PHP: Weird function behavior depends on how I include the script it's contained in? I'm at a loss

Question

I honestly have little clue what facts are relevant to this issue, so I'll go from the beginning.

I have bits of text that I'd like to format Markdown-style urls in before plugging into a template, but I don't want them to be wrapped in <p>'s or any other block-level tags.

To do this, I worked my way through the PHP Markdown code, isolated the stuff needed for the url formatting and threw the functions together into a file called 'MarkdownUrl.php' in a folder called 'functions' at my include path.

I'll post that code at the bottom.

For now, let's test it:

<?php
require_once 'functions/MarkdownUrl.php';
$text = "Hello, world. [This is a link.](http://www.google.com)";
echo MarkdownUrl($text);
?>

Output:

Hello, world. This is a link.

So far so good.

I also have a file called 'loader.php' at the include path. It contains a class autoloading function which it registers with spl_autoload_register() and another function called loadFunc() which will require_once a php file from the 'functions' directory according to the name it's passed. The latter is what matters here.

<?php
function loadFunc($funcName)
{
 if (!function_exists($funcName)) {
 require_once 'functions/' . $funcName . '.php';
 }
}
function loadClass($className)
{
 if (!class_exists($className, false)) {
 $classPath = str_replace('_', '/', $className);
 require_once 'classes/' . $classPath . '.php';
 }
}
spl_autoload_register('loadClass');
?>

Ignoring the 'whys' and 'wherefores' of including code in this manner for the moment, the following should be effectively equivalent to the test code I posted earlier, right?

<?php
require_once 'loader.php';
loadFunc('MarkdownUrl');
$text = "Hello, world. [This is a link.](http://www.google.com)";
echo MarkdownUrl($text);
?>

And, yet... output:

Hello, world. [This is a link.](http://www.google.com)

No errors are issued on running the above.

And, I have confirmed that the function MarkdownUrl() is being called (at least, a test echo I inserted at the top of the function fired).

Yet, the output is unformatted. WTF?

A couple I'm-not-sure-if-these-matter-at-all details: shared web host, include path specified via php.ini file in directory the test script runs in.

---

Contents of MarkdownUrl.php:

<?php
$nested_brackets_depth = 6;
$nested_url_parenthesis_depth = 4;
$no_entities = false;
function doAnchors_inline_callback($matches) {
 $whole_match = $matches[1];
 $link_text = $matches[2];
 $url = $matches[3] == '' ? $matches[4] : $matches[3];
 $title =& $matches[7];
 $url = encodeAttribute($url);
 $result = "<a href=\"$url\"";
 if (isset($title)) {
 $title = encodeAttribute($title);
 $result .= " title=\"$title\"";
 }
 $result .= ">$link_text</a>";
 return $result;
}
function encodeAttribute($text) {
#
# Encode text for a double-quoted HTML attribute. This function
# is *not* suitable for attributes enclosed in single quotes.
#
 $text = encodeAmpsAndAngles($text);
 $text = str_replace('"', '&quot;', $text);
 return $text;
}
function encodeAmpsAndAngles($text) {
 global $no_entities;
#
# Smart processing for ampersands and angle brackets that need to 
# be encoded. Valid character entities are left alone unless the
# no-entities mode is set.
#
 if ($no_entities) {
 $text = str_replace('&', '&amp;', $text);
 } else {
 # Ampersand-encoding based entirely on Nat Irons's Amputator
 # MT plugin: <http://bumppo.net/projects/amputator/>
 $text = preg_replace('/&(?!#?[xX]?(?:[0-9a-fA-F]+|\w+);)/', 
 '&amp;', $text);;
 }
 # Encode remaining <'s
 $text = str_replace('<', '&lt;', $text);
 return $text;
}
function MarkdownUrl($text) {
 global $nested_brackets_depth, $nested_url_parenthesis_depth;
 $nested_brackets_re = 
 str_repeat('(?>[^\[\]]+|\[', $nested_brackets_depth).
 str_repeat('\])*', $nested_brackets_depth);
 $nested_url_parenthesis_re = 
 str_repeat('(?>[^()\s]+|\(', $nested_url_parenthesis_depth).
 str_repeat('(?>\)))*', $nested_url_parenthesis_depth);
 $text = preg_replace_callback('{
 ( # wrap whole match in 1ドル
 \[
 ('.$nested_brackets_re.') # link text = 2ドル
 \]
 \( # literal paren
 [ \n]*
 (?:
 <(.+?)> # href = 3ドル
 |
 ('.$nested_url_parenthesis_re.') # href = 4ドル
 )
 [ \n]*
 ( # 5ドル
 ([\'"]) # quote char = 6ドル
 (.*?) # Title = 7ドル
 6円 # matching quote
 [ \n]* # ignore any spaces/tabs between closing quote and )
 )? # title is optional
 \)
 )
 }xs','doAnchors_inline_callback', $text);
 return $text;
}
?>

Answer 1

That happens because you have different variable scopes in your cases.

In first case your variables are defined in global scope, thus $nested_brackets_depth = 6; and other vars defined in MarkdownUrl.php are visible for markdown functions

In other case - that variables are defined in function that includes the file scope, so they are not visible inside the functions.

The solution: change markdown file header to

global $nested_brackets_depth = 6;
global $nested_url_parenthesis_depth = 4;
global $no_entities = false;

Btw, it is a great sample why global is evil and why defining/accessing the variables in global scope is evil too.

• mutus: Of course! Well, with a minor change (inserting global $nested_brackets_depth, $nested_url_parenthesis_depth, $no_entities; rather than attempting to initialize values in the global declarations).
• mutus: Adding that this would probably be avoidable if I'd kept it encapsulated in a class rather than lazily hacking it together into procedural code.