BigSmoke » PHP

MediaWiki thumb.php and rewrite rules

Rowan Rodrik — Sun, 07 Mar 2010 18:40:26 +0000

May, last year, I created an empty draft for this post, because, around that time, I had gone through quite some effort before I got thumbnails for foreign file repos working just right. Now, I’m taking a dive into my MediaWiki working dirs in preparation of the creation of a separate development environment, so it’s a good moment to rehash the past experience (almost as good as when I’d have done it right away).

This is how I configured the foreign file repo to be able to use images uploaded to the English wiki from the Dutch wiki:

$wgHashedUploadDirectory = false;
 
$wgForeignFileRepos[] = array(
    'class' => 'ForeignDBRepo',
    'name' => 'en',
    'url' => "http://wiki.hardwood-investments.net/media",
    'hashLevels' => 0,
    //'thumbScriptUrl' => "http://wiki.hardwood-investments.net/thumb.php",
    'transformVia404' => true,//!$wgGenerateThumbnailOnParse,
    'dbType' => $wgDBtype,
    'dbServer' => $wgDBserver,
    'dbUser' => $wgDBuser,
    'dbPassword' => $wgDBpassword,
    'dbName' => 'hardwood',
    'tablePrefix' => 'mw_',
    'hasSharedCache' => false,
    'descBaseUrl' => 'http://wiki.hardwood-investments.net/Image:',
    'fetchDescription' => false
);

To make thumbnails be generated by thumb.php on request I added the following to my .htaccess at the other end (and visa versa, because the Dutch wiki actually contains most of the images):

RewriteCond %{REQUEST_FILENAME} !-f
RewriteCond %{REQUEST_FILENAME} !-d
RewriteRule ^media/thumb/([^/]+)/([0-9]+)px-.*$ /thumb.php?f=$1&width=$2 [L,QSA]

www.stichting-ecosafe.org

Rowan Rodrik — Wed, 16 Sep 2009 10:19:56 +0000

Stichting EcoSafe is a Dutch foundation for the safe-keeping of the funds that are necessary for the maintenance of hardwood plantations. In July of 2006, together with Johan Ockels, I created a website for the Foundation. Johan was responsible for the organization of the whole process. This went very smooth and the website ended up being an emblem of simplicity and clarity. That’s why I wanted to blog a bit about it now, even though there are a few things that I’d probably end up doing different if I were to start from scratch. [There's actually a disturbing number of things for which this is true, I'm coming to notice.]

The Welcome page of the EcoSafe website

EcoSafe page for plantations

EcoSafe cost structure

File structure

Like with most websites, I started with creating an SVN repo so that I wouldn’t have to be afraid of ever losing anything.

The file structure was pretty standard:

a css dir for stylesheets;
img for images;
inc for shared PHP and mod_include stuff and for AJAX partials;
jot for to-do’s and other notes;
and js for JavaScript files and libraries.

Possible file structure improvements

If I were to redesign this file structure, I’d collapse css, img and js into one directory called layout, because these are typically things that require the same robots.txt and caching policy. Also, it is meaningless to organize things by file extension. If you want to sort something by file extension, use ls -X (or ls --sort=extension if you’re on GNU).

Server-side includes

The site would be so simple that I felt that any type of CMS or content transformation would be completely unnecessary. Instead, I decided to rely on Apache’s mod_include and just use a few partials for repeating page elements such as the left sidebar containing the logo and the menu.

Also, because I didn’t need to transform the HTML files, I decided I could use good ol’ HTML 4 instead of XHTML 1 (which I’d have to send to the browser with the wrong mime-type anyway).

This is the HTML for contact.nl.shtml:

"-//W3C//DTD HTML 4.01//EN" "http://www.w3.org/TR/html4/loose.dtd">
 
"en">
  
    Contact EcoSafe
 
    "Content-Type" content="text/html; charset=UTF-8" />
 
    "stylesheet" type="text/css" href="/css/style.css">
  
 
  
    
 
    
 
    "content">
      Contact
 
      Your email to EcoSafe kan be sent to the following address:
      "mailto:service@stichting-ecosafe.org">service@stichting-ecosafe.org.
      Or, alternatively, you can fax us at +31 50 - 309 66 58.
 
