ASTq

Abstract Syntax Tree (AST) Query Engine

Installation

$ npm install astq

About

ASTq is an Abstract Syntax Tree (AST) query engine library for JavaScript, i.e., it allows you to query nodes of an arbitary AST-style hierarchical data structure with the help of a powerful XPath-inspired query language. ASTq can operate on arbitrary AST-style data structures through the help of pluggable access adapters.

Query Language

ASTq uses an XPath-inspired Domain Specific Language (DSL) for querying the supplied AST-style hierarchical data structure.

By Example

At its simplest form, a query looks like a POSIX filesystem path:

Foo/Bar/Quux

This means: query and return all nodes of type Quux, which in turn are childs of nodes of type Bar, which in turn are childs of nodes of type Foo, which in turn has to be the start node.

A little bit more sophisticated query, showing more features, like axis, filter and optional whitespaces for padding:

// Foo [ /Bar [ @bar == 'baz1' || @bar == 'baz2' ] && /Quux ]

This means: query and return all nodes anywhere under the start node which are of type Foo and which have both childs of type Bar -- and with an attribute bar of values baz1 or baz2 -- and childs of type Quux.

By Grammar

In general, a query consists of one or more individual query paths, separated by comma. A path consists of a mandatory initial query step and optionally zero or more subsequent query steps.

The difference between initial and subsequent query steps is that an initial query step does not need an axis while all subsequent query steps require it. A query step consists of an (optional) AST node search axis, a (mandatory) AST node type match, an (optional) result marker "!" and an (optional) AST node filter expression:

query            ::= path (, path)*
path             ::= step-initial step-subsequent*
step-initial     ::= axis? match result? filter?
step-subsequent  ::= axis  match result? filter?

The search axis can be either...

• / for direct child nodes, or
• // for any descendant nodes, or
• ./ for current node plus direct child nodes, or
• .// for current node plus any descendant nodes, or
• -/ for direct left sibling node, or
• -// for any left sibling nodes, or
• +/ for direct right sibling node, or
• +// for any right sibling nodes, or
• ~/ for direct left and right sibling nodes, or
• ~// for all left and right sibling nodes, or
• ../ for direct parent node, or
• ..// for any parent nodes, or
• <// for any preceding nodes, or
• >// for any following nodes.

As an illustrating example: given an AST of the following particular nodes, ...

      A
      |
  +-+-+-+-+
 / /  |  \ \
B  C  D  E  F
      |
   +--+--+
  /   |   \
 G    H    I
      |
    +-+-+
   /     \
  J       K

...the following queries and their result exist:

Start Node | Query | Result Node(s) -----------|----------|--------------------------------- D | / * | G, H, I D | // * | G, H, J, K, I D | ./ * | D, G, H, I D | .// * | D, G, H, J, K, I D | -/ * | C D | -// * | C, B D | +/ * | E D | +// * | E, F D | ~/ * | C, E D | ~// * | B, C, E, F H | ../ * | D H | ..// * | D, A H | <// * | G, D, C B A H | >// * | J, K, I, E, F

A search axis usually walks along the references between nodes (at least in case of ASTy based AST). But in case the underlying AST and its adapter uses typed references, you can optionally constrain the search axis to take only references matching the type id into account.

axis               ::= axis-direction axis-type?
axis-direction     ::= axis-child
                     | axis-sibling-left
                     | axis-sibling-right
                     | axis-sibling
                     | axis-parent
                     | axis-preceding
                     | axis-following
axis-child         ::= ("/" | "//" | "./" | ".//")
axis-sibling-left  ::= ("-/" | "-//")
axis-sibling-right ::= ("+/" | "+//")
axis-sibling       ::= ("~/" | "~//")
axis-parent        ::= ("../" | "..//")
axis-preceding     ::= "<//"
axis-following     ::= ">//"
axis-type          ::= ":" (id | string)
result             ::= "!"
match              ::= id | string | "*"
filter             ::= "[" expr "]"

The real power comes through the optional filter expression: it can be applied to each query step and it recursively(!) can contain sub-queries with the help of embedded query paths! An illustrating combined example is:

// Foo / Bar [ / Baz [ @bar == 'baz' ] && / Quux ], // Foo2
+---------------------------------------------------------+  query
+------------------------------------------------+  +-----+  path
               +---------------------+    +-----+            path
+----+ +-----------------------------------------+  +-----+  step
++     +       +                          +         ++       axis
   +-+   +-+     +-+                        +--+       +--+  match
             +-----------------------------------+           filter
               +-------------------------------+             expr
                     +---------------+                       filter
                       +----------+                          expr

The result of a query is always all nodes which match against the last query step of any path (in case of no result marker on any step in the path) or all nodes of matched steps with a result marker. The queries in filter expressions just lead to a boolean decision for the filter, but never cause any resulting nodes theirself.

