A set of TSLint rules used on some Microsoft projects.
The project has been in use for over a year on multiple projects. Please report any bugs or false positives you might find!
See our Release Notes to find the latest new rules.
The Latest Development Version is available online.
npm install tslint-microsoft-contrib
Alternately, you can download the files directly from GitHub:
| TSLint version | tslint-microsoft-contrib version |
|---|---|
| >= 3.2.x | 2.x |
| 3.1.x | unsupported |
| 3.0.x | unsupported |
| 2.x | 1.x |
Add the new rulesDirectory to your tslint task:
grunt.initConfig({
tslint: {
options: {
rulesDirectory: 'node_modules/tslint-microsoft-contrib',
configuration: grunt.file.readJSON("tslint.json")
},
files: {
src: ['src/file1.ts', 'src/file2.ts']
}
}
})
The tslint.json file does not change format when using this package. Just add our rule definitions to your existing tslint.json file.
There certainly are a lot of options! Here are some links to get you started.
- Easiest Option - Our recommended ruleset is here: recommended_ruleset.js
- A nice blog post on the MSDN secure development blog can be found here: Automating Secure Development Lifecycle Checks in TypeScript with TSLint
- A wiki briefly describing the SDL and related rules is here: TSLint and the Microsoft Security Development Lifecycle
- And our configuration file with all options is available here: tslint.json
| Rule Name | Description | Since |
|---|---|---|
chai-prefer-contains-to-index-of |
Avoid Chai assertions that invoke indexOf and compare for a -1 result. It is better to use the chai .contain() assertion API instead because the failure message will be more clearer if the test fails. | 2.0.10 |
chai-vague-errors |
Avoid Chai assertions that result in vague errors. For example, asserting expect(something).to.be.true will result in the failure message "Expected true received false". This is a vague error message that does not reveal the underlying problem. It is especially vague in TypeScript because stack trace line numbers often do not match the source code. A better pattern to follow is the xUnit Patterns Assertion Message pattern. The previous code sample could be better written as expect(something).to.equal(true, 'expected something to have occurred'); |
1.0 |
export-name |
The name of the exported module must match the filename of the source file. This is case-sensitive but ignores file extension. Since version 1.0, this rule takes a list of regular expressions as a parameter. Any export name matching that regular expression will be ignored. For example, to allow an exported name like myChartOptions, then configure the rule like this: "export-name": [true, "myChartOptionsg"] | 0.0.3 |
function-name |
Applies a naming convention to function names and method names. You can configure the naming convention by passing parameters. Please note, the private-method-regex does take precedence over the static-method-regex, so a private static method must match the private-method-regex. The default values are: [ true, { "method-regex": "^[a-z][\w\d]+$", "private-method-regex": "^[a-z][\w\d]+$", "static-method-regex": "^[A-Z_\d]+$", "function-regex": "^[a-z][\w\d]+$" }] |
2.0.7 |
import-name |
The name of the imported module must match the name of the thing being imported. For example, it is valid to name imported modules the same as the module name: import Service = require('x/y/z/Service') and import Service from 'x/y/z/Service'. But it is invalid to change the name being imported, such as: import MyCoolService = require('x/y/z/Service') and import MyCoolService from 'x/y/z/Service'. Since version 2.0.9 it is possible to configure this rule with a list of exceptions. For example, to allow underscore to be imported as _, add this configuration: 'import-name': [ true, { 'underscore': '_' }] |
2.0.5 |
jquery-deferred-must-complete |
When a JQuery Deferred instance is created, then either reject() or resolve() must be called on it within all code branches in the scope. For more examples see the feature request. | 1.0 |
max-func-body-length |
Avoid long functions. The line count of a function body must not exceed the value configured within this rule's options. You can setup a general max function body length applied for every function/method/arrow function e.g. [true, 30] or set different maximum length for every type e.g. [true, { "func-body-length": 10 , "func-expression-body-length": 10 , "arrow-body-length": 5, "method-body-length": 15, "ctor-body-length": 5 }]. To specify a function name whose parameters you can ignore for this rule, pass a regular expression as a string(this can be useful for Mocha users to ignore the describe() function). Since version 2.0.9, you can also ignore single- and multi-line comments from the total function length, eg. [true, { "ignore-comments": true }] |
2.0.3 |
missing-jsdoc |