      About this website
 
      For comments and/or suggestions concerning this website,
      you can direct an email message at:
      "mailto:webmaster@stichting-ecosafe.org">webmaster@stichting-ecosafe.org.

Alternative language selection

I use to include the repeating parts. has several advantages over in that it allows for content-negotiation, execution of dynamic content etc., but here the only place were it holds an advantage is in the inclusion of /inc/alt-lang.phtml. alt-lang.phtml is a messy PHP script that figures out which language variants of a page are available and displays a selection of alternative language versions (variants with a language different from the current).

SSI and caching

Without the XBitHack directive set to full, all content handled by mod_include is sent without a Last-Modified header. However, I don’t want to use XBitHack at all, because I don’t want just any executable file to be handled by mod_include; that just too much of a … hack.

If I were to do something similar now, I’d use some kind of (sed) substitution to pre-process the includes locally so that more of what I end up uploading is simple static content. The dynamic part of the included PHP script, I would simply replace with JavaScript.

Visual design

As you can see in the HTML example, there’s hardly anything layout oriented in the HTML source. This is good, and means that I have to touch only the CSS for most minor and major lay-out modifications. (It is a pipe-dream to think that you only need to change the CSS to make the same HTML page look however you want as long as that HTML is rich enough in meaning, but for a site with pages of such simple structure, it’s a dream that comes pretty close to reality.)

I’m not much of a designer, but I think design is overrated anyway. Actually, I think that most website suffer from too much design.

The EcoSafe logo

To start the design, I got a logo made by Huite Zijlstra. Because the logo was pretty big and didn’t look good scaled down, I decided to put it at the left of the content area instead of at the top. This would still leave enough room for the menu (which actually takes less space horizontally than the logo).

Colors

For the color scheme, I just picked a few colors from the logo. As always, the base of the scheme would be black text on a white background for maximum readability. The print version hardly uses any colors.

@media screen {
body            { background-color: white;  }
*               { color: black;             }
a:link          { color: #585;              }
h1              { color: #880;              }
h2              { color: #888;              }
strong          { color: #a62;              }
#menu li a      { color: #660;              }
}

Underlines

I wanted an underline below the level 1 and 2 headings. Because I didn’t like the effect of text-decoration:underline (too thick for

s, too dark for

s and different from browser to browser) and because border-bottom was set too far from the text, I made two simple PNG images that I could repeat-x along the bottom edge.

@media screen {
h1 { background: url('/img/h1-border-bottom.png') bottom left repeat-x; }
h2 { background: url('/img/hx-border-bottom.png') bottom left repeat-x; }
}

The menu is very simple. The markup is part of inc/left-side.en.html for the English version and inc/left-side.nl.html for the Dutch version:

cat inc/left-side.en.html

"left" lang="en">
  class="logo" href="/index.en">class="logo" alt="[Logo]" src="/img/logo.jpg">
 
  "menu" class="menu">
    "/index.en" rel="start">Start page
    "/plantations.en">For plantations
    "/investors.en">For investors
    "/history.en">History
    
    "/methods.en">How it works
    "/cost-structure.en">Cost structure
    "/cost-calculator.en">Cost calculator
    
    "/contact.en">Contact

The EcoSafe menu (in English)

As is customary, I started by removing all the default list styles and made the anchors behave as block-level elements. I used the big O from the logo for bullets in the list (using background-image instead of list-style-image because the latter gives unpredictable cross-browser results and doesn’t make the bullet clickable).

#menu {
  margin-top: 2em;
  margin-left: 2em;
  list-style-type: none;
  padding: 0;
}
 
#menu li {
  margin: 0;
}
 
#menu li a {
  display: block;
  background-image: url('/img/o-21x16.png');
  background-position: top left;
  background-repeat: no-repeat;
  padding-left: 30px;
  text-decoration: none;
  font-family: sans-serif;
  font-weight: bold;
  color: #660;
}
 
#menu li a:hover,
#menu li.active a {
  background-image: url('/img/oSafe-21x16.png');
}
 
#menu a:hover {
  color: #787800;
}

JavaScript menu item activation

To add the active class to the currently active list item (

), I used a client-side solution using JavaScript. After all, it’s proper use of JavaScript to enhance your user interface with it (as long as, as many would say, it isn’t required for the UI to function (as it is in the Cost Calculator)).

