In words of Robert C. Martin, "Software architecture is the art of drawing lines that I call boundaries. Those boundaries separate software elements from one another, and restrict those on one side from knowing about those on the other." (*acknowledgements)
This plugin ensures that your architecture boundaries are respected by the elements in your project checking the folders and files structure and the dependencies between them. It is not a replacement for eslint-plugin-import, on the contrary, the combination of both plugins is recommended.
By default, the plugin works by checking import
statements, but it is also able to analyze "require", "exports" and dynamic imports, and can be configured to check any other AST nodes. (Read the main rules overview and configuration chapters for better comprehension)
This module is distributed via npm which is bundled with node and should be installed as one of your project's devDependencies:
npm install --save-dev eslint eslint-plugin-boundaries
eslint-plugin-boundaries
does not install eslint
for you. You must install it yourself.
Activate the plugin and one of the canned configs in your eslint.config.js
file:
import boundaries from "eslint-plugin-boundaries";
export default [
{
plugins: {
boundaries,
},
rules: {
...boundaries.configs.recommended.rules,
}
}
];
[!NOTE]
From version5.0.0
, this plugin is compatible with eslint v9 and above. It may be also compatible with previous eslint versions, but you might read the documentation of the4.2.2
version to know how to configure it properly using the legacy configuration format.
All of the plugin rules need to be able to identify the elements in the project, so, first of all you have to define your project element types by using the boundaries/elements
setting.
The plugin will use the provided patterns to identify each file as one of the element types. It will also assign a type to each dependency detected in the dependency nodes (import
or other statements), and it will check if the relationship between the dependent element and the dependency is allowed or not.
export default [{
settings: {
"boundaries/elements": [
{
type: "helpers",
pattern: "helpers/*"
},
{
type: "components",
pattern: "components/*"
},
{
type: "modules",
pattern: "modules/*"
}
]
}
}]
This is only a basic example of configuration. The plugin can be configured to identify elements being a file, or elements being a folder containing files. It also supports capturing path fragments to be used afterwards on each rule options, etc. Read the configuration chapter for further info, as configuring it properly is crucial to take advantage of all of the plugin features.
Once your project element types are defined, you can use them to configure each rule using its own options. For example, you could define which elements can be dependencies of other ones by configuring the element-types
rule as in:
export default [{
rules: {
"boundaries/element-types": [2, {
default: "disallow",
rules: [
{
from: "components",
allow: ["helpers", "components"]
},
{
from: "modules",
allow: ["helpers", "components", "modules"]
}
]
}]
}
}]
The plugin won't apply rules to a file or dependency when it does not recognize its element type, but you can force all files in your project to belong to an element type by enabling the boundaries/no-unknown-files rule.
This rule ensures that dependencies between your project element types are allowed.
Examples of usage:
Read the docs of the boundaries/element-types
rule for further info.
External dependencies used by each type of element in your project can be checked using this rule. For example, you can define that "helpers" can't import react
, or "components" can't import react-router-dom
, or modules can't import { Link } from react-router-dom
.
Read the docs of the boundaries/external
rule for further info.
This rule ensures that elements can't require other element's children. So, when an element B is children of A, B becomes a "private" element of A, and only A can use it.
Read the docs of the boundaries/no-private
rule for further info.
This rule ensures that elements can't import another file from other element than the defined entry point for that type (index.js
by default)
Read the docs of the boundaries/entry-point
rule for further info.
