a simple configuration utility for NodeJS featuring smart module wiring for unobtrusive dependency injection
wires
augments the filename resolution mechanisms used by import
and require()
under the hood as a means to transparently inject configuration into your code.
For simplicity's sake, we only provide examples using require()
in this documentation but everything described here also applies to import
with minor specificities discussed in a dedicated section.
require( "#port" ); // <= http port of the server
require( ":db_driver" ); // <= database driver
require( ":models/user" ); // <= user model
require( ":models/article" ); // <= article model
Hash-lead expressions are used to recover settings. Colon-lead expressions are used to leverage routes and inject modules.
Routes and settings are defined in JSON files:
// the following JSON
{
"port": 80,
":db_driver": "mysql/driver",
":models/": "./db/model-"
}
// will yield the following results
require( "#port" ) === 80;
require( ":db_driver" ) === require( "mysql/driver" );
require( ":models/user" ) === require ( "/.../db/model-user" );
require( ":models/article" ) === require ( "/.../db/model-article" )
// this alternate JSON
{
"port": 8080,
":db_driver": "./lib/dbDriver",
":models/": "models/dbo/"
}
// will yield the following results
require( "#port" ) === 8080;
require( ":db_driver" ) === require( "/.../lib/dbDriver" );
require( ":models/user" ) === require ( "models/dbo/user" );
require( ":models/article" ) === require ( "models/dbo/article" )
Not that settings which keys are at-sign-lead are considered as directives and will not appear in your configuration. Currently supported directives are @namespace
and @root
.
Typically,
- install
wires
globally:npm -g install wires
- use the
wires
command in lieu ofnode
.
So, node ...
becomes wires ...
.
As of version 0.4.0
, wires
will look for local installations based on the current working directory. This means you can add wires as a dependency to your project and have this specific version take precedence whenever the global command is ran from within said project.
> wires --version
v0.4.0 (node v0.12.2)
> npm install --save wires@0.2.0
wires@0.2.0 node_modules\wires
├── temp@0.8.1 (rimraf@2.2.8)
└── lodash@3.6.0
> wires --version
v0.2.0 (node v0.12.2)
This behavior is compatible with version 0.2.0
and up.
If you don't want to install wires
globally, you should use
npx wires ...
which will find and execute your locally installed wires
command.
If you don't want to use the wires
command, you can use the node
executable as follows:
wires
>=5
:node --require=wires --loader=wires ...
wires
<5
:node --require=wires ...
This will enable both the CommonJS and ECMAScript module resolution override.
Alternatively, if you only need CommonJS support, you can manually require wires
at the beginning of the entry file of your project and use node
as usual:
require( "wires" );
// rest of your code
wires uses four types of configuration file:
wires-defaults.json
contains default settingswires-defaults.XXX.json
contains default settings specific to when theWIRES_ENV
environment variable is set to"XXX"
wires.json
contains actual settingswires.XXX.json
contains default settings specific to when theWIRES_ENV
environment variable is set to"XXX"
Prior to version 3.0
, wires
did use the NODE_ENV
environment variable rather than WIRES_ENV
. Please adjust your projects accordingly when upgrading to version 3.0
and higher.
Actual configuration depends on the position of the file requiring it in the filesystem hierarchy. Each directory gets its own configuration which is computed as follows:
- start with an empty object
- override with
wires-defaults.json
if it exists - if
WIRES_ENV
is set to"XXX"
, override withwires-defaults.XXX.json
if it exists - override with the configuration of the parent directory if it exists and if the
@root
directive isn't set totrue
- override with
wires.json
if it exists - if
WIRES_ENV
is set to"XXX"
, override withwires.XXX.json
if it exists
In practice, you'll rarely use all files at once. Typically, parts of your application will define default values in their respective directories while the main app will set actual settings in the root directory, as in the following example:
// /app/wires.json
{
"mysql": {
"user": "root",
"password": "{#>PASSWORD}",
"database": "test3"
}
}
// /app/database/wires-defaults.json
{
"mysql": {
"host": "localhost",
"port": 3306,
"user": "",
"password": ""
}
}
// /app/database/lib/someFile.js
require( "#mysql.host" ) === "localhost";
require( "#mysql.port" ) === 3306,
require( "#mysql.user" ) === "root",
require( "#mysql.password" ) === String( process.env.PASSWORD ),
require( "#mysql.database" ) === "test3"
Adding in one of your configuration files the directive @namespace
which value is a string creates a namespace. The namespace isolates the directory and limits its access to the part of the parent configuration defined within said namespace.
