Runtime (recursive) type-checking for JavaScript.
$ npm install rttc --save
Want to coerce a value to match a particular type?
var rttc = require('rttc');
rttc.coerce({ firstName: 'string'}, {firstName: 13375055});
// => { firstName: "13375055" }
rttc.coerce({ firstName: 'string'}, {something: 'totally incorrect'});
// => { firstName: "" }
// (when confronted with something totally weird, `.coerce()` returns the "base value" for the type)
Want to throw an Error if a value doesn't match a particular type?
rttc.validateStrict({ firstName: 'string'}, {firstName: 13375055});
// throws error
// (`.validateStrict()` demands a value that is precisely the correct type)
rttc.validateStrict({ firstName: 'string'}, {firstName: '13375055'});
// does not throw
Or if you want to be a little more forgiving:
rttc.validate({ firstName: 'string'}, {something: 'totally incorrect'});
// throws error
rttc.validate({ firstName: 'string'}, {firstName: 45});
// => "45"
// (when confronted with minor differences, `.validate()` coerces as needed to make stuff fit)
Not sure how to build a type schema for use with .coerce() or .validate()? Use .infer() to build one from an example value you've got laying around:
rttc.infer({ firstName: 'Rosella', lastName: 'Graham', friends: ['Valencia', 'Edgar', 'Attis'] });
// => { firstName: 'string', lastName: 'string', friends: ['string'] }
This special example value is called an exemplar.
Note that when inferring the schema for an array in an exemplar, the array item is considered the pattern. It is used to indicate the expected type of each item in the array. Consequently, if an array in an exemplar has an item, that means it is homogeneous (i.e. all its items have the same schema).
You can use exemplars to specify just about any type-- here's a more advanced schema to show what I mean:
rttc.infer([{ upstream: '===', fieldName: 'photos', files: [{getFile: '->', fileName: 'whatever', numBytes: 34353, meta: '*' }] }]);
// =>
// [
// {
// upstream: 'ref',
// fieldName: 'string',
// files: [
// {
// getFile: 'lamda',
// fileName: 'string',
// numBytes: 'number',
// meta: 'json'
// }
// ]
// }
// ]
Note that all of the validation and coercion strategies used in this modules are recursive through the keys of plain old JavaScript objects and the indices of arrays.
Each type can be validated or coerced against. If coercion fails, the "base value" for the type will be used.
Also note that all types below may be expressed recursively within faceted dictionaries and patterned arrays. If those words don't make sense, keep reading, you'll see what I mean.
There are 10 different types recognized by rttc:
| type | rttc example notation | base value |
|---|---|---|
| string | 'any string like this' | '' |
| number | 1337 (any number) | 0 |
| boolean | false (or true) | false |
| lamda | '->' | function () { throw new Error('Not implemented! (this function was automatically created by rttc'); }; |
| generic dictionary | {} | {} (empty dictionary) |
| generic array | [] | [] (empty array) |
| json | '*' | null |
| ref | '===' | undefined |
| faceted dictionary (recursive) | {...} (i.e. w/ keys) | {...} (w/ all expected keys and their base values) |
| pattern array (recursive) | [...] (i.e. w/ 1 item) | [] (empty array) |
example: 'stuff'
The string type accepts any string.
example: 323
The number type accepts numbers like 0, -4, or 235.3. Anathemas like Infinity, -Infinity, NaN, and -0 are all coerced to zero.
example: false
The boolean type accepts true or false.
example: ->
The lamda type accepts any function.
example: {}
The generic dictionary type accepts any JSON-serializable dictionary.
Dictionaries that have been validated/coerced against the generic dictionary type:
Error instances get stringified into empty objects. Instead, rttc turns them into human-readable strings by reducing them to their .stack property (this includes the error message and the stack trace w/ line numbers)RegExp instances get stringified into empty objects. Instead, rttc turns them into human-readable strings like '/some regexp/gi'function() instances get stringified into empty objects. Instead, rttc turns them into human-readable strings like 'function doStuff (a,b) { console.log(\'wow I can actually read this!\'); }'example: {...}
The faceted dictionary type is any dictionary type schema with at least one key. Extra keys in the actual value that are not in the type schema will be stripped out. Missing keys will cause .validate() to throw.
