args: default negatable to false

[DartBot]

Issue by seaneagan Originally opened as dart-lang/sdk#16022

---

I can't prove this, but I feel like non-negatable is more common.

For example, --help, --verbose, --quiet, --version are all non-negatable. Also, since flags should generally default to false, specifying --no-foo is redundant. I would prefer to explicitly ask for this feature when I need it, but I'm not sure when that would be.

[DartBot]

Comment by iposva-google

---

In my opinion why you want negatable boolean options is so that you can easily experiment on the command line without having to remove the flag by just adding the "no-" to form a --no-xyz. It is especially useful if you want to call out that a flag is off when comparing two outcomes.

Also the premise that the default is set to false in general is not entirely correct. An example: The --inline flag should probably default to true in a compiler, with the option to turn off inlining if needed. A false default would mean that you would have the awkwardly named flag --dont_inline and the even more strange use within the code.

---

Added Area-Pkg, Library-Args, Triaged labels.

[DartBot]

Comment by seaneagan

---

Thanks, I wasn't aware of the use case to quickly switch between --foo and --no-foo. And I do see how explicitly specifying what is already the default can be helpful.

My main gripe though is that since I get a negated version by default, it's easy to introduce non-sensical negated options, like --no-quiet or --no-version, but then to remove them later is a potentially breaking change. For flags that default to true, you could have only the negated version, e.g. --no-inline, which makes it much more obvious that inlining is the default behavior.

[DartBot]

Comment by seaneagan

---

FWIW, I looked up chrome's command line flags (since it has a ton of them):

http://peter.sh/experiments/chromium-command-line-switches/

and they seem to follow the practice of only including --no-foo, instead of both, e.g. --[no-]foo

[DartBot]

Comment by seaneagan

---

Also forgot to mention that multiple flags are also often used in an enum type way, where only one of the flags should ever be specified at a time, for example see:

https://github.com/dart-lang/spark/blob/master/ide/tool/test_runner.dart#L129

.. and these type of flags are never negatable.

[DartBot]

Comment by anders-sandholm

---

Removed Library-Args label. Added Pkg-Args label.

[DartBot]

Comment by munificent

---

Yeah, I agree with you Sean. I thought defaulting to allow negation would make things more flexible, but in practice it feels like the wrong default.

Alas, it would be a breaking change. I'll probably do it at some point, but I'm going to sit on it for a while.

[Hixie]

Another reason to have negatable flags is if the tool supports saving defaults (e.g. in a .xxxrc file). If you save a positive value, you need a way to negate it on the command line.

For this reason, --verbose and --quiet should indeed be negatable.

I think the real solution here is to make addCommand support taking commands that start with --. --version is really a weird special case, in arg's convention it really should be a command. Similarly with --help, in a way.

Though if --version and --help are the only special cases, then having them require negatable:false seems fine to me.

[nex3]

I think the real solution here is to make addCommand support taking commands that start with --.

Can you expand on this? I'm not 100% sure what you mean.

[Hixie]

When you say "foo --version", you're really saying "foo version", that is, you're trying to run the "version" subcommand of "foo".