Update MessageNano#toString() to return mostly valid TextFormat.

The output of toString is now aligned with that used by non-nano and C++
runtimes, with the exception of groups. Groups should be serialized using a
camelized name (e.g. "FooBar" rather than "foo_bar") however the nano runtime
does not have information on which fields are groups.

Changes are:
  - bytes fields are output within double-quotes, non-printable characters are
    output as octal escape sequences (i.e. \NNN);
  - field identifiers are output in underscored format;
  - unset fields are not output (rather than printing "null");
  - the type name of the root message is not output.

With these changes the nano toString, normal toString, and C++'s DebugString all
produce equivalent output when given the same message. (Provided that message
uses no deprecated features.)

Change-Id: Id4791d73822846db29344db9f7bc3781c3e183a6

1 files changed, 4 insertions, 1 deletions

diff --git a/java/src/main/java/com/google/protobuf/nano/MessageNano.java b/java/src/main/java/com/google/protobuf/nano/MessageNano.java
index 8d4ec39..0885cdb 100644
--- a/java/src/main/java/com/google/protobuf/nano/MessageNano.java
+++ b/java/src/main/java/com/google/protobuf/nano/MessageNano.java
@@ -127,7 +127,10 @@ public abstract class MessageNano {
     }
 
     /**
-     * Intended for debugging purposes only. It does not use ASCII protobuf formatting.
+     * Returns a string that is (mostly) compatible with ProtoBuffer's TextFormat. Note that groups
+     * (which are deprecated) are not serialized with the correct field name.
+     *
+     * <p>This is implemented using reflection, so it is not especially fast.
      */
     @Override
     public String toString() {