dvc repro: change output format?

[skshetry]

Recently we made a change in the dvc repro output by removing the indent to > in front of the command scripts.

My reasoning for this is that when indenting the script, the output of the script starts from the leftmost part of the next line which makes it look ugly and congested.

Adding a ">" at least separates the DVC's output with the script's.

Before:

$ dvc repro
Running stage 'copy':
	echo "foo"
foo
	cp foo bar
	echo "bar"
bar
Generating lock file 'dvc.lock'
Updating lock file 'dvc.lock'

To track the changes with git, run:

	git add dvc.lock
Use `dvc push` to send your updates to remote storage.

Note that multiline commands weren't supported before.

After:

$ dvc repro
Running stage 'copy':
> echo "foo"
foo
> cp foo bar
> echo "bar"
bar
Generating lock file 'dvc.lock'
Updating lock file 'dvc.lock'

To track the changes with git, run:

	git add dvc.lock
Use `dvc push` to send your updates to remote storage.

This brings me to the main point of this discussion.

1. Is it necessary to print the output during repro? Should that be hidden by default (and, show whenever an error occurs or via some flag)?
2. If we should print them by default, should we introduce some UI widget to separate the outputs?
   Like a pytest's divider/header or should we keep it to the minimum >/indent?

Example 1

============== Running 'cp foo bar' ==============
foo
bar

Example 2

JK :)

   ___                  ___
  (o o)    Running     (o o)
 (  V  )    'copy'    (  V  )
/--m-m- -------------/--m-m- \
|                            |
|         echo "foo"         |
|         cp foo bar         |
|         echo "bar"         |
|____________________________|
foo
bar

@iterative/engineering

[jorgeorpinel]

OMG yes

   ___                  ___
  (o o)    Running     (o o)
 (  V  )    'copy'    (  V  )
/--m-m- -------------/--m-m- \
|                            |
|         echo "foo"         |
|         cp foo bar         |
|         echo "bar"         |
|____________________________|

[pared]

I would implement that for an easter egg, really, this is magnificent, @skshetry.

[jorgeorpinel]

Oh dear... This thing has threads 🙀

[jorgeorpinel]

Running stage 'copy':

(As discussed in iterative/dvc.org#1980 (comment)) I think we should still say "with command(s)" though, in case users don;t realize that stages are comprised of terminal commands.

[jorgeorpinel]

And actually, on the > matter...

1. As @shcheklein mentioned somewhere, it makes it harder to copy/paste that line as a working command
2. A > character at the beginning of a command line is actually a valid command, which redirects nothing to a file name provided next, e.g.

$ > echo hi
bash: hi: command not found
$ ls
echo

So I'm back to being of the opinion that some sort of indentation is better for this. My only idea is tabs (as we use in other commands) but each command output would need to be padded (with 2 tabs) also.

[pared]

So, maybe we should go with Example#1 or #2? When I am doing anything in DVC - (as a user) and running repro, I usually look for output of particular stage. To do that I need to find the execution command, and marking it as in Examples makes it easily distiguishable.

[jorgeorpinel]

Yes as long as the commands are by themselves in a terminal line (easy to copy/paste) any format around them I don't see any problem with.

[pared]

1. Is it necessary to print the output during repro? Should that be hidden by default (and, show whenever an error occurs or via some flag)?

I would say we should not hide output by default. It would make sense in most cases, but in the most important one - long-lasting training, one usually needs to see the progress. But optional turn off for stages outputs sounds reasonable.