boundaries/element-types
Define patterns to recognize each file in the project as one of this element types. All rules need this setting to be configured properly to work. The plugin tries to identify each file being analyzed or import
statement in rules as one of the defined element types. The assigned element type will be that with the first matching pattern, in the same order that elements are defined in the array, so you should sort them from the most accurate patterns to the less ones. Properties of each element
:
type
: <string>
Element type to be assigned to files or imports matching the pattern
. This type will be used afterwards in the rules configuration.pattern
: <string>|<array>
micromatch
pattern. By default the plugin will try to match this pattern progressively starting from the right side of each file path. This means that you don't have to define patterns matching from the base project path, but only the last part of the path that you want to be matched. This is made because the plugin supports elements being children of other elements, and otherwise it could wrongly recognize children elements as a part of the parent one. src/helpers/awesome-helper/index.js
, it will try to assign the element to a pattern matching index.js
, then awesome-helper/index.js
, then helpers/awesome-helper/index.js
, etc. Once a pattern matches, it assign the correspondent element type, and continues searching for parents elements with the same logic until the full path has been analyzed. This behavior can be disabled setting the mode
option to full
, then the provided pattern will try to match the full path.basePattern
: <string>
Optional micromatch
pattern. If provided, the left side of the element path must match also with this pattern from the root of the project (like if pattern is [basePattern]/**/[pattern]
). This option is useful when using the option mode
with file
or folder
values, but capturing fragments from the rest of the full path is also needed (see baseCapture
option below).mode
: <string> file|folder|full
Optional.
folder
(default value), the element type will be assigned to the first file's parent folder matching the pattern. In the practice, it is like adding **/*
to the given pattern, but the plugin makes it by itself because it needs to know exactly which parent folder has to be considered the element.file
, the given pattern will not be modified, but the plugin will still try to match the last part of the path. So, a pattern like *.model.js
would match with paths src/foo.model.js
, src/modules/foo/foo.model.js
, src/modules/foo/models/foo.model.js
, etc.full
, the given pattern will only match with patterns matching the full path. This means that you will have to provide patterns matching from the base project path. So, in order to match src/modules/foo/foo.model.js
you'll have to provide patterns like **/*.model.js
, **/*/*.model.js
, src/*/*/*.model.js
, etc. (the chosen pattern will depend on what do you want to capture from the path)capture
: <array>
Optional. This is a very powerful feature of the plugin. It allows to capture values of some fragments in the matching path to use them later in the rules configuration. It uses micromatch
capture feature under the hood, and stores each value in an object with the given capture
key being in the same index of the captured array.pattern: "helpers/*/*.js"
, capture: ["category", "elementName"]
, and a path helpers/data/parsers.js
, it will result in { category: "data", elementName: "parsers" }
.baseCapture
: <array>
Optional. micromatch
pattern. It allows capturing values from basePattern
as capture
does with pattern
. All keys from capture
and baseCapture
can be used in the rules configuration.export default [{
settings: {
"boundaries/elements": [
{
type: "helpers",
pattern: "helpers/*/*.js",
mode: "file",
capture: ["category", "elementName"]
},
{
type: "components",
pattern: "components/*/*",
capture: ["family", "elementName"]
},
{
type: "modules",
pattern: "module/*",
capture: ["elementName"]
}
]
}
}]
Tip: You can enable the debug mode when configuring the plugin, and you will get information about the type assigned to each file in the project, as well as captured properties and values.
boundaries/dependency-nodes
This setting allows to modify built-in default dependency nodes. By default, the plugin will analyze only the import
statements. All the rules defined for the plugin will be applicable to the nodes defined in this setting.
The setting should be an array of the following strings:
'require'
: analyze require
statements.'import'
: analyze import
statements.'export'
: analyze export
statements.'dynamic-import'
: analyze dynamic import statements.If you want to define custom dependency nodes, such as jest.mock(...)
, use additional-dependency-nodes setting.
For example, if you want to analyze the import
and dynamic-import
statements, you should use the following value:
"boundaries/dependency-nodes": ["import", "dynamic-import"],
boundaries/additional-dependency-nodes
This setting allows to define custom dependency nodes to analyze. All the rules defined for the plugin will be applicable to the nodes defined in this setting.