Dictionary type schemas (i.e. plain old JavaScript objects nested like {a:{}}) can be infinitely nested. Type validation and coercion will proceed through the nested objects recursively.
{
id: 'number',
name: 'string',
isAdmin: 'boolean',
mom: {
id: 'number',
name: 'string',
occupation: {
title: 'string',
workplace: 'string'
}
}
}
example: []
Arrays that have been validated/coerced against the generic array type:
Error instances get stringified into empty objects. Instead, rttc turns them into human-readable strings by reducing them to their .stack property (this includes the error message and the stack trace w/ line numbers)RegExp instances get stringified into empty objects. Instead, rttc turns them into human-readable strings like '/some regexp/gi'function() instances get stringified into empty objects. Instead, rttc turns them into human-readable strings like 'function doStuff (a,b) { console.log(\'wow I can actually read this!\'); }'example: ['Margaret']
example: [123]
example: [true]
example: [[...]]
example: [{...}]
Array type schemas may be infinitely nested and combined with dictionaries or any other types.
Runtime arrays being validated/coerced against array type schemas will be homogeneous (meaning every item in the array will have the same type).
Undefined items will always be stripped out of arrays.
Also note that, because of this, when providing a type schema or type-inference-able example for an array, you only need to provide one item in the array, e.g.:
[
{
id: 'number',
name: 'string',
email: 'string',
age: 'number',
isAdmin: 'boolean',
favoriteColors: ['string'],
friends: [
{
id: 'number',
name: 'string'
}
]
}
]
example: '*'
This works pretty much like the generic array or generic dictionary type, with one major difference: the top-level value can be a string, boolean, number, dictionary, array, or null value.
Other than the aforementioned exception for null, the generic JSON type follows the JSON-serializability rules from generic arrays and generic dictionaries.
example: '==='
This special type allows anything except undefined at the top level (undefined is permitted at any other level). It also does not rebuild objects, which means it maintains the original reference (i.e. is ===). It does not guarantee JSON-serializability.
The following is a high-level overview of important conventions used by the rttc module. For detailed coverage of every permutation of validation and coercion, check out the declarative tests in the spec/ folder of this repository.
undefined and null valuesundefined is never valid as a top-level value against ANY type, even mutable reference (===)undefined IS, however, allowed as an item in a nested array or value in a nested dictionary, but only against the mutable reference type (===)null is only valid at the top level against the JSON (*) and mutable reference (===) types.NaN is only valid against the mutable reference type ('===')Infinity and -Infinity is only valid against the mutable reference type ('===')Infinity and -Infinity are only valid against example: '==='+0 and -0 are always coerced to 0 (except against the mutable reference type)When coerced against the generic dictionary, generic array, or the generic json types, the following is true:
Error instances are coerced to the string value of their .stack property (i.e. the message + stack trace you're used to seeing in the terminal)Date instances are coerced to the string value of running their .toJSON() method (a ISO-8601 timestamp, e.g. '2015-05-24T15:16:48.999Z'. This reflects the Date in GMT/UTC time, so is therefore timezone-agnostic).RegExp instances are coerced to the string value you get from running their .toString() method (e.g. '/foo/' or '/^bar/gi').toString() method (e.g. 'function someFunction (some,args,like,this,maybe){ /* and some kind of implementation in here prbly */ }')Stream and Buffer instances (from Node.js) are only valid against the mutable reference type.null against the generic dictionary, generic array, or the generic json types.As mentioned above, every type has a base value.
""0false'->'), base value is a function that uses the standard machine fn signature and triggers its "error" callback w/ a message about being the rttc default (e.g. function(inputs,exits,env) { return exits.error(new Error('not implemented')); }){}) or a faceted dictionary type (e.g. {foo:'bar'}), the base value is {}.[]), or a faceted/homogenous array type (e.g. [3] or [{age:48,name: 'Nico'}]), the base value is []'*'), base value is null.'==='), base value is undefined.Note that, for both arrays and dictionaries, any keys in the schema will get the base value for their type (and their keys for their type, etc. -- recursive)
This package exposes a number of different utility methods. If you're interested in using any of these directly, we highly recommend you consider looking at machinepack-rttc, which provides a higher-level abstraction with better documentation.
