`debug.enable()` flushes enabled namespaces

[dbo]

Calling enable is disruptive to any enabled namespaces, which has changed in the recent minor version, caused by #409

$ DEBUG=foo node -e 'var dbg = require("debug"); dbg.enable("bar"); console.log(dbg.enabled("foo"))'
=> false

[geonanorch]

The new implementation allows full control (i.e. override of the original DEBUG settings), which seems desireable (even more in the context of changing this dynamically later on, see issue #433 ).
I submit that the main problem here is that public method enable() is not documented.
Actually debug lacks an API reference document – does obviously not prevent it from being appreciated and widely used, but makes integration with other tools more difficult...

[Qix-]

I disagree with @geonanorch slightly - I feel like .enable() should be able to take -* and .disable() should be able to take * in order to clear the enabled namespaces.

In fact, the default behavior if no namespaces are passed (e.g. literally debug.enable() and debug.disable() alike) should be to implicitly pass *, which would enable/disable all namespaces, respectively.

[geonanorch]

@Qix- I don't think that we disagree, your proposal for a reset makes plenty of sense!
Personally I prefer the 'no parameter provided' approach to a wildcard, i.e. .enable() rather than .enable(*), but that is cosmetic really.

[Qix-]

Well realistically both should be supported, I think.

[mblarsen]

It seems this issue is a wont-fix? @Qix- @geonanorch ?

In other words: the bug is a feature :)

The implicit and explicit * and -* seems to be a different feature/issue.

[Qix-]

@mblarsen Definitely not a wontfix, it's just a breaking change and requires a lot of work.

[mblarsen]

Related #451

[Qix-]

Yeah, same thing really. The fact that .enable() overwrites everything is what trips people up, methinks. Hence why I said debug.disable('*'); debug.enable(...) should be how users would migrate, and that calls to .enable() and .disable() should instead modify the existing namespaces.

In fact, I'd almost say that we get rid of .disable() as you merely need to negate any patterns passed to .enable() to get the exact same functionality in that case.

[mblarsen]

.disable() is nice for convenience and code readability though.