The setting should be an array of objects with the following structure:
selector
: The esquery selector for the Literal
node in which dependency source are defined. For example, to analyze jest.mock(...)
calls you could use this AST selector: CallExpression[callee.object.name=jest][callee.property.name=mock] > Literal:first-child
.kind
: The kind of dependency, possible values are: "value"
or "type"
. It is available only when using TypeScript.Example of usage:
export default [{
settings: {
"boundaries/additional-dependency-nodes": [
// jest.requireActual('source')
{
selector: "CallExpression[callee.object.name=jest][callee.property.name=requireActual] > Literal",
kind: "value",
},
// jest.mock('source', ...)
{
selector: "CallExpression[callee.object.name=jest][callee.property.name=mock] > Literal:first-child",
kind: "value",
},
],
}
}]
boundaries/include
Files or dependencies not matching these micromatch
patterns will be ignored by the plugin. If this option is not provided, all files will be included.
export default [{
settings: {
"boundaries/include": ["src/**/*.js"]
}
}]
boundaries/ignore
Files or dependencies matching these micromatch
patterns will be ignored by the plugin.
export default [{
settings: {
"boundaries/ignore": ["**/*.spec.js", "src/legacy-code/**/*"]
}
}]
Note: The
boundaries/ignore
option has precedence overboundaries/include
. If you defineboundaries/include
, useboundaries/ignore
to ignore subsets of included files.
boundaries/root-path
Use this setting only if you are facing issues with the plugin when executing the lint command from a different path than the project root.
The plugin is distributed with two different predefined configurations: "recommended" and "strict".
We recommend to use this setting if you are applying the plugin to an already existing project. Rules boundaries/no-unknown
, boundaries/no-unknown-files
and boundaries/no-ignored
are disabled, so it allows to have parts of the project non-compliant with your element types, allowing to refactor the code progressively.
import boundaries from "eslint-plugin-boundaries";
export default [{
rules: {
...boundaries.configs.recommended.rules,
}
}]
All rules are enabled, so all elements in the project will be compliant with your architecture boundaries. 😃
import boundaries from "eslint-plugin-boundaries";
export default [{
rules: {
...boundaries.configs.strict.rules,
}
}]
Some rules require extra configuration, and it has to be defined in each specific rule
property of the eslint.config.js
file. For example, allowed element types relationships has to be provided as an option to the boundaries/element-types
rule. Rules requiring extra configuration will print a warning in case they are enabled without the needed options.
The docs of each rule contains an specification of their own options, but the main rules share the format in which the options have to be defined. The format described here is valid for options of element-types
, external
and entry-point
rules.
Options set an allow
or disallow
value by default, and provide an array of rules. Each matching rule will override the default value and the value returned by previous matching rules. So, the final result of the options, once processed for each case, will be allow
or disallow
, and this value will be applied by the plugin rule in the correspondent way, making it to produce an eslint error or not.
export default [{
rules: {
"boundaries/element-types": [2, {
// Allow or disallow any dependency by default
default: "allow",
// Define a custom message for this rule
message: "${file.type} is not allowed to import ${dependency.type}",
rules: [
{
// In this type of files...
from: ["helpers"],
// ...disallow importing this type of elements
disallow: ["modules", "components"],
// ..for this kind of imports (applies only when using TypeScript)
importKind: "value",
// ...and return this custom error message
message: "Helpers must not import other thing than helpers"
},
{
from: ["components"],
disallow: ["modules"]
// As this rule has not "message" property, it will use the message defined at first level
}
]
}]
}
}]
Remember that:
allow
and disallow
properties, the disallow
one has priority. It will not try to match the allow
one if disallow
matches. The result for that rule will be disallow
in that case.from/target
: <element matchers>
Depending of the rule to which the options are for, the rule will be applied only if the file being analyzed matches with this element matcher (from
), or the dependency being imported matches with this element matcher (target
).disallow/allow
: <value matchers>
If the plugin rule target matches with this, then the result of the rule will be "disallow/allow". Each rule will require a type of value here depending of what it is checking. In the case of the element-types
rule, for example, another <element matcher>
has to be provided in order to check the type of the local dependency.importKind
: <string>
Optional. It is useful only when using TypeScript, because it allows to define if the rule applies when the dependency is being imported as a value or as a type. It can be also defined as an array of strings, or a micromatch pattern. Note that possible values to match with are "value"
, "type"
or "typeof"
. For example, you could define that "components" can import "helpers" as a value, but not as a type. So, import { helper } from "helpers/helper-a"
would be allowed, but import type { Helper } from "helpers/helper-a"
would be disallowed.message
: <string>
Optional. If the rule results in an error, the plugin will return this message instead of the default one. Read error messages for further info.Tip: Properties
from/target
anddisallow/allow
can receive a single matcher, or an array of matchers.
