WIP - Add ability to merge output

[JuanCaicedo]

I started work on addressing https://github.com/kimmobrunfeldt/concurrently/issues/75. This change makes this almost possible but it has a bug that I'll detail. instructions. I've made a repo that imports my changes and allows you test them.

Changes

• Adds a -g or --group flag to enable output aggregation
• No breaking changes

Working scenario

• If the commands finish in succession (first command finishes first, then second, then third). This works great.

Bugs

• However, if a previous command runs slower than one of the commands that follow it, the output of the latter command(s) are swallowed up.

Could you use some guidance on

• How to tackle the bug above
• How to add tests
• Code improvements

TODO

• Update readme to list group option
• Combine stderr in the same way

[JuanCaicedo]

Though the config is available as a global, I chose to pass it as a parameter to this new function. I prefer this style, but I'm open to changing it

[gustavohenke]

It's okay :)

[JuanCaicedo]

Tests are failing because of node v0.12, mentioned in https://github.com/kimmobrunfeldt/concurrently/issues/76

[JuanCaicedo]

Does the bug maybe have to do with the fact that merge is also used on line 280? If I try changing that use concat though, all output of later commands gets swallowed

[JuanCaicedo]

My current thought right now is that this has to do with RX not handling back pressure out of the box. Here is a link with more information

[gustavohenke]

Hi @JuanCaicedo! Thanks for this PR. As you may have noted, I'm getting started here as a maintainer :)

The node v0.12 build problem was fixed. If you push something new, the status will be green.

Regarding the tests for this feature... if you could emulate the example given at #75, I guess it would be enough.

And what do you mean by the output being "swallowed"? Does this means it will never show up?

[JuanCaicedo]

@gustavohenke thanks for your work, excited to see this project moving forward! Yes, that's what I mean

[gustavohenke]

Okay, so I might have found a way to fix that problem.

A process with index i needs to buffer until all processes from index 0 to i (inclusive) close.
This means:

• Process 0 buffers until 0 closes;
• Process 1 buffers until 0 and 1 close;
• Process 2 buffers until 0, 1 and 2 close;
• ...and so on.

Possible with some wise usage of buffer operator together with the close streams.

[gustavohenke]

Of course, there may be easier/simpler ways. I'm not a heavy user of Rx...

[JuanCaicedo]

@gustavohenke I have a some more time over these next few weeks to work on this, I'll take another stab at it 😃

[alerkesi]

Guys, which status for this PR? How can I help something to push this faster?

[JuanCaicedo]

@alerkesi Do you want to prototype an approach like @gustavohenke mentioned above? Or if you can think of a different way to do it, that would be good too (I'm also not too experienced with Rx)