An expression can be either a ternary/binary conditional expression, logical expression, bitwise expression, relational expression, arithmethical expression, functional call, attribute reference, query parameter, literal value, parenthesis expression or path of a sub-query.

expr             ::= conditional
                   | logical
                   | bitwise
                   | relational
                   | arithmentical
                   | function-call
                   | attribute-ref
                   | query-parameter
                   | literal
                   | parenthesis
                   | sub-query
conditional      ::= expr "?" expr ":" expr
                   | expr "?:" expr
logical          ::= expr ("&&" | "||") expr
                   | "!" expr
bitwise          ::= expr ("&" | "|" | "<<" | ">>") expr
                   | "~" expr
relational       ::= expr ("==" | "!=" | "<=" | ">=" | "<" | ">" | "=~" | "!~") expr
arithmethical    ::= expr ("+" | "-" | "*" | "/" | "%" | "**") expr
function-call    ::= id "(" (expr ("," expr)*)? ")"
attribute-ref    ::= "@" (id | string)
query-parameter  ::= "{" id "}"
id               ::= /[a-zA-Z_][a-zA-Z0-9_-]*/
literal          ::= string | regexp | number | value
string           ::= /"(\\"|.)*"/ | /'(\\'|.)*'/
regexp           ::= /`(\\`|.)*`/
number           ::= /\d+(\.\d+)?$/
value            ::= "true" | "false" | "null" | "NaN" | "undefined"
parenthesis      ::= "(" expr ")"
sub-query        ::= path           // <-- ESSENTIAL RECURSION !!

Notice that the function call parameters can be full expressions theirself, including (through the recursion over sub-query above) full query paths. The available pre-defined standard functions are:

• type(): String: Return type of current node. Example: type() == "foo"

• attrs(sep: String): String: Return the sep-separated concatenation of all attribute names of current node. The sep string is alway also prepended and appended for easier comparison of the result string. Example: attr(",") == ",foo,bar,"

• depth(): Number: Return depth in AST of current node (counting from 1 for the root node). Example: depth() <= 3

• pos(): Number: Return position of current node among sibling (counting from 1 for the first sibling). Example: pos() == 2

• nth(pos: Number): Boolean: Check whether position of current node among sibling is pos (counting from 1 for the first sibling). Negative values for pos count from the last sibling backward, i.e., -1 is the last sibling. Example: nth(3)

• first(): Boolean: Shorthand for nth(1).

• last(): Boolean: Shorthand for nth(-1).

• count(array: Object[]): Number: Return the number of elements in array. The array usually is either an externally passed-in parameter or a sub-query. Example: count({nodes}) <= count(// *)

• below(node: Node): Boolean: Checks whether current node is somewhere below node, i.e., whether current node is a child or descendant of node. Usually, this makes sense only if node is an externally passed-in parameter. Example: below({node}).

• follows(node: Node): Boolean: Checks whether current node is following node, i.e., whether current node comes after node in a standard depth-first tree visit (where parents are visited before childs). Usually, this makes sense only if node is an externally passed-in parameter. Example: follows({node}).

• in(nodes: Node[]): Number: Checks whether current node is in nodes. Usually, nodes is either an externally passed-in parameter or a sub-query. Example: in({nodes}).

• substr(str: String, pos: Number, len: Number): String: Returns the sub-string of str, starting at pos with length len. Negative values for pos count from the end of the string, i.e., -1 is the last character. Example: substr(@foo, 0, 1) == "A"

• index(str: String, sub: String, pos: Number): Number: Returns the index position of sub-string sub in string str, starting at pos. Example: indexof(@foo, "bar", 0) >= 0

• trim(str: String): String: Returns the string str with whitespaces removed from begin and end. Example: trim(@foo) == "bar"

• lc(str: String): String: Returns the lower-case variant of str. Example: lc(@foo) == "bar"

• uc(str: String): String: Returns the upper-case variant of str. Example: uc(@foo) == "BAR"

Application Programming Interface (API)

The ASTq API, here assumed to be exposed through the variable ASTQ, provides the following methods (in a notation somewhat resembling TypeScript type definitions):

ASTQ API

• new ASTQ(): ASTQ: Create a new ASTQ instance.

• ASTQ#adapter(adapter: (ASTQAdapter | ASTQAdapter[]), force: Boolean): ASTQ: Register one or more custom tree access adapter(s) to support arbitrary AST-style data structures. The ASTQAdapter has to conform to a particular duck-typed interface. See below for more information. By default ASTq has built-in adapters for ASTy, XML DOM, Parse5, Cheerio, UniST, JSON and Mozilla AST. All those "taste" the node passed to ASTQ#query and hence are auto-selected. Calling adapter() causes these to be replaced with a single custom adapter. Its "tasting" can be disabled with option force set to true. The ASTQ#adapter teturns the API itself.

