README

Specify annotations in docblocs as JSON.

Requires PHP 5.4 or greater. Install using composer:

composer require shabbyrobe/nope

This example demonstrates almost all of what Nope has to offer:

<?php
require 'vendor/autoload.php';

/**
 * Hellos your hellos
 *
 * :c1 = {
 *     "foo": true,
 *     "bar": ["baz", "qux"],
 *     "ding": {"dong": "woohoo"}
 * };
 * :c2 = {"foo": false};
 *
 * @author Mr. Pants
 */
class Hello
{
    /** :p1 = {"pants": true}; */
    public $pants;

    /** :m1 = {"foo": true}; */
    function world($foo, $bar)
    {
        return "Hello, World!";
    }
}

$parser = new Nope\Parser();
$t = microtime(true);
$out = $parser->parseClass('Hello');

var_dump($out);
var_dump(microtime(true) - $t);

Result:

stdClass#1 (
    [notes] => array (
        'c1' => array (
            'foo' => true
            'bar' => array ( '0' => 'baz' '1' => 'qux' )
            'ding' => array ( 'dong' => 'woohoo' )
        )
        'c2' => array ( 'foo' => false )
    )
    [methods] => array (
        'world' => array(
            'm1' => array ( 'foo' => true )
        )
    )
    [properties] => array (
        'pants' => array (
            'p1' => array ( 'pants' => true )
        )
    )
)
float(0.00057697296142578)

It's plenty quick, but you will still probably want to cache that result on a production server.

Nope annotations follow a simple format:

/**
 * :namespace = JSON;
 */

The parsing rules are straightforward:

Docblock margins are stripped away.
The parser looks for any line that starts with a : character and begins parsing.
Everything from the colon to the = sign (excluding trailing whitespace) is taken as the namespace.
Anything after the = sign (excluding leading whitespace) must be valid JSON. Parsing ends when a ; is encountered as the last character on a line.
Multiple annotations are parsed, but if you specify the same namespace twice, the second definition will overwrite the first.

Everything else inside a docblock is ignored, so this should not interfere with your documentation.

If you need a line in your docblock to start with :, escape it with a backslash: \:.

Due to the fortunate property that JSON strings cannot span multiple lines, it is not possible for the string ;\n to appear anywhere inside valid JSON. This is the property we exploit to make Nope possible. So the following example cannot be parsed successfully by Nope:

<?php
/** :foo = true; :bar = true; */
function impossible() {}

Please, please, please be careful when adding new namespaces to your libraries and applications. Ideally, you should define one namespace for your entire application and embed your annotations as an object inside that. This solves the "one annotation per line" problem as well:

<?php
/** :myapp = {"foo": "bar", "baz": "qux"}; */
function good() {}

/**
 * :foo = "bar";
 * :baz = "qux";
 */
function please_dont() {}

API

Parse all annotations from a class, trait, or interface:

<?php
/** :foo = {"yep": true}; */
class Pants
{
    /** :foo = {"yep": true}; */
    public $property;

    /** :foo = {"yep": true}; */
    public function test() {}
}

$result = $parser->parseClass('Pants');
$result = $parser->parseClass(new \ReflectionClass('Pants'));

This results in:

stdClass#1 (
    [notes] => array (
        'foo' => array (
            'yep' => true
        )
    )
    [properties] => array (
        'property' => array (
            'foo' => array (
                'yep' => true
            )
        )
    )
    [methods] => array (
        'test' => array (
            'foo' => array (
                'yep' => true
            )
        )
    )
)

Parse all annotations from a doc comment:

<?php
/** :foo = {"bar": true}; */
function func()
{}

$function = new ReflectionFunction('func');
$notes = $parser->parseDocComment($function->getDocComment());
$parsesTo = array(
    'foo'=>['bar'=>true],
);

Parse all annotations from a string:

<?php
$string = ':foo = {"bar": true};';
$notes = $parser->parse($string);
$parsesTo = array(
    'foo'=>["bar"=>true],
);

Parse all annotations from an array of Reflectors (must support the name property and the getDocComment() method):

<?php
$rc = new ReflectionClass('Pants');
$notes = $parser->parseReflectors($rc->getMethods(ReflectionMethod::IS_STATIC));

Method and property filters can be passed to parseClass:

<?php
$rc = new ReflectionClass('Pants');
$notes = $parser->parseClass(
    \Pants::class,
    \ReflectionProperty::IS_PUBLIC,
    \ReflectionMethod::IS_STATIC
);

Isn't this a solved problem?

Nope!

I've had about half a dozen goes at this one over the years, and I'm not satisfied with the available solutions. I like attribute-based metaprogramming and think it should be supported natively, but it doesn't look like that's coming to PHP any time soon.

There are already indeed heaps of tools for this already, several of which I have unleashed on the world myself (I'm sorry).

A common approach is to define a complex new language. These languages are often slightly different from vanilla PHP, which imposes a cognitive load each time you have to switch in and out of using them. You also tend to write annotations far less frequently than you write other code, so there is much time spent looking at manuals to fill in the blanks.

They also require complex PHP-based implementations of slow parsers to even be read in the first place. I have remained uncomfortable with these kinds of solutions for a long time - they are far too slow and have way too many moving parts.

I've even had two failed attempts at a leaner alternative to this in my Data Mapper project Amiss (see v3 and v4), both of which fell down because they were too unfamiliar and/or inflexible.

I've remained convinced that there was a native C-based solution to this lurking in PHP's standard library for a good long while, and I'm stunned that it took me this long to realise json_decode has been staring me in the face the whole time.

It's a perfect fit for the job: it can represent complex data structures that map well to pure PHP, the language is ubiquitous and widely understood, and there is a fast C-based parser available to PHP in a single function call.

Nope takes advantage of these properties by finding a way to unambiguously embed JSON into the unstructured text strings you find in doc comments.

shabbyrobe / nope

Maintainers

Details

README

API

Isn't this a solved problem?