The low-level reference below assumes you are willing/able to dig into the source code of this module for more information. So continue at your own risk!
Throws if the provided value is not the right type (recursive).
Either returns a (potentially "lightly" coerced) version of the value that was accepted, or it throws. The "lightly" coerced value turns "3" into 3, "true" into true, -4.5 into "-4.5", etc.
ALWAYS returns an acceptable version of the value, even if it has to mangle it to get there (i.e. by using the "base value" for the expected type. More on that below.)
undefined])This function will use the provided typeSchema to figure out where "lamda" values (functions) are expected, then will use eval() to bring them back to life. Use with care.
false], [dontStringifyFunctions=false])This takes care of a few serialization edge-cases, such as:
.stack property), and functions (unless dontStringifyFunctions is set)[Circular])-Infinity, Infinity, and NaN with 0undefined or null values. If allowNull is set to true, null values will not be stripped from the encoded string.undefined], [unsafeMode=false])Parse a stringified value back into a usable value.
This is basically just a variation on JSON.parse that calls rttc.hydrate() first if unsafeMode is enabled.
false])Encode a value into a string.
This is basically just a variation on JSON.stringify that calls rttc.dehydrate() first.
undefined], [unsafeMode=false])Parse a string from a human into something usable. If provided, typeSchema will be used to make a better guess. If unsafeMode is enabled, lamda functions will be hydrated.
The inverse of .parseHuman(), this function encodes a string that, if run through .parseHuman() would result in the given value.
undefined])Determine whether two values are equivalent using _.isEqual(), but also look for expected lamda values in the optional type schema and call toString() on functions before comparing them.
This is the method used by
rttc's own tests to validate that expected values and actual values match.
Guess the type schema from an example value.
Determine whether the given type schema is "specific". String, number, boolean, lamda, faceted dictionary, or patterned array types are "specific". Everything else is "generic".
If the second argument (recursive) is set to true, then also recursively check the subkeys of faceted dictionaries and patterns of arrays in the type schema.
If the third argument (isExemplar) is set to true, then treat the provided schema as an rttc example rather than a type schema.
For reference
| type | is specific? |
|---|---|
| string | yes (always) |
| number | yes (always) |
| boolean | yes (always) |
| lamda | yes (always) |
{} (generic) | no |
[] (generic) | no |
{...} (faceted) | yes (maybe recursively) |
[...] (patterned) | yes (maybe recursively) |
| json | no |
| ref | no |
Given a type schema, return an array of up to n unique sample values that would validate against it (in random order). n defaults to 2 if left undefined.
Given a value, return its type as a human-readable string (this is not limited to rttc types-- it can return strings like "Error" and "Date").
If special rttc exemplar syntax is used, it is respected.
Given a value, return a human-readable string which represents it. This string is equivalent to a JavaScript code snippet which would accurately represent the value in code.
This is a lot like util.inspect(val, false, null), but it also has special handling for Errors, Dates, and RegExps (using dehydrate() with allowNull enabled), as well as for Functions (making them eval()-ready.) The biggest difference is that everything you get from rttc.compile() is ready for use as values in *, {}, or [] type machines, Treeline, Angular's rendering engine, and JavaScript code in general (i.e. if you were to append it on the right-hand side of var x = , or if you ran eval() on it)
Note that undefined values in arrays and undefined values of keys in dictionaries will be stripped out, and circular references will be handled as they are in util.inspect(val, false, null)
Useful for:
Here's a table listing notable differences between util.inspect() and rttc.compile() for reference:
| value | util.inspect() | rttc.compile() |
|---|---|---|
| a function | [Function: foo] | function foo (){} |
| a Date | Tue May 26 2015 20:05:37 GMT-0500 (CDT) | '2015-05-27T01:06:37.072Z' |
| a RegExp | /foo/gi | '/foo/gi/' |
| an Error | [Error] | 'Error\n at repl:1:24\n...' |
| a deeply nested thing | { a: { b: { c: [Object] } } } | { a: { b: { c: { d: {} } } } } |
| a circular thing | { y: { z: [Circular] } } | { y: { z: '[Circular ~]' } } |
| undefined | undefined | null |
| [undefined] | [undefined] | [] |
| {foo: undefined} | {foo: undefined} | {} |
| Infinity | Infinity | 0 |
| -Infinity | -Infinity | 0 |
| NaN | NaN | 0 |
| Readable (Node stream) | { _readableState: { highWaterMar..}} | null |
| Buffer (Node bytestring) | <Buffer 61 62 63> | null |
The following functions are newly implemented, experimental, and tend to be a bit more advanced. They may undergo frequent changes over the coming months, so use with care. You have been warned!
