jscodeshift is a toolkit for running codemods over multiple JavaScript or Typescript files. It provides:
- A runner, which executes the provided transform for each file passed to it. It also outputs a summary of how many files have (not) been transformed.
- A wrapper around recast, providing a different API. Recast is an AST-to-AST transform tool and also tries to preserve the style of original code as much as possible.
Get jscodeshift from npm:
$ npm install -g jscodeshift
This will install the runner as jscodeshift
.
The CLI provides the following options:
$ jscodeshift --help
Usage: jscodeshift [OPTION]... PATH...
or: jscodeshift [OPTION]... -t TRANSFORM_PATH PATH...
or: jscodeshift [OPTION]... -t URL PATH...
or: jscodeshift [OPTION]... --stdin < file_list.txt
Apply transform logic in TRANSFORM_PATH (recursively) to every PATH.
If --stdin is set, each line of the standard input is used as a path.
Options:
"..." behind an option means that it can be supplied multiple times.
All options are also passed to the transformer, which means you can supply custom options that are not listed here.
--(no-)babel apply babeljs to the transform file
(default: true)
-c, --cpus=N start at most N child processes to process source files
(default: max(all - 1, 1))
-d, --(no-)dry dry run (no changes are made to files)
(default: false)
--extensions=EXT transform files with these file extensions (comma separated list)
(default: js)
-h, --help print this help and exit
--ignore-config=FILE ... ignore files if they match patterns sourced from a configuration file (e.g. a .gitignore)
--ignore-pattern=GLOB ... ignore files that match a provided glob expression
--parser=babel|babylon|flow|ts|tsx the parser to use for parsing the source files
(default: babel)
--parser-config=FILE path to a JSON file containing a custom parser configuration for flow or babylon
-p, --(no-)print print transformed files to stdout, useful for development
(default: false)
--(no-)run-in-band run serially in the current process
(default: false)
-s, --(no-)silent do not write to stdout or stderr
(default: false)
--(no-)stdin read file/directory list from stdin
(default: false)
-t, --transform=FILE path to the transform file. Can be either a local path or url
(default: ./transform.js)
-v, --verbose=0|1|2 show more information about the transform process
(default: 0)
--version print version and exit
This passes the source of all passed through the transform module specified
with -t
or --transform
(defaults to transform.js
in the current
directory). The next section explains the structure of the transform module.
The transform is simply a module that exports a function of the form:
module.exports = function(fileInfo, api, options) {
// transform `fileInfo.source` here
// ...
// return changed source
return source;
};
As of v0.6.1, this module can also be written in Typescript.
Holds information about the currently processed file.
Property | Description |
---|---|
path | File path |
source | File content |
This object exposes the jscodeshift
library and helper functions from the
runner.
Property | Description |
---|---|
jscodeshift | A reference to the jscodeshift library |
stats | A function to collect statistics during --dry runs |
report | Prints the passed string to stdout |
jscodeshift
is a reference to the wrapper around recast and provides a
jQuery-like API to navigate and transform the AST. Here is a quick example,
a more detailed description can be found below.
/**
* This replaces every occurrence of variable "foo".
*/
module.exports = function(fileInfo, api) {
return api.jscodeshift(fileInfo.source)
.findVariableDeclarators('foo')
.renameTo('bar')
.toSource();
}
Note: This API is exposed for convenience, but you don't have to use it. You can use any tool to modify the source.
stats
is a function that only works when the --dry
options is set. It accepts
a string, and will simply count how often it was called with that value.
At the end, the CLI will report those values. This can be useful while developing the transform, e.g. to find out how often a certain construct appears in the source(s).
report
allows you do print arbitrary strings to stdout. This can be
useful when other tools consume the output of jscodeshift. The reason to not
directly use process.stdout
in transform code is to avoid mangled output when
many files are processed.
Contains all options that have been passed to runner. This allows you to pass additional options to the transform. For example, if the CLI is called with
$ jscodeshift -t myTransforms fileA fileB --foo=bar
options
would contain {foo: 'bar'}
.
The return value of the function determines the status of the transformation:
- If a string is returned and it is different from passed source, the transform is considered to be successful.
- If a string is returned but it's the same as the source, the transform is considered to be unsuccessful.
- If nothing is returned, the file is not supposed to be transformed (which is ok).
The CLI provides a summary of the transformation at the end. You can get more
detailed information by setting the -v
option to 1
or 2
.
You can collect even more stats via the stats
function as explained above.
The transform can let jscodeshift know with which parser to parse the source files (and features like templates).
To do that, the transform module can export parser
, which can either be one
of the strings "babel"
, "babylon"
, "flow"
, "ts"
, or "tsx"
,
or it can be a parser object that is compatible with recast.
For example:
module.exports.parser = 'flow'; // use the flow parser
// or
module.exports.parser = {
parse: function(source) {
// return estree compatible AST
},
};
$ jscodeshift -t myTransform.js src
Processing 10 files...
Spawning 2 workers with 5 files each...
All workers done.
