TSLint
A linter for the TypeScript language.
Installation
CLI
sudo npm install tslint -g
Library
npm install tslint
Usage
Please first ensure that the TypeScript source files compile correctly.
CLI
usage: tslint
Options:
-c, --config configuration file
-f, --file file to lint [required]
-o, --out output file
-r, --rules-dir rules directory
-s, --formatters-dir formatters directory
-t, --format output format (prose, json) [default: "prose"]
By default, configuration is loaded from tslint.json
, if it exists in the current path, or the user's home directory, in that order.
tslint accepts the following commandline options:
-f, --file:
The location of the TypeScript file that you wish to lint. This
option is required. Muliptle files are processed consecutively.
-c, --config:
The location of the configuration file that tslint will use to
determine which rules are activated and what options to provide
to the rules. If no option is specified, the config file named
tslint.json is used, so long as it exists in the path.
The format of the file is { rules: { /* rules list */ } },
where /* rules list */ is a key: value comma-seperated list of
rulename: rule-options pairs. Rule-options can be either a
boolean true/false value denoting whether the rule is used or not,
or a list [boolean, ...] where the boolean provides the same role
as in the non-list case, and the rest of the list are options passed
to the rule that will determine what it checks for (such as number
of characters for the max-line-length rule, or what functions to ban
for the ban rule).
-o, --out:
A filename to output the results to. By default, tslint outputs to
stdout, which is usually the console where you're running it from.
-r, --rules-dir:
An additional rules directory, for user-created rules.
tslint will always check its default rules directory, in
node_modules/tslint/build/rules, before checking the user-provided
rules directory, so rules in the user-provided rules directory
with the same name as the base rules will not be loaded.
-s, --formatters-dir:
An additional formatters directory, for user-created formatters.
Formatters are files that will format the tslint output, before
writing it to stdout or the file passed in --out. The default
directory, node_modules/tslint/build/formatters, will always be
checked first, so user-created formatters with the same names
as the base formatters will not be loaded.
-t, --format:
The formatter to use to format the results of the linter before
outputting it to stdout or the file passed in --out. The core
formatters are prose (human readable) and json (machine readable),
and prose is the default if this option is not used. Additional
formatters can be added and used if the --formatters-dir option
is set.
--help:
Prints this help message.
Library
var options = {
formatter: "json",
configuration: configuration,
rulesDirectory: "customRules/",
formattersDirectory: "customFormatters/"
};
var Linter = require("tslint");
var ll = new Linter(fileName, contents, options);
var result = ll.lint();
Supported Rules
A sample configuration file with all options is available here.
ban
bans the use of specific functions. Options are ["object", "function"] pairs that ban the use of object.function()class-name
enforces PascalCased class and interface names.comment-format
enforces rules for single-line comments. Rule options:"check-space"
enforces the rule that all single-line comments must begin with a space, as in// comment
- note that comments starting with
///
are also allowed, for things such as ///
- note that comments starting with
"check-lowercase"
enforces the rule that the first non-whitespace character of a comment must be lowercase, if applicable
curly
enforces braces forif
/for
/do
/while
statements.eofline
enforces the file to end with a newline.forin
enforces afor ... in
statement to be filtered with anif
statement.*indent
enforces consistent indentation levels (currently disabled).interface-name
enforces the rule that interface names must begin with a capital 'I'jsdoc-format
enforces basic format rules for jsdoc comments -- comments starting with/**
- each line contains an asterisk and asterisks must be aligned
- each asterisk must be followed by either a space or a newline (except for the first and the last)
- the only characters before the asterisk on each line must be whitepace characters
label-position
enforces labels only on sensible statements.label-undefined
checks that labels are defined before usage.max-line-length
sets the maximum length of a line.no-any
diallows usages ofany
as a type decoration.no-arg
disallows access toarguments.callee
.no-bitwise
disallows bitwise operators.no-console
disallows access to the specified functions onconsole
. Rule options are functions to ban on the console variable.no-consecutive-blank-lines
disallows having more than one blank line in a row in a fileno-construct
disallows access to the constructors ofString
,Number
, andBoolean
.no-debugger
disallowsdebugger
statements.no-duplicate-key
disallows duplicate keys in object literals.no-duplicate-variable
disallows duplicate variable declarations.no-empty
disallows empty blocks.no-eval
disallowseval
function invocations.no-string-literal
disallows object access via string literals.no-switch-case-fall-through
disallows falling through case statements.no-trailing-comma
disallows trailing comma within object literals.no-trailing-whitespace
disallows trailing whitespace at the end of a line.no-unreachable
disallows unreachable code afterbreak
,catch
,throw
, andreturn
statements.no-unused-expression
disallows unused expression statements, that is, expression statements that are not assignments or function invocations (and thus no-ops).no-unused-variable
disallows unused imports, variables, functions and private class members."check-parameters"
disallows unused function and constructor parameters.- NOTE: this option is experimental and does not work with classes that use abstract method declarations, among other things. Use at your own risk.
