Text.to_display_text returns Text

[JaroslavTulach]

Especially when handling exceptions it is important to have proper implementation of to_display_text (which is invoked from PanicException.getMessage(). However Text! has some wrong implementation. Try:

main = IO.println "Ahoj".to_display_text

and it prints Text. That's wrong. I should print Ahoj.

[JaroslavTulach]

Proposed fix:

diff --git engine/runtime/src/main/java/org/enso/interpreter/node/expression/builtin/text/AnyToDisplayTextNode.java engine/runtime/src/main/java/org/enso/interpreter/node/expression/builtin/text/AnyToDisplayTextNode.java
index e0bed6343..589f9f49a 100644
--- engine/runtime/src/main/java/org/enso/interpreter/node/expression/builtin/text/AnyToDisplayTextNode.java
+++ engine/runtime/src/main/java/org/enso/interpreter/node/expression/builtin/text/AnyToDisplayTextNode.java
@@ -31,6 +31,11 @@ public abstract class AnyToDisplayTextNode extends Node {
       throw new IllegalStateException(e);
     }
   }
+  
+  @Specialization
+  Text identity(Text self) {
+    return self;
+  }

[radeusgd]

I think there was some story here.

I remember to_display_text was being added because to_text would sometimes return too big payloads and overwhelm the IDE. After all the default Text visualization is planned to be Lazy to handle big texts.

If we do just identity here, stuff like error messages which want to display this text can get huge (for example if my Text is a contents of a whole book).

So I suggest that maybe we should truncate the returned Text at some character limit? (though if we do so we need to be careful to not cut the text between two codepoints belonging to a single grapheme cluster - so some slight care will need to be taken to correctly find a Break between proper Characters (BreakIterator.getCharacterInstance should be our friend)

[JaroslavTulach]

Related pull requests, issues and discussions:

• going with to_display_text instead of to_pretty_text
• Error message representations #1498
• it's just a type name most of the time...
• Make To Text Conversion Identity for Text #4009
• shorten very much and generate labels that would fit on a node.
• Visualization of DataflowError #4076

CCing @4e6, @kustosz, @hubertp. Also related:

• The same method is dispatched for instances of a type and the Atom representing the type #1538
• IO.println: Type error: expected str to be Text, but got Function #5961

[jdunkerley]

Agree with Radek's comment on truncating the text - maybe first 80 grapheme clusters.
Otherwise @JaroslavTulach's fix looks like the right one.