the get_items() function, to control the prefix and suffix of each item, something like:
$options = Array('item_prefix' => '<div class="item">', 'item_suffix' => '</div>'...)
echo get_items();
Your views should definitely contain most of the html tags, so I do not think that a function get_items should be placed somewhere else. Just clean up your templates a little bit and work with the alternative loop syntax as suggested by Karpie. You could also think about using some existing templating engine like Twig but with that the views have to contain some references to your data as well...
What you want to achieve is that non-coders can edit your html files without having to understand too much scripting logic.
I have been designing websites for a while now, but there is one thing that I have never been quite sure of when using PHP and HTML. Is it better to have the whole document in PHP and echo HTML like so:
<?php
doSomething();
echo "<div id=\"some_div\">Content</div>";
?>
Or have a HTML file like so and just add in the PHP:
<html>
<body>
<?php doSomething(); ?>
<div id="some_div">Content</div>
</body>
</html>
It seems tidier to echo HTML, especially if lots of PHP gets used throughout the page, but doing so loses all formatting of the HTML i.e. colors in the IDE etc.
There are varying opinions on this. I think there are two good ways:
Use a templating engine like Smarty that completely separates code and presentation.
Use your second example, but when mixing PHP into HTML, only output variables. Do all the code logic in one block before outputting anything, or a separate file. Like so:
<?php $content = doSomething();
// complex calculations
?>
<html>
<body>
<?php echo $content; ?>
<div id="some_div">Content</div>
</body>
</html>
Most full-fledged application frameworks bring their own styles of doing this; in that case, it's usually best to follow the style provided.
I think this would depend on your group's or your own decided convention. And it can and should vary depending on what type of file you're working in. If you follow the MVC pattern then your views should be the latter. If you're writing a class or some non-output script/code then you should use the former.
Try to keep a separation of display or formatting of output and the logic that provides the data. For instance let's say you need to make a quick page that runs a simple query and outputs some data. In this case (where there is no other existing infrastructure or framework) you could place the logic in an include or in the top or the bottom of the file. Example:
<?php
# define some functions here that provide data in a raw format
?>
<html>
<body>
<?php foreach($foo = data_function($some_parameter) as $key => $value): ?>
<p>
<?=$value;?>
</p>
<?php endforeach; ?>
</body>
</html>
Or you could place the logic and function definitions in an include file or at the bottom of the file.
Now if you're producing some sort of class that has output (it really shouldn't) then you would echo the HTML or return it from the method being called. Preferably return it so that it can be output whenever and however the implementer would like.
The syntax highlighting is an important benefit of the second method, as you said. But also, if you're following good practices where logic and presentation are separated, you will naturally find that your files that contain HTML are almost entirely HTML, which then, naturally, leads to your second method again. This is the standard for MVC frameworks and the like. You'll have a bunch of files that are all PHP, doing logic, and then when that's done they'll include a presentation file which is mostly HTML with a sprinkling of PHP.
Simple:
More PHP - close HTML in PHP. When you generate HTML code in PHP, when you are doing something like a class, so it is better to make it in echo.
Less PHP - close PHP in HTML. This is stuff like just putting vars into fields of HTML stuff, like forms... And such.
The best approach is to separate the HTML from the PHP using template system or at least some kind of HTML skeleton like:
<main>
<header/>
<top-nav/>
<left-col>
<body />
</left-col>
<right-col />
<footer/>
</main>
Each node represents a template file e.g. main.php, hrader.php and so on. Than you have to separate the PHP code from the templates as something like functions.php and fineally use your second approach for template files and keeping functions clean of "echos" and HTML.
If you can, use a template engine instead.
Although it is slightly easier at first to mix your HTML and PHP, separating them makes things much easier to maintain later on.
I would recommend checking out TemplateLite which is based on Smarty but is a little more light weight.
I've reached a conclusion that using views in MVC framework e.g. Laravel, Yii, CodeIgniter is the best approach even if you are not displaying the html straight away.
Inside the view do all the echoing and looping of prepared variables, don't create or call functions there, unless formatting existing data e.g. date to specific format date('Y-m-d', strtodate(123456789)). It should be used only for creating HTML, not processing it. That's what frameworks have controllers for.
If using plain PHP, create you own view function to pass 3 variables to - html file, array of variables, and if you want to get output as string or print it straight away for the browser. I don't find a need for it as using frameworks is pretty much a standard. (I might improve the answer in the future by creating the function to get view generated HTML) Please see added edit below as a sample.
Frameworks allow you to get the HTML of the view instead of displaying it. So if you need to generate separate tables or other elements, pass the variables to a view, and return HTML.
Different fremeworks may use various type of templating languages e.g. blade. They help formatting the data, and essentially make templates easier to work with. It's also not necessary to use them for displaying data, or if forced to use it by the framework, just do required processing before posting the variables, and just "print" it using something like {{ yourVariable }} or {{ yourVariable.someProperty }}
Edit: here's a plain PHP (not framework PHP) - simple-php-view repository as a sample view library that allows to generate HTML using variables. Could be suitable for school/university projects or such where frameworks may not be allowed.
