cider-test: don't render a newline between expected and actual

[vemv]

Before:

After:

Personally I just didn't like this newline. For large reports it can make quite a difference.

Cheers - V

[bbatsov]

For short results it looks better without the newline, but I wondering if it helps readability for multi-line results. Let me think a bit about this.

[vemv]

Now that diff is conditionally present, maybe our heuristic can be:

• if there's a diff, consider that the expected/values are complex enough, and therefore deserve a newline between them
• else, no newline

[vemv]

Now the newline is inserted conditionally, following my previous comment.

[aisamu]

NB: Commenting because bbatsov asked on slack

I also found no whitespace to look better for short results, as it's easier to eyeball differences that way.

Perhaps diff would benefit from always having whitespace (between itself and actual, that is): the diff is self-contained and I don't see myself comparing it back-and-forth against actual. It's also large and makes the usual expected vs actual back-and-forth a bit harder/noisier than the no-diff case.

Would it be possible to post a screenshot of a multiline output as well, @vemv?

[vemv]

Good input, much appreciated!

So after that, the logic is simple and visually consistent:

• if there's a diff, there are newlines between expected/actual/diff
• if there's not a diff, there are no newlines between expected/actual

A screenshot showing both paths:

[aisamu]

That indeed looks nicer!

Just to clarify what I originally meant: we could also always have whitespace between actual and diff, and, orthogonally, only add a newline between expected and actual when they're multi-line.

More concretely, your example would then be:

expected: {1 (0 1), 3 (0 1 2 3), 5 (0 1 2 3 4 5)}
  actual: {5 (0 1 2 3 21 5), 3 (0 1 2 3), 1 (0 1)}

    diff: - {5 [nil nil nil nil 47}
          + {5 [nil nil nil nil 217}

That way we'd still benefit from the easier eyeballing of actual vs expected without the diff output interfering.
(Since the diff is "self-contained", I'm assuming it relies less on back-and-forths.)

But, as bbatsov's mentions, I'm unsure how the lack of whitespace between actual and expected would play out when expected/actual are multi-line.

[bbatsov]

I meant the same @aisamu, but I don't feel strongly about this.

[bbatsov]

I guess we can always change this down the road.

[bbatsov]

@vemv Just add some changelog entry and this one is good to go.

[vemv]

Makes sense @aisamu - implemented!

[mattiase]

What does this regexp really match?

[vemv]

Escaped newlines as coming from Clojure / cider-nrepl. These aren't vanilla elisp newlines.

[mattiase]

It matches the letter "n" in a string.

[vemv]

Thanks! will fix

[vemv]

605e253

[mattiase]

Thank you!