Let's go back to the previous example:
// /app/wires.json
{
"mysql": {
"user": "root",
"password": "{#>PASSWORD}",
"database": "test3"
}
}
// /app/database/wires-defaults.json
{
"@namespace": "mysql",
"host": "localhost",
"port": 3306,
"user": "",
"password": ""
}
// /app/database/lib/someFile.js
require( "#host" ) === "localhost";
require( "#port" ) === 3306,
require( "#user" ) === "root",
require( "#password" ) === String( process.env.PASSWORD ),
require( "#database" ) === "test3"
Note that routes are impervious to namespaces.
Sometimes, it is desirable to isolate your configuration. Set the @root
directive to true and wires
will stop climbing any further up the file hierarchy.
See the following example:
// /parent/wires.json
{
"parentKey": "parent value",
"childKey": "overidden value"
}
// /parent/child/wires-defaults.json
{
"@root": true,
"childKey": "child value"
}
// /parent/child/someFile.js
require( "#parentKey" ) === undefined;
require( "#childKey" ) === "child value";
- usual file paths
require( "./lib/myUtil.js" )
require( "fs" )
- tilde-slash-lead, for file paths relative to home directory
require( "~/.jshint.json" )
- chevron-slash-lead, for file paths relative to current working directory
require( ">/logger.js" )
- hash-lead, to recover settings
require( "#username" )
require( "#server.db.host" )
- hash-then-chevron-lead, to recover environment variables
require( "#>PATH" )
require( "#>WIRES_ENV" )
- colon-lead, to inject route-based modules
require( ":model/person.js" )
require( ":cacheFactory" )
- templated, to construct paths dynamically
require( "./lib/{#vendor}/main.js" )
require( "nodeunit/reporters/{#unit.reporter}" )
- two-colons-lead, to bypass
wires
entirelyrequire( "::wires-will-not-transform-this" )
Targeting undefined
or null
values in expressions may yield potentially undesirable "undefined"
or "null"
in the resulting string. If and when you change the leading #
to a leading ?
and the targeted value is "falsy" (undefined
, null
, false
, etc...), then the result is an empty string (""
). For instance:
- the expression
"value={#undefinedValue}"
yields"value=undefined"
- the expression
"value={?undefinedValue}"
yields"value="
This is especially handy for environment variables that may or may not be set. While "#>UNSET_VAR"
would yield "undefined"
, ?>UNSET_VAR
yields ""
.
In your configuration files, every object property which name is not colon-lead is a setting.
A setting can be of any type, including an object. When the value of a setting is a string, it accepts the template syntax seen in the previous section.
// /app/wires.json
{
"number": 56,
"boolean": false,
"string": "some string",
"templateString": "number is {#number}",
"array": [ 1, 2 ],
"object": {
"templateString": "boolean is {#boolean}"
},
"env": "{?>SOME_VAR}"
}
// /app/file.js
require( "#number" ) === 56;
require( "#string" ) === "some string";
require( "#templateString" ) === "number is 56";
require( "#array" ); // [ 1, 2 ]
require( "#object.templateString" ) === "boolean is false";
require( "#env" ) === ( process.env.SOME_VAR || "" );
As of version 2.0
, every object property which name ends with a question mark in your configuration files is a fallback. Fallbacks are useful in situations where a setting may be empty (""
, NaN
, null
or undefined
) yet you still need a default value for it.
Prior to version 4.0
, wires
did consider any falsy value as empty (0
, false
, etc). This was changed in order to accommodate casting as introduced in the very same version.
Let's considering the following situation:
// /app/wires.json
{
"mysql_user": "{?>SQL_USER}"
}
// /app/mysql/wires-defaults.json
{
"mysql_user": "root"
}
// /app/mysql/index.js
require( "#mysql_user" ) === ( process.env.SQL_USER || "" );
The environment variable SQL_USER
may not be set and so the setting mysql_user
may end up as an empty string. Yet, it is still set and the default value defined in wires-defaults.json
will never be used.
The problem can easily be worked around using a fallback:
// /app/wires.json
{
"mysql_user": "{?>SQL_USER}"
}
// /app/mysql/wires-defaults.json
{
"mysql_user?": "root"
}
// /app/mysql/index.js
require( "#mysql_user" ) === ( process.env.SQL_USER || "root" );
Fallbacks act like any other setting and can be overridden using the cascading nature of configurations. They can also be recovered programmatically if and when needed.
