Check for io.Writer

[cardil]

Fixes #80

[cardil]

@mattn Can you review this?

[dolmen]

@cardil You have a typo in filenames: writter.go -> writer.go (same for test file).

Please fix the typo and git commit --amend && git push -f.

[dolmen]

Use a type alias to avoid defining a real type:

Suggested change

	type filelike interface {
	type filelike = interface {

[cardil]

Definitely can do it. I'm just wondering why? The type is private, so what's the point?

[dolmen]

Log test result:

Suggested change

	isatty.IsWriterTerminal(os.Stdout)
	t.Log("os.Stdout:", isatty.IsWriterTerminal(os.Stdout))

[dolmen]

Log test result:

Suggested change

	isatty.IsWriterCygwinTerminal(os.Stdout)
	t.Log("os.Stdout:", isatty.IsWriterCygwinTerminal(os.Stdout))

[dolmen]

I think it would be helpful to add the symetric support for io.Reader.

[cardil]

I think it would be helpful to add the symetric support for io.Reader.

I can file another PR for that

[dolmen]

@cardil

1. The filelike interface will be common to both.
2. Both should be added at the same time in the API. io.Writer has no priority over io.Reader. (In fact I personally use go-isatty more to check os.Stdin than os.Stdout)

[cardil]

1. The filelike interface will be common to both.

Yes, maybe. That doesn't mean we need to have that in single PR.

2. Both should be added at the same time in the API. io.Writer has no priority over io.Reader. (In fact I personally use go-isatty more to check os.Stdin than os.Stdout)

There's no reason to add those at the same time? Why? It's a new API, that I'm proposing here, so why would be need to add additional features to this PR. I rather have separate, smaller PRs, that can ve reviewed independently.

[dolmen]

My point is that new API for io.Writer and io.Reader are NOT independent. Consistency matters.

[dolmen]

Anyway, I'm fully against this feature as I explained here: #80 (comment)