Download the PHP package nikic/php-ast without Composer

php-ast

This extension exposes the abstract syntax tree generated by PHP 7 and 8.

This is the documentation for version 1.x.y. See also documentation for version 0.1.x.

Table of Contents

• Installation
• API overview
• Basic usage
• Example
• Metadata
• Flags
• AST node kinds
• AST versioning
• Differences to PHP-Parser

Installation

Windows: Download a prebuilt Windows DLL and move it into the ext/ directory of your PHP installation. Furthermore, add extension=php_ast.dll to your php.ini file. (The php-ast PECL page also links to the same DLLs for Windows.)

Unix (PECL): Run pecl install ast and add extension=ast.so to your php.ini.

Unix (PIE): Run pie install nikic/ast and add extension=ast.so to your php.ini.

Unix (Compile): Compile and install the extension as follows.

Additionally add extension=ast.so to your php.ini file.

API overview

Defines:

• ast\Node class
• ast\Metadata class
• ast\AST_* kind constants
• ast\flags\* flag constants
• ast\parse_file(string $filename, int $version)
• ast\parse_code(string $code, int $version [, string $filename = "string code"])
• ast\get_kind_name(int $kind)
• ast\kind_uses_flags(int $kind)
• ast\get_metadata()
• ast\get_supported_versions(bool $exclude_deprecated = false)

Basic usage

Code can be parsed using either ast\parse_code(), which accepts a code string, or ast\parse_file(), which accepts a file path. Additionally, both functions require a $version argument to ensure forward-compatibility. The current version is 120.

The abstract syntax tree returned by these functions consists of ast\Node objects. ast\Node is declared as follows:

The kind property specifies the type of the node. It is an integral value, which corresponds to one of the ast\AST_* constants, for example ast\AST_STMT_LIST. See the AST node kinds section for an overview of the available node kinds.

The flags property contains node specific flags. It is always defined, but for most nodes it is always zero. See the flags section for a list of flags supported by the different node kinds.

The lineno property specifies the starting line number of the node.

The children property contains an array of child-nodes. These children can be either other ast\Node objects or plain values. There are two general categories of nodes: Normal AST nodes, which have a fixed set of named child nodes, as well as list nodes, which have a variable number of children. The AST node kinds section contains a list of the child names for the different node kinds.

Example

Simple usage example:

The util.php file defines an ast_dump() function, which can be used to create a more compact and human-readable dump of the AST structure:

To additionally show line numbers pass the AST_DUMP_LINENOS option as the second argument to ast_dump().

A more substantial AST dump can be found in the tests.

Metadata

There are a number of functions which provide meta-information for the AST structure:

ast\get_kind_name() returns a string name for an integral node kind.

ast\kind_uses_flags() determines whether the flags of a node kind may ever be non-zero.

ast\get_metadata() returns metadata about all AST node kinds. The return value is a map from AST node kinds to ast\Metadata objects defined as follows.

The kind is the integral node kind, while name is the same name as returned by the get_kind_name() function.

flags is an array of flag constant names, which may be used by the node kind. flagsCombinable specifies whether the flag should be checked using === (not combinable) or using & (combinable). Together these two values provide programmatic access to the information listed in the flags section.

The AST metadata is intended for use in tooling for working the AST, such as AST dumpers.

Flags

This section lists which flags are used by which AST node kinds. The "combinable" flags can be combined using bitwise or and should be checked by using $ast->flags & ast\flags\FOO. The "exclusive" flags are used standalone and should be checked using $ast->flags === ast\flags\BAR.

AST node kinds

This section lists the AST node kinds that are supported and the names of their child nodes.

AST Versioning

The ast\parse_code() and ast\parse_file() functions each accept a required AST $version argument. The idea behind the AST version is that we may need to change the format of the AST to account for new features in newer PHP versions, or to improve it in other ways. Such changes will always be made under a new AST version, while previous formats continue to be supported for some time.

Old AST versions may be deprecated in minor versions and removed in major versions of the AST extension.

A list of currently supported versions is available through ast\get_supported_versions(). The function accepts a boolean argument that determines whether deprecated versions should be excluded.

In the following the changes in the respective AST versions, as well as their current support state, are listed.

120 (current)

Supported since 1.1.3 (TBD).

• clone $expr is now represented like a function call (using AST_CALL instead of AST_CLONE).
• exit($expr) is now represented like a function call (using AST_CALL instead of AST_EXIT).

110 (stable)

Supported since 1.1.2 (2024-08-10).

• Add a hooks child node for AST_PROP_ELEM (PHP 8.4 property hooks) and AST_PARAM (constructor property promotion can have property hooks)
• Add new node kinds AST_PROPERTY_HOOK and AST_PROPERTY_HOOK_SHORT_BODY.
• Remove the name child node from the AST_ARROW_FUNC and AST_CLOSURE nodes (previously "{closure}")

100 (stable)

Supported since 1.1.1 (2023-11-12).