// menu.js
 
var menu = document.getElementById('menu');
var anchors = menu.getElementsByTagName('a');
var locationHref = window.location.pathname.toString();
  
for (i = anchors.length - 1; i >= 0; i--) {
  a = anchors[i];
  aHref = a.href;
    
  // Does this menu item link to the current page?
  // We find out by looking if the window location contains the URL in the anchor
  // or the other way arround. The reason to look at both is content-negotiation.
  // It's also true if the location is just '/' and we're looking at the anchor of
  // the 'start' page.
  if ( (locationHref === '/' && a.rel === 'start') ||
       (locationHref !== '/' && ( locationHref.indexOf(aHref) !== -1 ||
                                  aHref.indexOf(locationHref) !== -1 ) ) ) {
    a.parentNode.className = 'active';
    break;
  }
}

I actually just fixed a long-standing bug that was caused by me not being able to fully rely on HTTP language negotiation for the selection of the appropriate language variant, which made me change all links from being language-neutral to including the language in the link target (e.g.: http:///history became http:///history.en and http:///history.nl), the problem with this being that, instead of being able to link to link to http:/// (http://www.stichting-ecosafe.org/), I had to link to http:///index.en or http:///index.nl, making it more difficult to detect the active anchor if we’re requesting the home page through http:/// instead of on of its language-specific URLs.

The JavaScript rested on the assumption that by reverse iterating through all the anchors in the menu and thus processing the link to http:/// as last, I’d know that I had struck the home page and wouldn’t need to worry that any of the links contain a slash. (I don’t know if I intended it to work this way, but it sure seems to me now that the only way this could ever have worked was as an apparent side-effect of the looping order; the SVN logs seem to agree.)

I could have solved this by redirecting all requests for http:/// to the appropriate variant. Maybe I should have (to avoid duplicate content). Instead I chose to add a rel="start" attribute to the links to the home page, as can be deduced from the JavaScript above. (To resolve the duplicate content issue, I could also add a canonical link to the header of the two language variants.)

Anyway, all this brings me to the messy subject of content negotiation.

Content and language negotiation

The EcoSafe website would be bi-lingual (English and Dutch) from the onset. Initially, I wanted to use language negotiation to the extend of having completely language-neutral URLs. For example: http:///cost-calculator instead of http:///cost-calculator.en and http:///cost-calculator.nl. In the end, you can make this work properly in the browser with the help of a cookie, but it’s still a pipe-dream because nothing else will work if you do not also offer another navigational path to the different variants. Maybe, we’ll revisit this topic for a later experiment.

Content-type negotiation is almost effortless with Apache thanks to mod_negotiation. If, like me, you despise to have .html, .htm, .xhtml, .phtml, .pxhtml. .sxhtml, .php, .xml in your URL (I actually used all of these at some time or other), you only have to make sure that MultiViews is in your options.

I’ve configured SSI by means of the following instead of a “magic mime-type”:

AddType         text/html       .shtml
AddHandler      server-parsed   .shtml
AddCharset      UTF-8           .shtml
AddOutputFilter Includes        .shtml

For PHP I couldn’t do the same because my web host was still at Apache 1.3. Otherwise, the following should have worked equally well for PHP:

# This doesn't work with Apache 1.3
AddType        text/html       .phtml
AddHandler     php-script      .phtml
AddCharset     UTF-8           .phtml

Configuring language priority is easy with Apache:

Integrating PHP and SSI

The integration of PHP with all the weirdness that I had configured and created around SSI took some figuring out. Luckily, PHP offers a virtual() function that works roughly the same as mod_include's . Here’s an example:


   virtual('/inc/left-side.en.html'); ?>
   $uri = '/cost-calculator.en.phtml'; include('inc/alt-lang.phtml'); ?>

In retrospect, it’s pretty much bullshit to use it. I could have just as well require()d the partials (which I actually did for the alternate language selection), but I probably started out using virtual on a more generic URL without language and content-type selection in it.