Given a type schema, return an exemplar which accepts precisely the same set of values.
Convert a normal value into an exemplar representative of the most specific type schema which would accept it. In most cases, this leaves the value untouched-- however it does take care of a few special cases:
{}). The most specific exemplar which can accept an empty dictionary is the generic dictionary.[]). Since we don't know the contents, we have to assume this array could be heterogeneous (i.e. have items with different types).null becomes '*'.undefined, it becomes '==='.NaN, Infinity, -Infinity, and -0 become 0.undefined values are stripped.dehydrate internally.If the allowSpecialSyntax flag is enabled, then *, ->, and === will take on their traditional meaning instead of being replaced with strings (e.g. "a star symbol").
rttc.coerceExemplar([{a:null}, {b: [[74,39,'surprise string!']] }])
// => [ {} ]
rttc.coerceExemplar([74,39,'surprise string!'])
// => [ 'surprise string!' ]
rttc.coerceExemplar({x:'*'})
// => { x: 'a star symbol' }
rttc.coerceExemplar({x:'*'}, true)
// => { x: '*' }
Return true if the provided value is NOT a valid rttc exemplar.
Given an exemplar schema and a keypath, return information about the specified segment. If the path is inside of a generic, then the exemplar is '*', and this path is optional. If the path is inside of a ref, then the exemplar is '===', and this path is optional. If the path is not reachable (i.e. inside of a string, or lamda... or something) then throw an error.
var SOME_EXEMPLAR = {
salutation: 'Mr.',
hobbies: ['knitting'],
medicalInfo: {
numYearsBlueberryAbuse: 12.5,
latestBloodWork: {}
}
};
rttc.getPathInfo(SOME_EXEMPLAR, 'hobbies.238');
// =>
// {
// exemplar: 'knitting',
// optional: false
// }
rttc.getPathInfo(SOME_EXEMPLAR, 'medicalInfo.latestBloodWork.whiteBloodCellCount');
// =>
// {
// exemplar: '*',
// optional: true
// }
Given two rttc schemas (e.g. A and B), return the most specific schema that would accept the superset of what both schemas accept normally (A ∪ B).
validateStrict())Given two rttc schemas, return the most specific schema that accepts the shared subset of values accepted by both. Formally, this subset is the intersection of A and B (A ∩ B), where A is the set of values accepted by schema0 and B is the set of values accepted by schema1. If A ∩ B is the empty set, then this function will return null. Otherwise it will return the schema that precisely accepts A ∩ B.
validateStrict())Given a type schema, strip out generics ("ref", "json", {}, and []) to convert it into a "specific" type. In other words, the result of this function always passes rttc.isSpecific().
A convenience method to return the base value for the given exemplar.
A convenience method that calls rttc.infer() on the provided exemplar to get the type schema, then uses it to rttc.coerce() the actualValue provided.
MIT
© 2014 Mike McNeil, Cody Stoltman; © 2015 The Treeline Company
FAQs
Runtime type-checking for JavaScript.
The npm package rttc receives a total of 32,137 weekly downloads. As such, rttc popularity was classified as popular.
We found that rttc demonstrated a not healthy version release cadence and project activity because the last version was released a year ago. It has 2 open source maintainers collaborating on the project.
Did you know?
Socket for GitHub automatically highlights issues in each pull request and monitors the health of all your open source dependencies. Discover the contents of your packages and block harmful activity before you install or update your dependencies.
Security News
Socket's threat research team has detected five malicious npm packages targeting Roblox developers, deploying malware to steal credentials and personal data.
Security News
vlt introduced its new package manager and a serverless registry this week, innovating in a space where npm has stagnated.
Security News
Research
The Socket Research Team uncovered a malicious Python package typosquatting the popular 'fabric' SSH library, silently exfiltrating AWS credentials from unsuspecting developers.