Elements matchers used in the rules options can have the next formats:
<string>
: Will return true
when the element type matches with this micromatch
pattern. It supports templating for using values from captured values.[<string>, <capturedValuesObject>]
: Will return true
whe when the element type matches with the first element in the array, and all of the captured values also match. <capturedValuesObject>
has to be an object containing capture
keys from the boundaries/element-types
setting of the element as keys, and micromatch
patterns as values. (values also support templating) { type: "helpers", pattern": "helpers/*/*.js", "capture": ["category", "elementName"]}
, you could write element matchers as:
["helpers", { category: "data", elementName: "parsers"}]
: Will only match with helpers with category "data" and elementName "parsers" (helpers/data/parsers.js
).["helpers", { category: "data" }]
: Will match with all helpers with category "data" (helpers/data/*.js
)["data-${from.elementName}", { category: "${from.category}" }]
: Will only match with helpers with the type equals to the elementName
of the file importing plus a data-
prefix, and the category being equal to the category
of the file importing the dependency.When defining Element matchers, the values captured both from the element importing ("from") and from the imported element ("target") are available to be replaced. They are replaced both in the main string and in the <capturedValuesObject>
.
Templates must be defined with the format ${from.CAPTURED_PROPERTY}
or ${target.CAPTURED_PROPERTY}
.
The plugin returns a different default message for each rule, check the documentation of each one for further info. But some rules support defining custom messages in their configuration, as seen in "Main format of rules options".
When defining custom messages, it is possible to provide information about the current file or dependency. Use ${file.PROPERTY}
or ${dependency.PROPERTY}
, and it will be replaced by the correspondent captured value from the file or the dependency:
{
"message": "${file.type}s of category ${file.category} are not allowed to import ${dependency.category}s"
// If the error was produced by a file with type "component" and captured value "category" being "atom", trying to import a dependency with category "molecule", the message would be:
// "components of category atom are not allowed to import molecules"
}
Available properties in error templates both from file
or dependency
are:
type
: Element's type.internalPath
: File path being analyzed or imported. Relative to the element's root path.source
: Available only for dependency
. The source of the import
statement as it is in the code.parent
: If the element is child of another element, it is also available in this property, which contains correspondent type
, internalPath
and captured properties as well.importKind
: Available only for dependency
when using TypeScript. It contains the kind of import being analyzed. Possible values are "value"
, "type"
or "typeof"
.Tip: Read "Global settings" for further info about how to capture values from elements.