Coming back to the previous example, if we redefine the fallback in the parent configuration then this new setting will be used in place of root
as the "fallback value":
// /app/wires.json
{
"mysql_user": "{?>SQL_USER}",
"mysql_user?": "admin"
}
// /app/mysql/wires-defaults.json
{
"mysql_user?": "root"
}
// /app/mysql/index.js
require( "#mysql_user" ) === ( process.env.SQL_USER || "admin" );
require( "#mysql_user?" ) === "admin";
Fallbacks also work within settings that are objects themselves. They just disappear when you require the object in its entirety:
// /app/wires.json
{
"mysql": {
"user": "{?>SQL_USER}"
}
}
// /app/mysql/wires-defaults.json
{
"mysql": {
"user?": "root"
}
}
// /app/mysql/index.js
require( "#mysql.user" ) === ( process.env.SQL_USER || "root" );
require( "#mysql.user?" ) === "root";
assert.deepEqual(
require( "#mysql" ),
{
"user": process.env.SQL_USER || "root"
}
);
As of version 4.0
, it is possible to cast string values into booleans or numbers in configuration files:
// wires.json
{
"bool": "(boolean) true",
"number": "(number) 1204"
}
// file.js
require( "#bool" ) === true;
require( "#number" ) === 1204;
It is especially useful when dealing with environment variables:
// wires.json
{
"size": "(number){?>SIZE}"
}
// file.js
const size = ( process.env.SIZE || "" ).trim();
require( "#size" ) === ( size ? Number( size ) : NaN );
Casting follows the following rules:
- For booleans (cast using
(bool)
or(boolean)
):"true"
becomestrue
,"false"
becomesfalse
,- any other string becomes
null
.
- For numbers (cast using
(num)
or(number)
), any string that cannot be parsed as a number will giveNaN
as a result.
// wires.json
{
"true": "(bool) true",
"false": "(bool) false",
"null": "(bool) not a boolean",
"number": "(num) 1204",
"NaN": "(num) not a number",
}
// file.js
require( "#true" ) === true;
require( "#false" ) === false;
require( "#null" ) === null;
require( "#number" ) === 1204;
isNaN( require( "#NaN" ) );
You can put an arbitrary number of spaces after the opening parenthesis and around the closing parenthesis. So the following expressions are all valid and strictly equivalent:
"(bool)true"
"(bool) true"
"( bool)true"
"(bool )true"
"( bool )true"
"( bool) true"
"(bool ) true"
"( bool ) true"
In your configuration files, every object property which name is colon-lead and does not end with a slash defines a route.
Routes must be strings. Like settings, they do accept the template syntax. The final string must be a path to a file that actually exists.
File paths may refer to a file:
- globally (relying on
NODE_PATH
) - relatively to the directory of the configuration file (start with
"./"
or"../"
) - relatively to the home directory (start with
"~/"
) - relatively to the current working directory (start with
">/"
)
// /app/wires.json
{
":dbRequest": "mysql/request",
":cacheFactory": "../cache/factory",
":data": ">/data.json"
":eslint": "~/.eslintrc.json"
}
// /app/some/path/inside/index.js
require( ":dbRequest" ) === require( "mysql/request" );
require( ":cacheFactory" ) === require( "/cache/factory" );
require( ":data" ) === require( "/working/dir/data.json" );
require( ":eslint" ) === require( "/home/me/.eslintrc.json" );
As of version 2.1.0
, it is possible to set a route to null
. Requiring such a route would result in null
.
// /app/wires.json
{
":not-implemented-yet": null
}
// /app/some/path/inside/index.js
require( ":not-implemented-yet" ) === null;
In your configuration files, every object property which name is colon-lead and ends with a slash defines a generic route.
Like normal routes, they must be strings, they do accept the templated syntax and they have the same semantics (NODE_PATH
, "./"
, "../"
, "~/"
, ">/"
). However, they may not point to the path of an existing file since the final string will be used as a replacement for the property name in require expressions.
// /app/wires.json
{
":dbo/": "./db/model-",
}
// /app/mvc/controllers/mainPage.js
require( ":dbo/client" ) === require( "/app/db/model-client" );
require( ":dbo/car/bmw" ) === require( "/app/db/model-car/bmw" );
As of version 2.1.0
, it is possible to set a generic route to null
. Requiring such a route and its children would result in null
.
// /app/wires.json
{
":to-do/": null,
}
// /app/mvc/controllers/mainPage.js
require( ":to-do/child" ) === null;
require( ":to-do/other/path" ) === null;
As of version 2.1.0
, in your configuration files, every object property which name is colon-lead and ends with a slash followed by an opening then a closing parenthesis defines a computed route.