Results: 0 errors 2 unmodified 3 skipped 5 ok
As already mentioned, jscodeshift also provides a wrapper around recast. In order to properly use the jscodeshift API, one has to understand the basic building blocks of recast (and ASTs) as well.
An AST node is a plain JavaScript object with a specific set of fields, in
accordance with the Mozilla Parser API. The primary way to identify nodes
is via their type
.
For example, string literals are represented via Literal
nodes, which
have the structure
// "foo"
{
type: 'Literal',
value: 'foo',
raw: '"foo"'
}
It's OK to not know the structure of every AST node type. The (esprima) AST explorer is an online tool to inspect the AST for a given piece of JS code.
Recast itself relies heavily on ast-types which defines methods to traverse the AST, access node fields and build new nodes. ast-types wraps every AST node into a path object. Paths contain meta-information and helper methods to process AST nodes.
For example, the child-parent relationship between two nodes is not explicitly
defined. Given a plain AST node, it is not possible to traverse the tree up.
Given a path object however, the parent can be traversed to via path.parent
.
For more information about the path object API, please have a look at ast-types.
To make creating AST nodes a bit simpler and "safer", ast-types defines a couple
of builder methods, which are also exposed on jscodeshift
.
For example, the following creates an AST equivalent to foo(bar)
:
// inside a module transform
var j = jscodeshift;
// foo(bar);
var ast = j.callExpression(
j.identifier('foo'),
[j.identifier('bar')]
);
The signature of each builder function is best learned by having a look at the definition files.
In order to transform the AST, you have to traverse it and find the nodes that need to be changed. jscodeshift is built around the idea of collections of paths and thus provides a different way of processing an AST than recast or ast-types.
A collection has methods to process the nodes inside a collection, often resulting in a new collection. This results in a fluent interface, which can make the transform more readable.
Collections are "typed" which means that the type of a collection is the
"lowest" type all AST nodes in the collection have in common. That means you
cannot call a method for a FunctionExpression
collection on an Identifier
collection.
Here is an example of how one would find/traverse all Identifier
nodes with
jscodeshift and with recast:
// recast
var ast = recast.parse(src);
recast.visit(ast, {
visitIdentifier: function(path) {
// do something with path
return false;
}
});
// jscodeshift
jscodeshift(src)
.find(jscodeshift.Identifier)
.forEach(function(path) {
// do something with path
});
To learn about the provided methods, have a look at the Collection.js and its extensions.
jscodeshift provides an API to extend collections. By moving common operators into helper functions (which can be stored separately in other modules), a transform can be made more readable.
There are two types of extensions: generic extensions and type-specific extensions. Generic extensions are applicable to all collections. As such, they typically don't access specific node data, but rather traverse the AST from the nodes in the collection. Type-specific extensions work only on specific node types and are not callable on differently typed collections.
// Adding a method to all Identifiers
jscodeshift.registerMethods({
logNames: function() {
return this.forEach(function(path) {
console.log(path.node.name);
});
}
}, jscodeshift.Identifier);
// Adding a method to all collections
jscodeshift.registerMethods({
findIdentifiers: function() {
return this.find(jscodeshift.Identifier);
}
});
jscodeshift(ast).findIdentifiers().logNames();
jscodeshift(ast).logNames(); // error, unless `ast` only consists of Identifier nodes
Passing options to recast
You may want to change some of the output settings (like setting '
instead of "
).
This can be done by passing config options to recast.
.toSource({quote: 'single'}); // sets strings to use single quotes in transformed code.
You can also pass options to recast's parse
method by passing an object to
jscodeshift as second argument:
jscodeshift(source, {...})
More on config options here
jscodeshift comes with a simple utility to allow easy unit testing with Jest, without having to write a lot of boilerplate code. This utility makes some assumptions in order to reduce the amount of configuration required:
- The test is located in a subdirectory under the directory the transform itself is located in (eg.
__tests__
) - Test fixtures are located in a
__testfixtures__
directory
This results in a directory structure like this:
/MyTransform.js
/__tests__/MyTransform-test.js
/__testfixtures__/MyTransform.input.js
/__testfixtures__/MyTransform.output.js
To define a test, use defineTest
or defineInlineTest
from the testUtils
module. A simple example is bundled in the sample directory.
jest.autoMockOff();
const defineTest = require('jscodeshift/dist/testUtils').defineTest;
defineTest(__dirname, 'MyTransform');
An alternate fixture filename can be provided as the fourth argument to defineTest
. This also means that multiple test fixtures can be provided:
defineTest(__dirname, 'MyTransform', null, 'FirstFixture');
defineTest(__dirname, 'MyTransform', null, 'SecondFixture');
This will run two tests: One for __testfixtures__/FirstFixture.input.js
and one for __testfixtures__/SecondFixture.input.js
const transform = require('../myTransform');
defineInlineTest(transform, {}, 'input', 'expected output', 'test name (optional)');
- react-codemod - React codemod scripts to update React APIs.
- js-codemod - Codemod scripts to transform code to next generation JS.
- js-transforms - Some documented codemod experiments to help you learn.
- Discord - Reactiflux