406 handling

Because I deployed on Apache 1.3 and the ForceLanguagePriority directive was only introduced with Apache 2.0.30, I had to write an ugly hack to avoid visitors getting 406 errors. To that end, I added a 406 handler to my .htaccess file:

LanguagePriority en nl
ForceLanguagePriority Prefer Fallback # This doesn't work with 1.3
 
ErrorDocument 406 /error-406.php # Luckily, this does

error-406.php is a PHP file that figures out the available variants based on $_SERVER['REQUEST_URI']. Then, it simply picks the first one (which works because, accidentally, that’s the one I’ve given priority using the LanguagePriority directive as well), outputs a 200 OK header instead of the 406, and virtual()s the file of the variant. The code looks somewhat like this:


chdir($_SERVER['DOCUMENT_ROOT']);
$filenames = glob(basename($_SERVER['REQUEST_URI']) . ".*");
 
$filename = $filenames[0];
 
apache_setenv('DOCUMENT_URI', "/$filename");
 
header('HTTP/1.1 200 OK');
virtual("$filename");

EcoSafe Cost Calculator

EcoSafe Cost Calculator results

The Cost Calculator

The EcoSafe Cost Calculator is some of the least neatly contained and most procedurally oriented PHP code I’ve ever produced while knowing full well what I was doing. It does almost everything it does in global scope. Yet, it does it well.

The thing is designed as a dynamic web page rather than a web application. What I mean by this is that it’s simply two pages (one for English and one for Dutch) using PHP among a number of pages using SSI. In an application, it’s usual to have just one ‘view’ that is the same for all languages, but here I chose to put the different language versions in different language pages and then include everything reusable (and language-neutral) from within these files.

Most of the actual processing and calculating is done in inc/costs-functions.php. (The part about gotos is a joke. (Labeled blocks would have been quite sufficient. ))

 # costs-functions.php - Stuff that's includes by cost-calculator.{nl,en}.phtml
/**
 * Just remember that this code was never meant to be general purpose or anything.
 * So, relaxeeee and keep your OO-axe burried where it belongs.
 * Oh, if only PHP would support GOTO's ... Sigh ...
 */

The rest of the file is just a whole lot of processing of form data and turning it into something that can be easily traversed for display to the user. There are even the function calls without arguments doing all their work on globals. These are actually only added to make it clearer em a piece of code is doing. And—I must say—after a few years it’s still remarkably clear to me what each part of the code is doing. There’s no deep, confusing nesting structures or anything. There’s just a whole lot of very simple code.

Some simple AHAH increases form interactivity

Users of the calculator can add any number of plantings and locations. When the user decides to add a planting or a location, the onClick event triggers the execution of addExtraPlanting() or addExtraLocation(). Here’s how addExtraPlanting() looks:

function addExtraPlanting() {
  lang = document.documentElement.lang;
 
  new Ajax.Updater(
    'plantings', '/inc/planting.' + lang, {
      method: 'get',
      insertion: Insertion.Bottom
    }
  );
}

Ajax.Updater comes from the Prototype JavaScript framework.

Here’s what inc/planting.en.phtml looks like. The same file is also included in a loop to rebuild the form’s state after submitting.


  "num_hectares[]" type="text" size="5" value="" />
 
  hectares have been planted in
 
  
 
  ("Remove this planting" href="#" onclick="removePlanting(this); return false;">x)

I think that I’ve gone into small enough detail by now to get to the conclusion. Also showing the contents of planting_options.php would be pushing it. Ah, well…


 
if ( !isset($this_year) ) $this_year = intval(date('Y'));
if ( !isset($plant_year) ) $plant_year = $this_year;
 
for ($i = $this_year; $i >= $this_year - 20; $i--)
  echo " . ($i == $plant_year ? " selected='1'" : "") . ">$i\n";

(Yesterday, I couldn’t resist the temptation of turning this into a simple file to require() instead of the function definition it was. I think it’s funny to refactor something to remove encapsulation.)