They are very similar to generic routes except there is no automatic concatenation performed by wires
. Rather, the value must be a string pointing to a module that exports a function. This function will be called by wires
with the path segments as arguments and is expected to return the resulting path.
This sounds quite complicated but let's re-implement the generic route example from the previous section with a computed route:
// /app/wires.json
{
":dbo/()": "./helpers/dbo.js",
}
// /app/helpers/dbo.js
module.exports =
( ...pathSegments ) =>
"../db/model-" + pathSegments.join( `/` );
// /app/mvc/controllers/mainPage.js
require( ":dbo/client" ) === require( "/app/db/model-client" );
require( ":dbo/car/bmw" ) === require( "/app/db/model-car/bmw" );
Please note that paths returned by the function are resolved relatively to the location of the file where the function is defined.
Computed Route functions must be defined in CommonJS modules. ECMAScript Modules are not supported here.
It is possible for a computed route function to return null
: requiring such a route and its children would result in null
.
// /app/wires.json
{
":dbo/()": "./helpers/dbo.js",
}
// /app/helpers/dbo.js
module.exports = () => null; // not implemented yet
// /app/mvc/controllers/mainPage.js
require( ":dbo/client" ) === null;
require( ":dbo/product" ) === null;
As of version 5.0
, wires
overrides both static and dynamic import statements. This makes it compatible with ES Module based projects.
Unlike require
, import()
needs fully defined paths, file extension included, and fails to fetch index.js
when targeted at a directory, so you have to keep that in mind when creating routes.
// /app/wires.json
{
":dir1": "./dir",
":dir2": "./dir/index",
":dir3": "./dir/index.js"
}
// /app/dir/index.js
module.exports = "You Found Me!";
// /app/cjs.js
require( `:dir1` ) === "You Found Me!"
require( `:dir2` ) === "You Found Me!"
require( `:dir3` ) === "You Found Me!"
// /app/esm.js
import message1 from ":dir1"; // fails with exception
import message2 from ":dir2"; // fails with exception
import message3 from ":dir3"; // succeeds!
message3 === "You Found Me!";
Also, for object values, wires
will create a named export for every property which name is a proper JavaScript identifier.
// /app/wires.json
{
"object": {
"property1": 1,
"property2": 2,
"no name export": 3
}
}
// /app/esm.js
import object, { property1, property2 } from "#object";
object.property1 === 1;
object.property2 === 2;
object[ "no name export" ] === 3;
property1 === 1;
property2 === 2;
A primary goal of wires
is to be compatible with as many bundlers as possible. The bundle does not need wires
to run and does not include the core code of the package.
While routes are computed statically, values that depend on environment variables will properly change if the corresponding environment variable changes. wires
will add a few short functions to the bundle to ensure as much.
For instance, if we have the following situation:
// /wires.json
{
"server": {
"keepAlive": "(number) {?>KEEP_ALIVE}",
"keepAlive": 60,
"port": "(number) {?>PORT}",
"port?": 8080
}
}
// /index.js
import { keepAlive, port } from "#server";
console.log( keepAlive, port );
then:
node index.js
will output60 8080
KEEP_ALIVE=30 PORT=80 node index.js
will output30 80
If index.js
is bundled into bundle.js
:
node bundle.js
will output60 8080
KEEP_ALIVE=30000 PORT=80 node bundle.js
will output30 80
As of version 5
, wires
exposes a rollup
plugin at wires/rollup
(it is part of the wires
package itself).
Simply create a rollup.config.js
configuration file and import the plugin:
import wires from "wires/rollup";
export default {
input: "src/index.js",
output: {
dir: "output",
format: "es"
},
plugins: [ wires() ],
};
Then call rollup
either via the CLI or the API.
The plugin is compatible with both @rollup/plugin-commonjs
and @rollup/plugin-node-resolve
which makes it possible to bundle CommonJS projects, provided you set the former's requireReturnsDefault
options to true
and put the latter after the wires
plugin in the list of plugins.
It can also make sense to include @rollup/plugin-json
.
import commonjs from "@rollup/plugin-commonjs";
import json from "@rollup/plugin-json";
import { nodeResolve } from "@rollup/plugin-node-resolve";
import wires from "wires/rollup";
export default {
input: "src/index.js",
output: {
dir: "output",
exports: "auto",
format: "cjs"
},
plugins: [
wires(),
nodeResolve(),
json(),
commonjs( {
requireReturnsDefault: true
} )
]
};
© Julian Aubourg, 2012-2022 – licensed under the MIT license.