.PHP File Format – Lexical structure

Tokens

The scanner produces named tokens such as T_VARIABLE, T_STRING, T_LNUMBER and T_CONSTANT_ENCAPSED_STRING. Parse errors name them, as in unexpected token "==". PHP code can read them with the Tokenizer extension (token_get_all(), and the PhpToken class added in PHP 8.0). The full list is in the manual’s List of parser tokens.

Since PHP 8.0 a namespaced name is one token: T_NAME_QUALIFIED for App\Model, T_NAME_FULLY_QUALIFIED for \App\Model and T_NAME_RELATIVE for namespace\Model. Spaces are no longer allowed around the separators in a name, and keywords may be parts of a namespaced name.

White space is the space, tab and line break. It is required only where two tokens would otherwise run together, such as echo $a.

Comments

• Comments do not nest. A /* inside a /* */ comment has no meaning, and the first */ ends it.
• Since PHP 8.0, #[ starts an attribute, so a # comment must not begin with [.
• A documentation comment placed directly before a function, class, method or property is kept by the engine and can be read with Reflection’s getDocComment(). The language gives its contents no meaning. PHPDoc tags such as @param are a convention read by tools.

Attribute syntax

An attribute group starts with the token #[ (T_ATTRIBUTE) and ends with ]. It holds one or more attributes separated by commas, with an optional trailing comma, and comes before the declaration it applies to.

attribute-group:  #[ attribute {, attribute} [,] ]
attribute:        name [ ( arguments ) ]

#[Route('/home', methods: ['GET'])]
#[Cache(60), Authorised]
function home(): string { return 'hello'; }

What attributes do, which declarations take them and how to read them is on the Attributes page.

Identifiers

Names of variables, functions, classes, constants, labels and so on follow one pattern, given in the manual as a regular expression:

^[a-zA-Z_\x80-\xff][a-zA-Z0-9_\x80-\xff]*$

• A name starts with a letter, an underscore or a byte from 128 to 255, then continues with those or digits. There is no length limit.
• Names are compared as bytes. UTF-8 names work because every byte of a multi-byte UTF-8 character is in the allowed range. Two names that look the same but are encoded differently, such as a precomposed ä and a plus a combining mark, are different names.
• A variable name always starts with $; the $ is not part of the name. Other kinds of name have no sigil.
• A variable name that does not fit the pattern can still exist if it is reached dynamically: ${'invalid-name'} = 1; See Variable variables.

Case rules

Part of the language ignores case and part of it does not.

Case-insensitive names are compared using ASCII rules. Autoloaders usually map class names to files, and file systems are often case-sensitive, so in practice a class should be used with the same case as its declaration.

Keywords

These words are reserved by the grammar. They cannot be used as the names of constants, functions or classes (manual: List of Keywords). Some look like functions but are language constructs: they need no parentheses in some cases, and cannot be called dynamically by name.

In alphabetical order the full list is: __halt_compiler abstract and array as break callable case catch class clone const continue declare default die do echo else elseif empty enddeclare endfor endforeach endif endswitch endwhile eval exit extends final finally fn for foreach function global goto if implements include include_once instanceof insteadof interface isset list match namespace new or print private protected public readonly require require_once return static switch throw trait try unset use var while xor yield, plus yield from, written as two words.

fn has been reserved since PHP 7.4, match since 8.0 and readonly since 8.1; readonly may still be used as a function name.

Keywords may be used as the names of properties, methods, class constants and enum cases, except that class cannot name a class constant, because Foo::class is the class-name fetch. They may also be named arguments. Since PHP 7.0 this applies inside a class body. Elsewhere a keyword remains reserved.

Literals

true, false and null are predefined constants that are treated specially and work in any case. TRUE and True are the same thing.

decimal:      [1-9][0-9]*(_[0-9]+)* | 0
hexadecimal:  0[xX][0-9a-fA-F]+(_[0-9a-fA-F]+)*
octal:        0[oO]?[0-7]+(_[0-7]+)*
binary:       0[bB][01]+(_[01]+)*

LNUM           [0-9]+(_[0-9]+)*
DNUM           ({LNUM}?"."{LNUM}) | ({LNUM}"."{LNUM}?)
EXPONENT_DNUM  (({LNUM} | {DNUM}) [eE][+-]? {LNUM})

$name = 'Ada';
$text = <<<EOT
    Hello, {$name}.
      The "quotes" need no escape.
    EOT;
// "Hello, Ada.\n  The \"quotes\" need no escape."

Interpolation

Double-quoted strings and heredocs replace variables with their values. The manual describes two forms, plus a third that is deprecated.

$fruit = ['apple', 'name' => 'pear'];
echo "I ate $fruit[0] and $fruit[name].";  // I ate apple and pear.
$p = new stdClass;
$p->x = 3;
echo "x is $p->x";                        // x is 3

echo "{$arr['key']}";          // quoted keys need this form
echo "{$arr[3][2]}";
echo "{$obj->values[3]->name}";
echo "{$square->width}00 cm";   // the braces end the name before the zeros
echo "{${$name}}";             // a variable variable

Punctuation

All operators, with their precedence, are listed on the Expressions page.