zio-test: Remove the `"... + x more"` feedback from zio-test outputs #9900

guizmaii · 2025-05-28T08:51:46Z

It's extremely frustrating to have a test framework which doesn't show/report the data when there's a diff between the result of the test execution and the expected result

I always have to add some debug println and relaunch my tests, to be able to see this diff.
IMO, that shouldn't be the case.
zio-test should show me the result data and the expected data in its output so I can fix my test

Effectively reverting #8692

I tried to optmise a bit the code so that performances aren't impacted too much by this change.

I tested the perfs with this test (coming from the initally reported issue #8644):

test("speed test - comes from https://github.com/zio/zio/issues/8644") {
      final case class ComplexObject(
        id: String,
        name: String,
        name2: String,
        name3: String,
        name4: String,
        name5: String,
        name6: String,
        name7: String,
        name8: String,
        name9: String,
        name10: String,
        name11: String
      )

      object ComplexObject {
        def of(i: Int): ComplexObject = {
          val s = i.toString
          ComplexObject(
            id = s,
            name = s,
            name2 = s,
            name3 = s,
            name4 = s,
            name5 = s,
            name6 = s,
            name7 = s,
            name8 = s,
            name9 = s,
            name10 = s,
            name11 = s
          )
        }
      }

      for {
        _ <- ZIO.unit
        // sample of complex objects
        hugeList = (1 to 1000).map(i => ComplexObject.of(i)).toList

        shouldNotHave20  = assertTrue(!hugeList.exists(_.name3 == "20"))
        shouldNotHave510 = assertTrue(!hugeList.exists(_.name3 == "510"))
        shouldNotHave780 = assertTrue(!hugeList.exists(_.name3 == "780"))
        shouldNotHave999 = assertTrue(!hugeList.exists(_.name3 == "999"))
      } yield TestResult.allSuccesses(
        shouldNotHave20,
        shouldNotHave510,
        shouldNotHave780,
        shouldNotHave999
      )
    }

With the code optimisations, the test indeed takes a bit longer to execute, but only around ~550-600ms, which isn't "forever" in my opinion (the author of the original ticket was reporting, I quote, "The exists assertions take forever")
I have a similar computer to the reporter of the issue: Macbook Pro M3 36Gb of RAM.

test/shared/src/main/scala/zio/test/PrettyPrint.scala

guizmaii · 2025-05-28T10:29:43Z

test/shared/src/main/scala/zio/test/PrettyPrint.scala

        s"""$name($spacer$indentedBody$spacer)"""

-      case string: String =>
-        val surround = if (string.split("\n").length > 1) "\"\"\"" else "\""


string.split("\n").length > 1 not a great way to check if a String contains a Character 😕

guizmaii · 2025-05-28T11:27:10Z

test/shared/src/main/scala/zio/test/PrettyPrint.scala

-        val body = labels0
-          .zip(product.productIterator)
-          .map { case (key, value) =>
-            s"${(key + " =").faint} ${PrettyPrint(value)}"
-          }
-          .toList
-          .mkString(",\n")
-        val isMultiline  = body.split("\n").length > 1
-        val indentedBody = indent(body, if (isMultiline) 2 else 0)


This wasn't great. We were building a String with \n as separator to then split it twice on this same character. I implemented an optimised version of this algorithm

test/shared/src/main/scala-2.13/zio/test/PrettyPrintVersionSpecific.scala

guizmaii · 2025-05-28T12:03:36Z

test/shared/src/main/scala-2.12/zio/test/PrettyPrintVersionSpecific.scala

    Iterator.continually("")
  }
+
+  def prettyPrintProduct(product: Product): String = {


previous, non-optimized, compatible with Scala 2.12, version of the algorithm

See #9900 (comment)

guizmaii · 2025-07-01T16:31:08Z

@kyri-petrou @hearnadam I took all your comments into account

hearnadam

LGTM aside from the test changes.

hearnadam · 2025-07-02T07:07:37Z

test/shared/src/main/scala/zio/test/PrettyPrint.scala


-      case set: Set[_] =>
-        prettyPrintIterator(set.iterator, set.size, className(set))
+      // For why `Nil.type` is used. See https://github.com/zio/zio/pull/9900#discussion_r2121380398


Instead of a comment, can you add a test for Vector.empty? (or any other non-List empty Seq)?

Done!

Interestingly, the previous code was printing Nil for Vector.empty while now it correctly prints Vector() 🙂

I keep the comment as using _: Nil.type is not usual code

hearnadam · 2025-07-02T07:09:20Z

test-tests/shared/src/test/scala/zio/test/PrettyPrintSpec.scala

+      }
+
+      for {
+        _ <- ZIO.unit


Can you avoid the ZIO/for completely?

val x = val y = ... assertTrue(

guizmaii marked this pull request as draft May 28, 2025 08:52

guizmaii self-assigned this May 28, 2025

guizmaii force-pushed the test_remove_+_more branch 4 times, most recently from 485e014 to e3430f6 Compare May 28, 2025 09:20