New cli

[hildjj]

Split into several chunks:

• Compile the existing CLI with rollup
• Add tests to the exisitng CLI
• Move the CLI to commander, updating tests
• Add -t/-T
• Fix Add -c / --config CLI option #161 by adding support for .js/.cjs config files and being careful about the command line overriding the config file, but the config file overriding defaults.

[hildjj]

This needs some testing on windows before it's ready.

[Mingun]

In general very good, but test file should be read after the parser generation and changing output type from source to parser can lead to some unintended results. I think, that this should be fixed.

The other stuff are minor but it will be great if you look at them.

[Mingun]

This two changes fixes the Windows build

[hildjj]

Sorry it's taken so long, but I'm finally addressing all of these comments.

@Mingun, you said "changing output type from source to parser can lead to some unintended results". Can you say more about that? My thought was that if you're supplying a test, we should get a parser back instead of source, so that we can call that generated code.

[hildjj]

I would expect that --test is fully additive, so parser is still generated and written to the --output

I can see this point of view. There have been a few times where I've wanted to save the output of the test, but I suppose I can always just capture stdout or add another option later if it becomes important.

[hildjj]

After thinking about it some more, I don't want to add variadic options because it would break backward-compatibility for some uses of the command line.

[Mingun]

After thinking about it some more, I don't want to add variadic options because it would break backward-compatibility for some uses of the command line.

Yes, I also thinking about this after you write this . So may be later.

I think it is almost done, but I don't like the way how the options combined from the config file, command line JSON and command line parameters -- it is too sophisticated and non-obvious. I suggest to change handling, as I did in hildjj/peggy@new-cli...Mingun:new-cli (and also fix couple of other issues).

Also, If I understood correctly, do you agree with me that it is worth reworking the way the --test is handled?

[hildjj]

I have reworked how --test behaves, as discussed above. I'm just fixing up the tests now before pushing the update.

I agree that the config file stuff is a little subtle, but I think it was always that way. Can you explain a little more about how you changed it? I'm not seeing how yours is different from mine in the link above.

[Mingun]

Conceptually I did the following:

1. I introduce DEFAULTS const with all options that have a default value
2. Then I save parsed config from the commander to the cliOptions.
3. Then remove from cliOptions all keys that have default values (as defined in the DEFAULTS)
4. Then build new options object as

Object.assign({}, DEFAULTS, extraOptionsFile, extraOptions, cliOptions)

To clarify: I just made forming options a little more straightforward (at least it looks so for me)

[hildjj]

That makes sense. I'll work in something like that.

[hildjj]

As a compromise on variadic options, and to make everything consistent, you can now specify a comma-separated list for --dependency and --plugin

[Mingun]

As for dancing around config, I've just create an issue in the commander, so, if it will be implemented, we could switch to the built-in solution: tj/commander.js#1584

[hildjj]

I think I have all of the requested changes in now. The defaults handling is better than what I had before, but wasn't able to be as clean as you had, because I wanted arrays and objects in the options to be additive instead of replacing what the earlier option was.

[Mingun]

Ok, it seems good. LGTM