n1k0 / node-deeper

slightly better structural equality testing

deeper

deeper is a library for structurally comparing the equality of JavaScript values. It supports recursive / cyclical data structures, is written to avoid try / catch / throw (for speed), and has no dependencies by default.

If you're running Node 0.12+ or io.js, deeper will use the built-in Buffer.equals(). If you're running an older version of Node and you install Ben Noordhuis's buffertools into a project using deeper, it will use that to speed up comparison of Buffers. This used to be installed as an optional dependency, but it gets in the way of browserification and also makes using deeper in your own projects harder, so I changed it to just try to use it if it's there.

It has some optimizations, but stresses correctness over raw speed (unless you're testing objects with lots of Buffers attached to them, in which case it plus buffertools is likely to be the fastest general-purpose deep-comparison tool available).

The core algorithm is based on those used by Node's assertion library and the implementation of cycle detection in isEqual in Underscore.js.

I like to think the documentation is pretty OK.

installation

npm install deeper

usage

// vanilla
var deepEqual = require('deeper')

if (!deepEqual(obj1, obj2)) console.log("yay! diversity!");

details

Copied from the source, here are the details of deeper's algorithm:

1. === only tests objects and functions by reference. null is an object. Any pairs of identical entities failing this test are therefore objects (including null), which need to be recursed into and compared attribute by attribute.
2. Since the only entities to get to this test must be objects, if a or b is not an object, they're clearly not the same. All unfiltered a and b getting past this are objects (including null).
3. null is an object, but null === null. All unfiltered a and b are non-null Objects.
4. Buffers need to be special-cased because they live partially on the wrong side of the C++ / JavaScript barrier. Still, calling this on structures that can contain Buffers is a bad idea, because they can contain multiple megabytes of data and comparing them byte-by-byte is hella expensive.
5. It's much faster to compare dates by numeric value (.getTime()) than by lexical value.
6. Compare RegExps by their components, not the objects themselves.
7. Treat argumens objects like arrays. The parts of an arguments list most people care about are the arguments themselves, not callee, which you shouldn't be looking at anyway.
8. Objects are more complex:
   1. Ensure that a and b are on the same constructor chain.
   2. Ensure that a and b have the same number of own properties (which is what Object.keys() returns).
   3. Ensure that cyclical references don't blow up the stack.
   4. Ensure that all the key names match (faster).
   5. Ensure that all of the associated values match, recursively (slower).

(somewhat untested) assumptions:

• Functions are only considered identical if they unify to the same reference. To anything else is to invite the wrath of the halting problem.
• V8 is smart enough to optimize treating an Array like any other kind of object.
• Users of this function are cool with mutually recursive data structures that are otherwise identical being treated as the same.

license

BSD. Go nuts.