Conclusion

As is usual when looking at old code, I see many things that I’d do (even just a little) different today, but I saw a surprising number of solutions that I actually still like now that I see them back after three years. Removing some of the remaining warts probably won’t do much good besides the masturbatory satisfaction it could give me. (It’s likely that the website won’t live much longer, making such extra attention very undeserved.) But, nevertheless, I’ve enjoyed blogging about it now to recoup the whole experience and to at least look at what I’d do different now and what I learned in the meantime.

Some links

PHP fgetcsv() behavior on empty lines

Rowan Rodrik — Sat, 05 Sep 2009 08:55:13 +0000

The PHP documentation for fgetcsv() states that A blank line in a CSV file will be returned as an array comprising a single null field, and will not be treated as an error. Here’s a quick demonstration of this behavior.

fgetcsv.php:


 
while ($fields = fgetcsv(STDIN, 0, ';'))
  print_r($fields);
 
exit(0);

Execute the script and feed it some CSV with empty lines:

php -q fgetcsv.php
"Veld 1";"Veld 2";"Veld 3";;"Veld 5"
 
"Field 1";;"Field 3";"Field 4";
;;;;
;"Campo 2";;;"Campo 5"

After pressing Ctrl+D, I’m presented with the following output:

Array
(
    [0] => Veld 1
    [1] => Veld 2
    [2] => Veld 3
    [3] => 
    [4] => Veld 5
)
Array
(
    [0] => 
)
Array
(
    [0] => Field 1
    [1] => 
    [2] => Field 3
    [3] => Field 4
    [4] => 
)
Array
(
    [0] => 
    [1] => 
    [2] => 
    [3] => 
    [4] => 
)
Array
(
    [0] => 
    [1] => Campo 2
    [2] => 
    [3] => 
    [4] => Campo 5
)
Array
(
    [0] => 
)

This behaviour on empty lines is a little bit annoying if you want to test if the line is empty():

$a = array(null);
print_r($a);
 
if ( empty($a) )
  echo '$a is empty';
else
  echo '$a is not empty';
 
echo "\n";

This code will print:

Array
(
    [0] => 
)
$a is not empty

Hence, the following function:

/**
 * This function tests if the given array (as returned by fgetcsv())
 * is the result of an empty line in the CSV file.
 *
 * It does not work for lines that contain only delimiters.
 * From the POV of this function, these are simply records with
 * many empty fields.
 */
function fgetcsv_empty_line($row_array) {
  return ( !isset($row_array[1]) and empty($row_array[0]) );
}

Now, if I change the call to empty() in my test to a call to fgetcsv_empty_line():

$a is empty

Enforcing Drupal URL aliases

Rowan Rodrik — Wed, 10 Jun 2009 12:09:07 +0000

I hate modules, especially core modules. I prefer code to be tightly integrated. I want it to work together. Is that too much to ask? In Drupal, most functionality has been stuffed in modules. There’s a Locale module, a Content Translation module and a Path module. What’s missing is a Working Together module.

For me, clean, meaningful URLs are a number one, two and three requirement for any website that I do. Drupal considers /node/54673 to be a cool URL. I don’t. So, as a kind of afterthought, Drupal comes with the Path module. This module allows you to set URL aliases per node.

The problem is that there’s no concept of a canonical URL. The URL alias works, but so does the node/3242 URL. Neither redirects to the other. In many cases this is not much of a problem (because regular visitors will not notice this) but for our current project it is.

We have a lot of blocks with URL dependent visibility settings. For example, for a section about investing we have a menu that is displayed on all URLs starting with /investing, such as /investing/projects and so on.

After editing a page, Drupal helpfully redirects the user to node/[nodenumber]. For us, this means that the menu is no longer displayed and even the theme will be wrong. (We use the Sections module to select a subtheme based on which section you’re in.)