no-use-before-declare
disallows usage of variables before their declaration.no-var-requires
disallows the use of require statements except in import statements, banning the use of forms such asvar module = require("module")
one-line
enforces the specified tokens to be on the same line as the expression preceding it. Rule options:"check-catch"
checks thatcatch
is on the same line as the closing brace fortry
"check-else"
checks thatelse
is on the same line as the closing brace forif
"check-open-brace"
checks that an open brace falls on the same line as its preceding expression."check-whitespace"
checks preceding whitespace for the specified tokens.
quotemark
enforces consistent single or double quoted string literals.radix
enforces the radix parameter ofparseInt
semicolon
enforces semicolons at the end of every statement.triple-equals
enforces === and !== in favor of == and !=.typedef
enforces type definitions to exist. Rule options:"callSignature"
checks return type of functions"indexSignature"
checks index type specifier of indexers"parameter"
checks type specifier of parameters"propertySignature"
checks return types of interface properties"variableDeclarator"
checks variable declarations"memberVariableDeclarator"
checks member variable declarations
typedef-whitespace
enforces spacing whitespace for type definitions. Each rule option requires a value of"space"
or"nospace"
to require a space or no space before the type specifier's colon. Rule options:"callSignature"
checks return type of functions"catchClause"
checks type in exception catch blocks"indexSignature"
checks index type specifier of indexers
use-strict
enforces ECMAScript 5's strict modecheck-module
checks that all top-level modules are using strict modecheck-function
checks that all top-level functions are using strict mode
variable-name
allows only camelCased or UPPER_CASED variable names. Rule options:"allow-leading-underscore"
allows underscores at the beginnning.
whitespace
enforces spacing whitespace. Rule options:"check-branch"
checks branching statements (if
/else
/for
/while
) are followed by whitespace"check-decl"
checks that variable declarations have whitespace around the equals token"check-operator"
checks for whitespace around operator tokens"check-separator"
checks for whitespace after separator tokens (,
/;
)"check-type"
checks for whitespace before a variable type specification
TSLint Rule Flags
You can enable/disable TSLint or a subset of rules within a file with the following comment rule flags:
/* tslint:disable */
- Disable all rules for the rest of the file/* tslint:enable */
- Enable all rules for the rest of the file/* tslint:disable:rule1 rule2 rule3... */
- Disable the listed rules for the rest of the file/* tslint:enable:rule1 rule2 rule3... */
- Enable the listed rules for the rest of the file
Rules flags enable or disable rules as they are parsed. A rule is enabled or disabled until a later directive commands otherwise. Disabling an already disabled rule or enabling an already enabled rule has no effect.
For example, imagine the directive /* tslint:disable */
on the first line of a file, /* tslint:enable:ban class-name */
on the 10th line and /* tslint:enable */
on the 20th. No rules will be checked between the 1st and 10th lines, only the ban
and class-name
rules will be checked between the 10th and 20th, and all rules will be checked for the remainder of the file.
Custom Rules
TSLint ships with a set of core rules that can be configured. However, users are also allowed to write their own rules, which allows them to enforce specific behavior not covered by the core of TSLint. TSLint's internal rules are itself written to be pluggable, so adding a new rule is as simple as creating a new rule file named by convention. New rules can be written in either TypeScript or Javascript; if written in TypeScript, the code must be compiled to Javascript before invoking TSLint.
