fix: deep assign should ignore inseparable object

[haoxins]

No description provided.

[haoxins]

also fix failed test: support Object.create(null) targets

[schnittstabil]

Thank you @coderhaoxin, but that's not how Object.assign works, e.g:

// this PR:
> deepAssign({regexp: /reg/}, {regexp: 'string'});
{regexp: 'string'}

// but:
> Object.assign(/reg/, 'string')
{ /reg/ '0': 's', '1': 't', '2': 'r', '3': 'i', '4': 'n', '5': 'g' }

[schnittstabil]

Nothing is wrong with this test. The first deep-assign argument is the target and this test asserts, that the following does not throw an Error:

deep-assign(Object.create(null), …)

[schnittstabil]

@coderhaoxin Thank you for mentioning that. You were right, this test had to be repaired (e252fa2).

[haoxins]

@schnittstabil Oh, I miss something. Thx

[haoxins]

@schnittstabil Object.assign() convert all the arguments to Object, but should deep-assign the same?

I mean, is it reasonable that

deepAssign({regexp: /reg/}, {regexp: 'string'});

should equal to

Object.assign(/reg/, 'string')

for the property regexp?

[schnittstabil]

The project description of deep-assign says:

Recursive Object.assign()

@coderhaoxin Therefore I would say yes, in my opinion it is the canonical way.

But I must admit that the behavior is not very obvious, which is shown by #3, #4, #6. I am open to new interpretations of recursive and assign, but deep-assign should behave like Object.assign in the base case…

[schnittstabil]

@coderhaoxin merge-options might be of interest to you.