    /*  the built-in implementation for supporting ASTy  */
    astq.adapter({
        taste:            function (node)       { return (typeof node === "object" && node.ASTy) },
        getParentNode:    function (node, type) { return node.parent()  },
        getChildNodes:    function (node, type) { return node.childs()  },
        getNodeType:      function (node)       { return node.type()    },
        getNodeAttrNames: function (node)       { return node.attrs()   },
        getNodeAttrValue: function (node, attr) { return node.get(attr) }
    })

• ASTQ#version(): { major: Number, minor: Number, micro: Number, date: Number }: Return the current ASTq library version details.

• ASTQ#func(name: String, func: (adapter: Adapter, node: Object, [...]) => Any): ASTQ: Register function named name by providing the callback func which has to return an arbitrary value and optionally can access the current node with the help of the selected adapter. Returns the API itself.

    /*  the built-in implementation for "depth"  */
    astq.func("depth", function (adapter, node) => {
        var depth = 1
        while ((node = adapter.getParentNode(node)) !== null)
            depth++
        return depth
    })

• ASTQ#cache(num: Number): ASTQ: Set the upper limit for the internal query cache to num, i.e., up to num ASTs of parsed queries will be cached. Set num to 0 to disable the cache at all. Returns the API itself.

• ASTQ#compile(selector: String, trace?: Boolean): ASTQQuery { Compile selectorDSL into an internal query object for subsequent processing byASTQ#execute. If traceistrue` the compiling is dumped to the console. Returns the query object.

• ASTQ#execute(node: Object, query: ASTQQuery, params?: Object, trace?: Boolean): Object[]: Execute the previously compiled query (see compile above) at node. The optional params object can provide parameters for the {name} query constructs. If trace is true the execution is dumped to the console. Returns an array of zero or more matching AST nodes.

• ASTQ#query(node: Object, selector: String, params?: Object, trace?: Boolean): Object[]: Just the convenient combination of compile and execute: execute(node, compile(selector, trace), params, trace). Use this as the standard query method except you need more control. The optional params object can provide parameters for the {name} query constructs. If trace is true the compiling and execution is dumped to the console. Returns an array of zero or more matching AST nodes.

ASTQAdapter API

For accessing arbitrary AST-style data structures, an adapter has to be provided. By default ASTq has adapters for use with ASTy, XML DOM, Parse5, Cheerio, UniST, JSON and Mozilla AST. The ASTQAdapter interface is:

• ASTQAdapter#taste(node: Object): Boolean: Taste node to be sure this adapter is intended to handle it.

• ASTQAdapter#getParentNode(node: Object): Object: Return parent node of node. In case the underyling data structure does not support traversing to parent nodes, throw an exception.

• ASTQAdapter#getChildNodes(node: Object): Object[]: Return the list of all child nodes of node.

• ASTQAdapter#getNodeType(node: Object): String: Return the type of node.

• ASTQAdapter#getNodeAttrNames(node: Object): String[]: Return the list of all attribute names of node.

• ASTQAdapter#getNodeAttrValue(node: Object, attr: String): Any: Return the value of attribute attr of node.

Example

$ cat sample.js
const acorn = require("acorn")
const ASTQ  = require("astq")

let source = `
    class Foo {
        foo () {
            const bar = "quux"
            let baz = 42
        }
    }
`

let ast = acorn.parse(source, { ecmaVersion: 6 })

let astq = new ASTQ()
astq.adapter("mozast")
astq.query(ast, `
    // VariableDeclarator [
           /:id   Identifier [ @name  ]
        && /:init Literal    [ @value ]
    ]
`).forEach(function (node) {
    console.log(`${node.id.name}: ${node.init.value}`)
})

$ babel-node sample.js
bar: quux
baz: 42

Implementation Notice

Although ASTq is written in ECMAScript 2018, it is transpiled to older environments and this way runs in really all current (as of 2018) JavaScript environments, of course.

Additionally, there are two transpilation results: first, there is a compressed astq.browser.js for Browser environments. Second, there is an uncompressed astq.node.js for Node.js environments.

The Browser variant astq.browser.js has all external dependencies asty, pegjs-otf, pegjs-util, and cache-lru directly embedded. The Node.js variant astq.node.js still requires the external dependencies asty, pegjs-otf, pegjs-util, and cache-lru.

License

Copyright © 2014-2024 Dr. Ralf S. Engelschall (http://engelschall.com/)

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.