Global Redirect doesn’t work

The Global Redirect module promises to solve this by allowing you to redirect node/[nodenumber] URLs to their alias if available. It kinda does, in some circumstances.

Our Drupal website sports two languages: English (EN) and Dutch (NL). English is the default language (not the fallback language; we don’t use a fallback) and doesn’t use a prefix. Dutch uses the nl prefix. Two example URLs:

URL alias	Generic URL
`http://www.example.com/investing/projects`	`http://www.example.com/node/288`
`http://www.example.com/nl/beleggen/projecten`	`http://www.example.com/nl/node/110`

When /node/288 is requested, the client is correctly redirected to /investing/projects, but when /node/110 is requested, no redirect takes place. It will take place when prefixing /nl, but this is completely useless since Drupal’s built-in actions such as edit don’t redirect using this prefix, and these actions were what we needed this module for in the first place.

A very simple hack that does work

We ended up tearing our hair out trying to fix Global Redirect until we decided that we could just delete the module and replace it with a RewriteRule and a simple PHP script.

Modify: `.htaccess`

# Put this after RewriteBase and before Drupal's default rewrite rules
RewriteRule ^(../)?node/([0-9]+)$ fixurl.php?nid=$2 [L]

Add: `fixurl.php`


 
require_once './includes/bootstrap.inc';
drupal_bootstrap(DRUPAL_BOOTSTRAP_DATABASE);
 
$result = db_query("SELECT * FROM {url_alias} WHERE src = 'node/%d' LIMIT 1", $_GET['nid']);
if ( db_error() ) die("O agony!");
 
$url_alias_object = db_fetch_object($result);
$destination = $url_alias_object->dst;
 
$result = db_query("SELECT prefix FROM {languages} WHERE language = '%s'", $url_alias_object->language);
if ( db_error() ) die("O agony!");
 
$prefix = db_result($result);
 
if ( !empty($prefix) )
  $prefix .= '/';
 
header("Location: /$prefix$destination",TRUE,301);
 
?>

Shortcomings in our hack

The code assumes that every content page has an URL alias. For us, this is okay, because we need these pretty URLs to even have menus show up or to have the right page be displayed with the right theme.

Also, this code is specifically tailored to language code in the URL prefix. For subdomain based language selection, for example, you’d need to modify it.

WordPress feed pagination

Rowan Rodrik — Mon, 16 Feb 2009 21:52:37 +0000

Wiebe uses his author Atom feed to generate a list of his blog posts here on his own website. WordPress feeds only display the latest 10 entries. He has written 16 so far. What he needs is a feed with all his entries.

First, I tried if pagination works for feeds. Appending “/page/2” (as is used an non-feed lists) to a feed URL gives a 404 so I was kind of stuck there. Four days ago, after Googling for some time, unable to find a solution anywhere, I asked on the forum. Still no answer today so I tried to find out which parameters WordPress accepts in the QUERY_STRING. The WordPress Codex does explain how queries are handled but not which parameters are accepted.

Digging into wp-includes/query.php, with much trial and error, I found out that I can append ?paged=2 to the URL to get the next page. At least I got that sorted then. There are a number of much more promising parameters supported by get_posts(), but these don’t seem to be parsed by parse_query(). Next time, I’d like to find out how how to use two of these: nopaging and posts_per_page.

Wiebe could complete his list by merging together all the pages of the feed, but I’d much prefer to find a relatively painless method to produce a feed with an unlimited number of posts.

Notes

http://codex.wordpress.org/Query_Overview
http://codex.wordpress.org/Function_Reference/WP_Query
http://codex.wordpress.org/Template_Tags/query_posts

New theme

Rowan Rodrik — Sun, 13 Jul 2008 15:26:45 +0000

After upgrading to WordPress 2.5.x, I had to fall back on a stock theme because my old customization of the Sandbox theme no longer worked with the upgrade. But, then, it was time to redo my theme anyway. So here you’re looking at the first version of my new theme. I might have let it stabilize some more before putting it on-line, but who cares? My reader maybe? Let’s just hope he or she doesn’t use IE.