All files must have a top level JSDoc comment. A JSDoc comment starts with /** (not one more or one less asterisk) and a JSDoc at the 'top-level' appears without leading spaces. Trailing spaces are acceptable but not recommended. | 1.0 |
missing-optional-annotation |
Deprecated - This rule is now enforced by the TypeScript compiler. A parameter that follows one or more parameters marked as optional is not itself marked optional | 0.0.1 |
mocha-avoid-only |
Do not invoke Mocha's describe.only or it.only functions. These functions are useful ways to run a single unit test or a single test case during your build, but please be careful to not push these methods calls to your version control repositiory because it will turn off any of the other tests. | 1.0 |
mocha-no-side-effect-code |
All test logic in a Mocha test case should be within Mocha lifecycle method and not defined statically to execute when the module loads. Put all assignments and initialization statements in a before(), beforeEach(), after(), afterEach(), or it() function. Code executed outside of these lifecycle methods can throw exceptions before the test runner is initialized and can result in errors or even test runner crashes. This rule can be configured with a regex to ignore certain initializations. For example, to ignore any calls to RestDataFactory configure the rule with: [true, { ignore: '^RestDataFactory\\..*' }] |
2.0.10 |
mocha-unneeded-done |
A function declares a MochaDone parameter but only resolves it synchronously in the main function. The MochaDone parameter can be safely removed from the parameter list. | 2.0.10 |
no-backbone-get-set-outside-model |
Avoid using model.get('x') and model.set('x', value) Backbone accessors outside of the owning model. This breaks type safety and you should define getters and setters for your attributes instead. | 1.0 |
no-banned-terms |
Do not use banned terms: caller, callee, eval, arguments. These terms refer to functions or properties that should not be used, so it is best practice to simply avoid them. | 0.0.1 |
no-constant-condition |
Do not use constant expressions in conditions. Similar to the ESLint no-constant-condition rule | 1.0 |
no-control-regex |
Do not use control characters in regular expressions . Similar to the ESLint no-control-regex rule | 1.0 |
no-cookies |
Do not use cookies | 0.0.1 |
no-delete-expression |
Do not delete expressions. Only properties should be deleted | 0.0.2 |
no-disable-auto-sanitization |
Do not disable auto-sanitization of HTML because this opens up your page to an XSS attack. Specifically, do not use the execUnsafeLocalFunction or setInnerHTMLUnsafe functions. | 0.0.1 |
no-document-domain |
Do not write to document.domain. Scripts setting document.domain to any value should be validated to ensure that the value is on a list of allowed sites. Also, if your site deals with PII in any way then document.domain must not be set to a top-level domain (for example, live.com) but only to an appropriate subdomain (for example, billing.live.com). If you are absolutely sure that you want to set document.domain then add a tslint suppression comment for the line. For more information see the Phase 4 Verification page of the Microsoft SDL | 2.0.3 |
no-document-write |
Do not use document.write | 0.0.1 |
no-duplicate-case |
Do not use duplicate case labels in switch statements. Similar to the ESLint no-duplicate-case rule | 1.0 |
no-duplicate-parameter-names |
Deprecated - This rule is now enforced by the TypeScript compiler. Do not write functions or methods with duplicate parameter names | 0.0.1 |
no-empty-interfaces |
Do not use empty interfaces. They are compile-time only artifacts and they serve no useful purpose | 1.0 |
no-empty-line-after-opening-brace |
Avoid an empty line after an opening brace. | 2.0.6 |
no-exec-script |
Do not use the execScript functions | 0.0.1 |
no-for-in |
Avoid use of for-in statements. They can be replaced by Object.keys | 1.0 |
no-function-constructor-with-string-args |
Do not use the version of the Function constructor that accepts a string argument to define the body of the function | 0.0.1 |
no-function-expression |
Do not use function expressions; use arrow functions (lambdas) instead. In general, lambdas are simpler to use and avoid the confusion about what the 'this' references points to. Function expressions that contain a 'this' reference are allowed and will not create a failure. | 1.0 |
no-http-string |
Do not use strings that start with 'http:'. URL strings should start with 'https:'. Http strings can be a security problem and indicator that your software may suffer from cookie-stealing attacks. Since version 1.0, this rule takes a list of regular expressions as a parameter. Any string matching that regular expression will be ignored. For example, to allow http connections to example.com and examples.com, configure your rule like this: "no-http-string": [true, "http://www.example.com/?.*", "http://www.examples.com/?.*"\] | 0.0.3 |
no-increment-decrement |