The repository allows to generate HTML at any time by calling a function and passing required variables to it, similar to frameworks. Separately generated HTML can then be combined by another view.
It depends on the context. If you are outputting a lot of HTML with attributes, you're going to get sick of escaping the quotation marks in PHP strings. However, there is no need to use ?><p><? instead of echo "<p>"; either. It's really just a matter of personal taste.
The second method is what I usually use. And it was the default method for me too. It is just to handy to get php to work inside html rather than echo out the html code. But I had to modify the httpd.conf file as my server just commented out the php code.
Most HTML in a large website is duplicated across pages (the header, footer, navigation menus, etc.). How do you design your code so that all this duplicate HTML is not actually duplicated in your code? For example, if I want to change my navigation links from a <ul> to a <ol>, I'd like to make that change in just one file.
Here's how I've seen one particular codebase handle this problem. The code for every page looks like this:
print_top_html();
/* all the code/HTML for this particular page */
print_bottom_html();
But I feel uncomfortable with this approach (partially because opening tags aren't in the same file as their closing tags).
Is there a better way?
I mostly work with PHP sites, but I'd be interested in hearing solutions for other languages (I'm not sure if this question is language-agnostic).
I'm not a php programmer, but I know we can use a templating system called Smarty that it works with templates(views), something like asp.net mvc does with Razor.
look here http://www.smarty.net/
One solution at least in the case of PHP (and other programming languages) is templates. Instead of having two functions like you have above it would instead be a mix of HTML and PHP like this.
<html>
<head>
<title><?php print $page_title ?></title>
<?php print $styles ?>
<?php print $scripts ?>
</head>
<body>
<div id="nav">
<?php print $nav ?>
</div>
<div id="content">
<?php print $content ?>
</div>
</body>
</html>
Each variable within this template would contain HTML that was produced by another template, HTML produced by a function, or also content from a database. There are a number of PHP template engines which operate in more or less this manner.
You create a template for HTML that you would generally use over and over again. Then to use it would be something like this.
<?php
$vars['nav'] = _generate_nav();
$vars['content'] = "This is the page content."
extract($vars); // Extracts variables from an array, see php.net docs
include 'page_template.php'; // Or whatever you want to name your template
It's a pretty flexible way of doing things and one which a lot of frameworks and content management systems use.
Here's a really, really simplified version of a common method.
layout.php
<html>
<body>
<?php echo $content; ?>
</body>
</html>
Then
whatever_page.php
<?php
$content = "Hello World";
include( 'layout.php' );
Sounds like you need to use include() or require()
<?php
include("header.inc.php");
output html code for page
include("footer.inc.php");
?>
The header and footer files can hold all the common HTML for the site.
You asked for how other languages handle this, and I didn't see anything other than PHP, so I encourage you to check out Rails. Rails convention is elegant, and reflects #codeincarnate 's version in PHP.
In the MVC framework, the current view is rendered inside of a controller-specific layout file that encapsulates the current method's corresponding view. It uses a "yield" method to identify a section where view content should be inserted. A common layout file looks like this:
<html>
<head>
<% #stylesheet and js includes %>
<body>
<div id="header">Header content, menus, etc…</div>
<%= yield %>
<div id="footer">Footer content</div>
</body>
</html>
This enables the application to have a different look and feel or different navigation based on the controller. In practice, I haven't used different layout files for each controller, but instead rely on the default layout, which is named "application".
However, let's say you had a company website, with separate controllers for "information", "blog", and "admin". You could then change the navigation for each in a clean and unobtrusive manner by handling the different layout views in their respective layout files that correspond to their controllers.
You can always set a custom layout in the controller method by stating:
render :layout => 'custom_layout'
There are also great helper methods built into Rails so you don't have to rely on $global variables in PHP to ensure your CSS and Javascript paths are correct depending on your development environment (dev, staging, prod…). The most common are:
#looks in public/stylesheets and assumes it's a css file
stylesheet_link_tag "filename_without_extension"
#looks in public/javascripts and assumes it's a js file
javascript_include_tag "jquery"
Of course, each of these sections could be expounded upon in much greater detail and this is just brushing the surface. Check out the following for more detail:
http://guides.rubyonrails.org/layouts_and_rendering.html
What you suggested works OK. As long as print_top_html and print_bottom_html stay in sync (and you can use automated tests to check this), then you never need to worry about them again, leaving you to focus on the real content of the site -- the stuff in the middle.
Alternatively, you can combine print_top_html and print_bottom_html into a single call, and send it HTML code (or a callback) to place in the middle.
I use the partials system of Zend_View (very similar to Rails). A partial is essentially a small HTML template that has its own variable scope. It can be called from inside views like:
<?php echo $this->partial('my_partial.phtml', array( 'var1' => $myvar ));
The variables that get passed into the construct get bound to local variables inside the partial itself. Very handy for re-use.
You can also render a partial from inside normal code, if you're writing a helper object where you have more complex logic than you'd normally feel comfortable putting in a view.
public function helperFunction()
{
// complex logic here
$html = $this->getView()->partial('my_partial.phtml', array('var1' => $myvar ));
return $html;
}
Then in your view
<?php echo $this->myHelper()->helperFunction(); ?>