Rule names are always camel-cased and must contain the suffix Rule
. Let us take the example of how to write a new rule to forbid all import statements (you know, for science). Let us name the rule file noImportsRule.ts
. Rules can be referenced in tslint.json
in their dasherized forms, so "no-imports": true
would turn on the rule.
Now, let us first write the rule in TypeScript. At the top, we reference TSLint's definition file. The exported class name must always be named Rule
and extend from Lint.Rules.AbstractRule
.
/// <reference path='tslint.d.ts' />
export class Rule extends Lint.Rules.AbstractRule {
public static FAILURE_STRING = "import statement forbidden";
public apply(syntaxTree: TypeScript.SyntaxTree): Lint.RuleFailure[] {
return this.applyWithWalker(new NoImportsWalker(syntaxTree, this.getOptions()));
}
}
The walker takes care of all the work.
class NoImportsWalker extends Lint.RuleWalker {
public visitImportDeclaration(node: TypeScript.ImportDeclarationSyntax) {
// get the current position and skip over any leading whitespace
var position = this.position() + node.leadingTriviaWidth();
// create a failure at the current position
this.addFailure(this.createFailure(position, node.width(), Rule.FAILURE_STRING));
// call the base version of this visitor to actually parse this node
super.visitImportDeclaration(node);
}
}
Given a walker, TypeScript's parser visits the AST using the visitor pattern. So the rule walkers only need to override the appropriate visitor methods to enforce its checks. For reference, the base walker can be found in syntaxWalker.generated.ts within the TypeScript source code.
We still need to hook up this new rule to TSLint. First make sure to compile noImportsRule.ts
: tsc -m commonjs noImportsRule.ts tslint.d.ts
. Then, if using the CLI, provide the directory that contains this rule as an option to --rules-dir
. If using TSLint as a library or via grunt-tslint
, the options
hash must contain "rulesDirectory": "..."
. If you run the linter, you'll see that we have now successfully banned all import statements via TSLint!
Now, let us rewrite the same rule in Javascript.
function Rule() {
Lint.Rules.AbstractRule.apply(this, arguments);
}
Rule.prototype = Object.create(Lint.Rules.AbstractRule.prototype);
Rule.prototype.apply = function(syntaxTree) {
return this.applyWithWalker(new NoImportsWalker(syntaxTree, this.getOptions()));
};
function NoImportsWalker() {
Lint.RuleWalker.apply(this, arguments);
}
NoImportsWalker.prototype = Object.create(Lint.RuleWalker.prototype);
NoImportsWalker.prototype.visitImportDeclaration = function (node) {
// get the current position and skip over any leading whitespace
var position = this.position() + node.leadingTriviaWidth();
// create a failure at the current position
this.addFailure(this.createFailure(position, node.width(), "import statement forbidden"));
// call the base version of this visitor to actually parse this node
Lint.RuleWalker.prototype.visitImportDeclaration.call(this, node);
};
exports.Rule = Rule;
As you can see, it's a pretty straightforward translation from the equivalent TypeScript code.
Finally, core rules cannot be overwritten with a custom implementation, and rules can also take in options (retrieved via this.getOptions()
).
Custom Formatters
Just like rules, additional formatters can also be supplied to TSLint via --formatters-dir
on the CLI or formattersDirectory
option on the library or grunt-tslint
. Writing a new formatter is simpler than writing a new rule, as shown in the JSON formatter's code.
/// <reference path='tslint.d.ts' />
export class Formatter extends Lint.Formatters.AbstractFormatter {
public format(failures: Lint.RuleFailure[]): string {
var failuresJSON: any[] = [];
for (var i = 0; i < failures.length; ++i) {
failuresJSON.push(failures[i].toJson());
}
return JSON.stringify(failuresJSON);
}
}
Such custom formatters can also be written in Javascript. Additionally, formatter files are always named with the suffix Formatter
, and referenced from TSLint without its suffix.
Development
To develop tslint simply clone the repository, install dependencies and run grunt:
git clone git@github.com:palantir/tslint.git
npm install
grunt
Creating a new Release
- Bump up the version number in package.json and tslint.ts
- Add a section for the new release in CHANGELOG.md
- Run
grunt
to build the latest sources - Commit
- Run
npm publish
- Create a git tag for the new release and push it