Some rules also provide extra information about the reported error. For example, no-external
rules provides information about detected forbidden specifiers. This information is available using ${report.PROPERTY}
. Check each rule's documentation to know which report properties it provides:
{
"message": "Do not import ${report.specifiers} from ${dependency.source} in helpers"
}
Just to illustrate the high level of customization that the plugin supports, here is an example of advanced options for the boundaries/element-types
rule based on the previous global elements
settings example:
export default [{
rules: {
"boundaries/element-types": [2, {
// disallow importing any element by default
default: "disallow",
rules: [
{
// allow importing helpers files from helpers files
from: ["helpers"],
allow: ["helpers"]
},
{
// when file is inside an element of type "components"
from: ["components"],
allow: [
// allow importing components of the same family
["components", { family: "${from.family}" }],
// allow importing helpers with captured category "data"
["helpers", { category: "data" }],
]
},
{
// when component has captured family "molecule"
from: [["components", { family: "molecule" }]],
allow: [
// allow importing components with captured family "atom"
["components", { family: "atom" }],
],
},
{
// when component has captured family "atom"
from: [["components", { family: "atom" }]],
disallow: [
// disallow importing helpers with captured category "data"
["helpers", { category: "data" }]
],
// Custom message only for this specific error
message: "Atom components can't import data helpers"
},
{
// when file is inside a module
from: ["modules"],
allow: [
// allow importing any type of component or helper
"helpers",
"components"
]
},
{
// when module name starts by "page-"
from: [["modules", { elementName: "page-*" }]],
disallow: [
// disallow importing any type of component not being of family layout
["components", { family: "!layout" }],
],
// Custom message only for this specific error
message: "Modules with name starting by 'page-' can't import not layout components. You tried to import a component of family ${target.family} from a module with name ${from.elementName}"
}
]
}]
}
}]
"With the advent of module bundlers and the current state of modules and module syntax specs, it's not always obvious where import x from 'module' should look to find the file behind module." (**Quote from the eslint-plugin-import
docs)
This plugin uses eslint-module-utils/resolve
module under the hood, which is a part of the eslint-plugin-import
plugin. So the import/resolver
setting can be used to use custom resolvers for this plugin too.
Read the resolvers
chapter of the eslint-plugin-import
plugin for further info.
export default [{
settings: {
"import/resolver": {
"eslint-import-resolver-node": {},
"some-other-custom-resolver": { someConfig: "value" }
}
}
}]
This plugin can be used also in TypeScript projects using @typescript-eslint/eslint-plugin
. Follow next steps to configure it:
Install dependencies:
npm i --save-dev @typescript-eslint/eslint-plugin @typescript-eslint/parser eslint-import-resolver-typescript
Configure @typescript-eslint/parser
as parser, load the @typescript-eslint
plugin, and setup the eslint-import-resolver-typescript
resolver in the eslint.config.js
config file:
import boundaries from "eslint-plugin-boundaries";
import typescriptParser from "@typescript-eslint/parser";
import typescriptEslintPlugin from "@typescript-eslint/eslint-plugin";
export default [{
languageOptions: {
parser: typescriptParser,
},
plugins: {
"@typescript-eslint": typescriptEslintPlugin,
boundaries,
},
settings: {
"import/resolver": {
typescript: {
alwaysTryTypes: true,
},
},
},
}];
Note that
eslint-import-resolver-typescript
detects even custom paths defined in thetsconfig.json
file, so its usage is also compatible with this plugin.
In case you face any issue configuring it, you can also use this repository as a guide. It contains a fully working and tested example.
v5.0.0 release is compatible with eslint v9 and above. It may be also compatible with previous eslint versions, but you might read the documentation of the 4.2.2
version to know how to configure it properly using the legacy configuration format. You may also be interested on reading the eslint guide to migrate to v9.
v4.0.0 release introduced breaking changes. If you were using v3.x, you should read the "how to migrate from v3 to v4" guide.
v2.0.0 release introduced many breaking changes. If you were using v1.x, you should read the "how to migrate from v1 to v2" guide.
In order to help during the configuration process, the plugin can trace information about the files and imports being analyzed. The information includes the file path, the assigned element type, the captured values, etc. So, it can help you to check that your elements
setting works as expected. You can enable it using the ESLINT_PLUGIN_BOUNDARIES_DEBUG
environment variable.
ESLINT_PLUGIN_BOUNDARIES_DEBUG=1 npm run lint
* Quote from Robert C. Martin's book "Clean Architecture: A Craftsman's Guide to Software Structure and Design".
** This plugin uses internally the eslint-module-utils/resolve
module, which is a part of the eslint-plugin-import
plugin. Thanks to the maintainers of that plugin for their awesome work.
Contributors are welcome. Please read the contributing guidelines and code of conduct.
MIT, see LICENSE for details.