Vertical navigation

Ever since the first time that I saw a blog which featured vertical navigation instead of the typical columns, I’ve wanted to implement this for myself. Well, finally…

Site-wide elements use the complete width of the page. The page content is centered in the middle at 87.5%. The identity stuff in the header and the navigation in the footer sits against a back blackground while the content area has the proven black on white for easy reading. I hope that the strong color-contrast as well as the clear difference in with between site-wide elements and page content makes it easy to keep focused on either reading or navigating without distractions.

… and a talkative footer

With this theme, I didn’t want another footer which consist of the odd logo and some loose copyright statements. I wanted a footer which you can actually read, even understand. And who cares if it takes up a little space? It’s at the bottom of the page.

I’ve written an (unpublished, unpolished) plug-in which can generate a list of posts that are chronologically related. Traditionally, most blogs have a next/previous post link at the top and bottom of each post. This works very well if you limit your blog to one subject (which is really a very good idea anyway), but if, like mine, your blog is a little bit messy, you could say that someone who stumbled here searching for an article about Subversion is not necessarily interested in the next post if this is a photo of my baby niece.

Hence the chronologically related posts plugin. With this plugin I can say wether I want a link to the first, previous and next post in the blog, within the same category, or matching a given number of tags. (The tag matching isn’t implemented yet, though. Also, matching on meta fields would be a kick-ass ass way to support explicit sequences.)

I put the list generated by this plug-in on top of a blue background besides the various context links of the post.

Issues left

I hope to have the first major revision of my theme ready soon. Here’s a list of some issues that I might address:

The CSS renders a bit psychedelically in MSIE 6 (only version I tested) at the moment. Sigh… Let’s just hope that IE 7 will give better results. Then I’ll gladly drop the IE 6 support.
When viewing a category, the tag cloud in the navigation panel at the bottom only shows tags for that category. This has to do with the use with me calling the st_tag_cloud() from within the category template.
Some of the elements that I just showed to you don’t really look that good and most elements that I didn’t can be said to be … hideously ugly. Some highlights: the header (should really be a few cool images), the comment form, and the Next/Previous Page links.

Comment!

I’d almost forget all about the clean, new look of the comment list. And, if you register a Gravatar, your comments will be accompanied by your avatar. Try it. Please!

Separate development/production environments for WordPress

Rowan Rodrik — Sat, 12 Jul 2008 00:10:32 +0000

When you’re out Googling on how to maintain a separate development environment for a WordPress installation, you will only stumble across information about how to install all kinds of WAMPP packages. Well, I don’t care about WAMP (or WAMPP). I want to be able to edit my theme, change my plugins, mess with my database locally and then deploy my changes when they’re ready and well-tested (as if I ever…)

Rails was the obvious inspiration for how to do this properly. In Rails, the whole development and deployment process is very intuitive and powerful. In WordPress documentation I never even see the awareness of the need for this separation. They usually tell you to download stuff, upload it and muck about with it on the life production server. But, I’m not the mucking-about-in-live-configurations type. I’m the I-fucked-this-up-so-often-I-want-a-staging-area type. This post is about how I managed to fulfill this wish with WordPress.

Changing the environment

The first thing I had to do was to find some way to decide which environment to go into. For some reason I decided to use Apache’s mod_rewrite to set an environment variable based on the HTTP Host header. This is in fact very illogical, but we’ll get to that later.


RewriteEngine On
RewriteBase /
RewriteCond %{HTTP_HOST} =bsblog [OR]
RewriteCond %{HTTP_HOST} =bsblog.molensteen
RewriteRule . - [env=WP_ENV:development]
RewriteCond %{HTTP_HOST} =blog.bigsmoke.us
RewriteRule . - [env=WP_ENV:production]

 
# Keep out of WP's own block of rewrite rules below
# BEGIN WordPress

. - looks like a needle because it’s voodoo. The dot says I match anything and the dash says I change nothing of what I match. I do set an environment variable to whether I want to be in development or in production.

So I now have an Apache environment variable available for querying from within PHP (as if PHP doesn’t have a $_SERVER['HTTP_HOST'] variable:-? ) and I can make use of that in my wp-config.php.

