strictEqual / deepStrictEqual behave differently on numbers

[rotu]

Chai's behavior of strictEqual and deepStrictEqual differ in how they treat numbers (particularly NaN and -0). It seems like "depth" should not affect the equality of two shallow objects.

chai.assert.strictEqual(NaN,NaN) // fails
chai.assert.deepStrictEqual(NaN,NaN) // passes

chai.assert.strictEqual(-0,0) // passes
chai.assert.deepStrictEqual(-0,0) // fails

Note that the documentation adds confusion, since it says that "deep equality" refers to deep-eql semantics and "strict equality" refers to ===:

chai/lib/chai/core/assertions.js

Lines 1716 to 1718 in 1a8247f

	* By default, strict (`===`) equality is used. Add `.deep` earlier in the
	* chain to use deep equality instead. See the `deep-eql` project page for
	* info on the deep equality algorithm: https://github.com/chaijs/deep-eql.

It appears deepStrictEqual was originally added as an alias for strictEqual to match Node: #653

[koddsson]

Damn this is weird. I feel like chai.assert.strictEqual(NaN,NaN) and should not fail and chai.assert.strictEqual(-0,0) should not pass.

I'll dig into this.

cc/ @keithamus as someone that might know if this behaviour is actually correct.

[koddsson]

I guess NaN !== NaN so it does make some sort of sense.

[koddsson]

Also -0 === 0 now that I think about it.

[koddsson]

Then reading the deep-eql README, it says:

• Strict equality for non-traversable nodes according to Object.is:
  . - eql(NaN, NaN).should.be.true;
  . - eql(-0, +0).should.be.false;

[koddsson]

Makes me think this is indeed working as expected but maybe a bit confusing.

[rotu]

Makes me think this is indeed working as expected but maybe a bit confusing.

I don't know what you mean by "as expected". The discrepancy was not the result of a typo, but it's also not a reasonable design choice.

I think I would expect chai to change strictEqual as node:assert did in v10. See relevant Node PR.