Expected/actual values in REPL-run tests are not valid Clojure data

[temochka]

When running tests in REPL via Calva’s various "Run test" commands, the "expected"/"actual" values printed in Calva’s REPL buffer are not valid Clojure data.

Example test

(deftest a-test
  (is (= {:foo "bar buz"} {:foo "buz bar"})))

Output

; Running test: a-test…
; FAIL in calva-bug.core-test/a-test (core_test.clj:6):
; expected:
{:foo bar buz}
; actual:
({:foo buz bar})
; 1 tests finished, problems found. 😭 errors: 0, failures: 1, ns: 1, vars: 1

Expected behavior

; Running test: a-test…
; FAIL in calva-bug.core-test/a-test (core_test.clj:6):
; expected:
{:foo "bar buz"}
; actual:
{:foo "buz bar"}
; 1 tests finished, problems found. 😭 errors: 0, failures: 1, ns: 1, vars: 1

Additional context

• I was able to reproduce this on a clean lein new calva-bug project running a Leiningen REPL with the default profile using a single test provided above.
• It is important to print values as a valid data, because it’s a common use-case to copy the actual data and update the test with it.
• The actual result ends up being wrapped in a list for some reason, would be nice to get rid of this.
• This looks as if the values are printed via print (for human consumption) instead of prn.

[bpringe]

Thanks for the report. We've seen this list issue reported before, and it seems to be an upstream effect of cider-nrepl or clojure.test. I created this issue about it.

I'd like to track that specific concern there, but if you think this issue should be kept open to track the other concern of test output being valid Clojure data, we can keep this one open too.

[temochka]

@bpringe I don’t consider the list wrapping an issue (and it might be unrelated), but the main issue with how the test output is a daily nuisance in my workflow. I’m curious how others are currently working around it. Not only the expected/actual values are not pretty-printed (making them difficult to read), but you can’t even copy-paste them back to REPL or back to the test. "Repair" long data structures copied from the test output (which involves a lot of guessing whether a certain numeral is a string or an integer) is just not fun.

[PEZ]

Looks like a mistake is made when we print those strings.

[vemv]

Hi!

OP's own clojure-emacs/cider-nrepl#713 appears to fix this. Before merging that PR I'd like to make sure that in fact that solves this issue in Calva as intended.

Do you have it easy to verify so? Not sure if your using lein or tools.deps or...

If that helps I could release a one-off .jar.

[temochka]

@vemv FWIW, the test output in my last comment in clojure-emacs/cider-nrepl#713 is from calva running against nrepl started with a local snapshot of cider containing the change. It does appear to fix the issue.

[vemv]

Alright! That's enough evidence :)

[PEZ]

Wonderful news. Great work @temochka!