Multiple configurations in one file

I love configuration files that share the program’s language; wp-config.php being simple PHP code is what make this whole thing so easy:


 
if ( getenv('WP_ENV') == 'production' ) {
  // ** MySQL settings ** //
  define('DB_NAME', 'blog');             // The name of the database
  define('DB_USER', 'wordpress');        // Your MySQL username
  define('DB_PASSWORD', '[my password]'); // ...and password
  define('DB_HOST', 'bigsmoke.db');        // 99% chance you won't need to change this value
  define('WP_SITEURL', 'http://blog.bigsmoke.us');
}
elseif ( getenv('WP_ENV') == 'development' ) {
  // ** MySQL settings ** //
  define('DB_NAME', 'bsblog');          // The name of the database
  define('DB_USER', 'root');            // Your MySQL username
  define('DB_PASSWORD', '[my password]'); // ...and password
  define('DB_HOST', '127.0.0.1');       // 99% chance you won't need to change this value
  define('WP_SITEURL', 'http://bsblog');
  //define('WP_DEBUG', true);
}
 
define('WP_HOME', WP_SITEURL);
 
// You can have multiple installations in one database if you give each a unique prefix
$table_prefix  = 'wp_';   // Only numbers, letters, and underscores please!
 
// The rest of the stuff in this config file just isn't interesting
?>

There’s a few things to note here. You have to use getenv() or $_SERVER instead of $_ENV because variables set by Apache end up in the former two. Another thing to note is that I should have just checked $_SERVER['HTTP_HOST'] instead of resorting to mod_rewrite voodoo. For the rest it’s all very straight-forward: I make some database settings depend on which environment I’m in and I set the URL accordingly.

Development URLs

I had some trouble putting the pieces back together when newer WordPress versions started doing automatic redirects for URL that didn’t match siteurl in the wp_options. This change meant that when going to http://bsblog/ (the development URL for this weblog) for example, I’d inevitably end up at http://blog.bigsmoke.us/.

Links had always been constructed according to this setting, so I had already been planning a plug-in to transform production URLs to development URLs. But, I learned (a little late) that this is completely unnecessary since wp-config.php supports the configuration of a base URL. I had wrongly assumed that settings that weren’t in the sample config file, simply didn’t exist.

Thus, after adding WP_SITEURL and WP_SITEHOME to wp-config.php, everything was working.

Ideas to further enhance your configuration

Don’t limit yourself to one development environment if you have more than one development server.
Automate your deployment process. I use rsync for this.
Write a script to clone your production database to your development database. There’s no substitute for actual data.

Apache’s ForceType directive overrides AddCharset directives

Rowan Rodrik — Mon, 26 Feb 2007 22:30:01 +0000

Yesterday, after uploading a refreshed www.sicirec.org, some character encoding issues popped up because I had converted the website’s content from ISO-8859-1 (Latin 1) to UTF-8. (I wanted to be able to type and paste special characters from PuTTY into VIM without worrying about the particular encoding of each file.)

The Apache HTTPD at InitFour, our webhosting provider, is configured to send ISO-8859-1 by default, while the one on our test server is configured for UTF-8. This caused a little bit of a surprise when I uploaded the refreshed website and saw all characters outside the ASCII range mangled on the life website!

I quickly dug into my .htaccess file to add the AddCharset utf-8 .xhtml directive. To my surprise, this didn’t do squat. A lot of fiddling, reloading and researching later, I realized that the following section in my .htaccess file rendered the AddCharset directive irrelevant:


ForceType text/html

I had to change the ForceType directive to include the charset as a MIME parameter:


ForceType 'text/html; charset=UTF-8'

Now, it all seemed to work. (Except that it didn’t really because I do some ridiculously complex content negotiation stuff involving a 406 handler in PHP that virtuals the most appropriate variant when no match is found. This script didn’t send a useful Content-Type header. After first adding it to the script, I noticed that the AddDefaultCharset is actually allowed in .htaccess context—a discovery which luckily rendered the other hacks useless.)