• Add a type child node for all AST_CLASS_CONST_GROUP nodes.

90 (stable)

Supported since 1.0.14 (2021-07-24)

• Same as AST version 85.

85 (stable)

Supported since 1.0.11 (2021-04-20)

• Add a type child node (for enum type) for all AST_CLASS nodes.

80 (stable)

Supported since 1.0.10 (2020-09-12).

• mixed type hints are now reported as an AST_TYPE with type TYPE_MIXED instead of an AST_NAME.
• AST_CLASS_CONST_GROUP nodes are emitted for class constant declarations wrapping the AST_CLASS_CONST_DECL and any attributes. Previously, AST_CLASS_CONST_DECL would be emitted.
• AST_PARAM, AST_CLASS_DECL, AST_METHOD, AST_PROP_DECL, AST_CLOSURE, AST_FUNC_DECL, and AST_ARROW_FUNC nodes now contain an attributes child.

70 (stable)

Supported since 1.0.1 (2019-02-11).

• AST_PROP_GROUP was added to support PHP 7.4's typed properties. The property visibility modifiers are now part of AST_PROP_GROUP instead of AST_PROP_DECL.

  Note that property group type information is only available with AST versions 70+.

• AST_CLASS_NAME is created instead of AST_CLASS_CONST for SomeClass::class.

60 (deprecated)

Supported since 0.1.7 (2018-10-06). Deprecated in php-ast 1.1.0.

• AST_FUNC_DECL and AST_METHOD no longer generate a uses child. Previously this child was always null.
• AST_CONST_ELEM now always has a docComment child. Previously it was absent on PHP 7.0. On PHP 7.0 the value is always null.

50 (deprecated)

Supported since 0.1.5 (2017-07-19). Deprecated in php-ast 1.1.0.

This is the oldest AST version available in 1.0.0. See the 0.1.x AST version changelog for information on changes prior to this version.

Differences to PHP-Parser

Next to php-ast I also maintain the PHP-Parser library, which has some overlap with this extension. This section summarizes the main differences between php-ast and PHP-Parser so you may decide which is preferable for your use-case.

The primary difference is that php-ast is a PHP extension (written in C) which exports the AST internally used by PHP 7 and 8. PHP-Parser on the other hand is library written in PHP. This has a number of consequences:

• php-ast is significantly faster than PHP-Parser, because the AST construction is implemented in C.
• php-ast needs to be installed as an extension, on Linux either by compiling it manually or retrieving it from a package manager, on Windows by loading a DLL. PHP-Parser is installed as a Composer dependency.
• php-ast only runs on PHP >= 7.0 (php-ast 1.0.16 was the last version supporting php <= 7.1), as prior versions did not use an internal AST. PHP-Parser supports PHP >= 5.5.
• php-ast may only parse code that is syntactically valid on the version of PHP it runs on. This means that it's not possible to parse code using features of newer versions (e.g. PHP 7.4 code while running on PHP 7.2). Similarly, it is not possible to parse code that is no longer syntactically valid on the used version (e.g. some PHP 5 code may no longer be parsed -- however most code will work). PHP-Parser supports parsing both newer and older (up to PHP 5.2) versions.
• php-ast can only parse code which is syntactically valid, while PHP-Parser can provide a partial AST for code that contains errors (e.g., because it is currently being edited).
• php-ast only provides the starting line number (and for declarations the ending line number) of nodes, because this is the only part that PHP itself stores. PHP-Parser provides precise file offsets.

There are a number of differences in the AST representation and available support code:

• The PHP-Parser library uses a separate class for every node type, with child nodes stored as type-annotated properties. php-ast uses one class for everything, with children stored as arrays. The former approach is friendlier to developers because it has very good IDE integration. The php-ast extension does not use separate classes, because registering hundreds of classes was judged a no-go for a bundled extension.
• The PHP-Parser library contains various support code for working with the AST, while php-ast only handles AST construction. The reason for this is that implementing this support code in C is extremely complicated and there is little other benefit to implementing it in C. The main components that PHP-Parser offers that may be of interest are:
  • Node dumper (human readable representation): While the php-ast extension does not directly implement this, a ast_dump function is provided in the util.php file.
  • Pretty printer (converting the AST back to PHP code): This is not provided natively, but the php-ast-reverter package implements this functionality.
  • Name resolution (resolving namespace prefixes and aliases): There is currently no standalone package for this.
  • AST traversal / visitation: There is currently no standalone package for this either, but implementing a recursive AST walk is easy.

The tolerant-php-parser-to-php-ast project can convert the AST produced by tolerant-php-parser (Another pure PHP parser library) into the format used by the php-ast extension. This can be used as a slow fallback in case the php-ast extension is not available. It may also be used to produce a partial php-ast output for code with syntax errors.