Avoid use of increment and decrement operators particularly as part of complicated expressions | 0.0.1 |
no-inner-html |
Do not write values to innerHTML, outerHTML, or set HTML using the JQuery html() function. Writing values to innerHTML can expose your website to XSS injection attacks. All strings must be escaped before being rendered to the page. | 2.0.4 |
no-invalid-regexp |
Do not use invalid regular expression strings in the RegExp constructor. Similar to the ESLint no-invalid-regexp rule | 1.0 |
no-jquery-raw-elements |
Do not create HTML elements using JQuery and string concatenation. It is error prone and can hide subtle defects. Instead use the JQuery element API. | 2.0.8 |
no-missing-visibility-modifiers |
Deprecated - This rule is in the TSLint product as member-access. Class members (both fields and methods) should have visibility modifiers specified. THe Principle of Least Visibility guides us to prefer private methods and fields when possible. If a developer forgets to add a modifier then TypeScript assumes the element should be public, which is the wrong default choice. |
1.0 |
no-multiline-string |
Do not declare multiline strings | 0.0.1 |
no-multiple-var-decl |
Deprecated - This rule is now part of the base TSLint product as the rule named 'one-variable-per-declaration'. Do not use comma separated variable declarations | 1.0 |
no-octal-literal |
Do not use octal literals or escaped octal sequences | 0.0.1 |
no-regex-spaces |
Do not use multiple spaces in a regular expression literal. Similar to the ESLint no-regex-spaces rule | 1.0 |
no-relative-imports |
Do not use relative paths when importing external modules or ES6 import declarations. The advantages of removing all relative paths from imports is that 1) the import name will be consistent across all files and subdirectories so searching for usages is much easier. 2) Moving source files to different folders will not require you to edit your import statements. 3) It will be possible to copy and paste import lines between files regardless of the file location. And 4) version control diffs will be simplified by having overall fewer edits to the import lines. | 2.0.5 |
no-reserved-keywords |
Do not use reserved keywords as names of local variables, fields, functions, or other identifiers. Since version 2.0.9 this rule accepts a parameter called allow-quoted-properties. If true, interface properties in quotes will be ignored. This can be a useful way to avoid verbose suppress-warning comments for generated d.ts files. | 0.0.1, 2.0.9 |
no-single-line-block-comment |
Avoid single line block comments and use single line comments instead. Block comments do not nest properly and have no advantages over normal single-line comments | 2.0.10 |
no-sparse-arrays |
Do not use sparse arrays. Sparse arrays contain empty slots, most frequently due to multiple commas being used in an array literal. Based on the ESLint no-sparse-arrays rule | 1.0 |
no-stateless-class |
A stateless class represents a failure in the object oriented design of the system. A class without state is better modeled as a module or given some state. A stateless class is defined as a class with only static members and no parent class. | 2.0.4 |
no-string-based-set-immediate |
Do not use the version of setImmediate that accepts code as a string argument. However, it is acceptable to use the version of setImmediate where a direct reference to a function is provided as the callback argument | 0.0.1 |
no-string-based-set-interval |
Do not use the version of setInterval that accepts code as a string argument. However, it is acceptable to use the version of setInterval where a direct reference to a function is provided as the callback argument | 0.0.1 |
no-string-based-set-timeout |
Do not use the version of setTimeout that accepts code as a string argument. However, it is acceptable to use the version of setTimeout where a direct reference to a function is provided as the callback argument | 0.0.1 |
no-suspicious-comment |
Do not use suspicious comments, such as BUG, HACK, FIXME, LATER, LATER2, TODO. We recommend that you run this rule before each release as a quality checkpoint. Reference: CWE-546 Suspicious Comment | 2.0.11 |
no-typeof-undefined |
Do not use the idiom typeof x === 'undefined'. You can safely use the simpler x === undefined or perhaps x == null if you want to check for either null or undefined. |
2.0.8 |
no-unexternalized-strings |
Ensures that double quoted strings are passed to a localize call to provide proper strings for different locales. The rule can be configured using an object literal as document in the feature request | 2.0.1 |
no-unnecessary-bind |
Do not bind 'this' as the context for a function literal or lambda expression. If you bind 'this' as the context to a function literal, then you should just use a lambda without the bind. If you bind 'this' as the context to a lambda, then you can remove the bind call because 'this' is already the context for lambdas. Works for Underscore methods as well. | 1.0 |
no-unnecessary-field-initialization |
Do not unnecessarily initialize the fields of a class to values they already have. For example, there is no need to explicitly set a field to undefined in the field's initialization or in the class' constructor. Also, if a field is initialized to a constant value (null, a string, a boolean, or some number) then there is no need to reassign the field to this value within the class constructor. |
2.0.9 |
no-unnecessary-local-variable |
Do not declare a variable only to return it from the function on the next line. It is always less code to simply return the expression that initializes the variable. | 2.0.4 |
no-unnecessary-override |
Do not write a method that only calls super() on the parent method with the same arguments. You can safely remove methods like this and Javascript will correctly dispatch the method to the parent object. | 2.0.4 |
no-unnecessary-semicolons |
Remove unnecessary semicolons | 0.0.1 |
no-unsupported-browser-code |
Avoid writing browser-specific code for unsupported browser versions. Browser versions are specified in the rule configuration options, eg: [true, [ "IE 11", "Firefox > 40", "Chrome >= 45" ] ]. Browser-specific blocks of code can then be designated with a single-line comment, like so: // Browser specific: IE 10, or with a jsdoc like this: @browserspecific chrome 40. |
2.0.10 |
no-unused-imports |
Deprecated - This rule is now covered by TSLint's no-unused-variables rule. However, it can still be useful to enable this rule and pair it with the fix-no-unused-imports formatter. | 0.0.1 |
no-with-statement |
Do not use with statements. Assign the item to a new variable instead | 0.0.1 |
no-var-self |
Do not use var self = this; instead, manage scope with arrow functions/lambdas. Self variables are a common practice in JavaScript but can be avoided in TypeScript. By default the rule bans any assignments of the this reference. If you want to enforce a naming convention or allow some usages then configure the rule with a regex. By default the rule is configured with (?!) which matches nothing. You can pass ^self$ to allow variables named self or pass ^(?!self$) to allow anything other than self, for example |
2.0.8 |
prefer-array-literal |
Use array literal syntax when declaring or instantiating array types. For example, prefer the Javascript form of string[] to the TypeScript form Array. Prefer '[]' to 'new Array()'. Prefer '[4, 5]' to 'new Array(4, 5)'. Prefer '[undefined, undefined]' to 'new Array(4)'. Since 2.0.10, this rule can be configured to allow Array type parameters. To ignore type parameters, configure the rule with the values: [ true, { 'allow-type-parameters': true } ] |
1.0, 2.0.10 |
prefer-const |
Use const to declare variables if they are only assigned a value once. |
2.0.6 |
prefer-type-cast |
Prefer the tradition type casts instead of the new 'as-cast' syntax. For example, prefer 'myVariable' instead of 'myVariable as string'. Rule ignores any file ending in .tsx. If you prefer the opposite and want to see the 'as type' casts, then enable the tslint rule named 'no-angle-bracket-type-assertion' | 2.0.4 |
promise-must-complete |
When a Promise instance is created, then either the reject() or resolve() parameter must be called on it within all code branches in the scope. For more examples see the feature request. | 1.0 |
react-iframe-missing-sandbox |
React iframes must specify a sandbox attribute. If specified as an empty string, this attribute enables extra restrictions on the content that can appear in the inline frame. The value of the attribute can either be an empty string (all the restrictions are applied), or a space-separated list of tokens that lift particular restrictions. You many not use both allow-scripts and allow-same-origin at the same time, as that allows the embedded document to programmatically remove the sandbox attribute in some scenarios. | 2.0.10 |
react-a11y-titles |
For accessibility of your website, HTML title elements must not be empty. References: * WCAG 2.0 - Requirement 2.4.2 Page Titled (Level A) * OAA-Accessibility Rule 13: Title element should not be empty |
2.0.11 |
react-no-dangerous-html |
Do not use React's dangerouslySetInnerHTML API. This rule finds usages of the dangerouslySetInnerHTML API (but not any JSX references). For more info see the react-no-dangerous-html Rule wiki page. | 0.0.2 |
react-this-binding-issue |
Several errors can occur when using React and React.Component subclasses. When using React components you must be careful to correctly bind the 'this' reference on any methods that you pass off to child components as callbacks. For example, it is common to define a private method called 'onClick' and then specify onClick={this.onClick} as a JSX attribute. If you do this then the 'this' reference will be undefined when your private method is invoked. The React documentation suggests that you bind the 'this' reference on all of your methods within the constructor: this.onClick = this.onClick.bind(this);. This rule will create a violation if 1) a method reference is passed to a JSX attribute without being bound in the constructor. And 2) a method is bound multiple times in the constructor. Another issue that can occur is binding the 'this' reference to a function within the render() method. For example, many people will create an anonymous lambda within the JSX attribute to avoid the 'this' binding issue: onClick={() => { this.onClick(); }}. The problem with this is that a new instance of an anonymous function is created every time render() is invoked. When React compares virutal DOM properties within shouldComponentUpdate() then the onClick property will look like a new property and force a re-render. You should avoid this pattern because creating function instances within render methods breaks any logic within shouldComponentUpdate() methods. This rule creates violations if 1) an anonymous function is passed as a JSX attribute. And 2) if a function instantiated in local scope is passed as a JSX attribute. This rule can be configured via the "allow-anonymous-listeners" parameter. If you want to suppress violations for the anonymous listener scenarios then configure that rule like this: "react-this-binding-issue": [ true, { 'allow-anonymous-listeners': true } ] |
2.0.8, 2.0.9 |
react-unused-props-and-state |
Remove unneeded properties defined in React Props and State interfaces. Any interface named Props or State is defined as a React interface. All fields in these interfaces must be referenced. | 2.0.10 |
underscore-consistent-invocation |
Enforce a consistent usage of the _ functions. By default, invoking underscore functions should begin with wrapping a variable in an underscore instance: _(list).map(...). An alternative is to prefer using the static methods on the _ variable: _.map(list, ...). The rule accepts single parameter called 'style' which can be the value 'static' or 'instance': [true, { "style": "static" }] |
2.0.10 |
use-isnan |
Deprecated - This rule is now part of the base TSLint product. Ensures that you use the isNaN() function to check for NaN references instead of a comparison to the NaN constant. Similar to the use-isnan ESLint rule. | 1.0 |
use-named-parameter |
Do not reference the arguments object by numerical index; instead, use a named parameter. This rule is similar to JSLint's Use a named parameter rule. | 0.0.3 |
valid-typeof |
Ensures that the results of typeof are compared against a valid string. This rule aims to prevent errors from likely typos by ensuring that when the result of a typeof operation is compared against a string, that the string is a valid value. Similar to the valid-typeof ESLint rule. | 1.0 |
These formatters assume that you use the UTF-8 file encoding. They may not work if you have a different encoding, especially if your encoding uses a 2-byte line ending (such as \r\n on Windows).
| Formatter Name | Description | Since |
|---|---|---|
fix-no-require-imports |
This formatter automatically converts imports from the require syntax to the ES6 syntax. For example import Utils = require('Utils'); becomes import {Utils} from 'Utils';. However, be warned that the fix assumes that your imported module exports the correct thing. If anything goes wrong with your exports then you'll get a compiler failure saying there is no default export. |
2.0.8 |
fix-no-unused-imports |
This formatter automatically fixes any unused imports found by the no-unused-imports rule. | 2.0.8 |
fix-no-var-keyword |
This formatter automatically converts var variable declarations into let variable declarations found by the no-var-keyword rule. | 2.0.8 |
fix-prefer-const |
This formatter automatically converts let variable declarations into const declarations found by the prefer-const rule. | 2.0.8 |
To develop tslint-microsoft-contrib simply clone the repository, install dependencies and run grunt:
git config --global core.autocrlf input
git config --global core.eol lf
git clone git@github.com:Microsoft/tslint-microsoft-contrib.git
cd tslint-microsoft-contrib
npm install
grunt all
grunt create-rule --rule-name=no-something-or-other
The repo requires you to use linefeed as a line ending, and asks that you checkout the code with linefeed endings enabled. This does change your global config, so please make a note of your previous values and revert them after the checkout if necessary.
If command fails because of file access permissions, prefix it with sudo.
npm install -g node-inspector
Then run:
node-debug grunt mochaTest
The node-debug command will load Node Inspector in your default browser (works in Chrome and Opera only).
Set a breakpoint somewhere in your code and resume execution. Your breakpoint should be hit.
Refer to the Releases Wiki Page
This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.