changes copy-pasted from ExtendedArrayBackedAttributes.java

changes copy-pasted from ExtendedArrayBackedAttributesBuilder.java

changes copy-pasted from ExtendedAttributeKey.java

changes copy-pasted from ExtendedAttributes.java

changes copy-pasted from ExtendedAttributesBuilder.java

Is your plan to keeep ExtendedAttributes around for a couple of releases to give folks a chance to migrate gracefully?

yeah, good point, deprecated them in c133eea

oh no, that spun out of control due to suppressing usages, reverted, I'd suggest that I do that as a follow-up PR

leaving this discussion open so I don't forget

Is the logic here based on any sort of reference or specification you can link to? Trying to get a sense of how to evaluate it for completeness / correctness.

https://datatracker.ietf.org/doc/html/rfc8259

Thanks. Here's the relevant section: https://datatracker.ietf.org/doc/html/rfc8259#section-7

We shouldn't need to duplicate this since ValueBytes.create already copies the array, so it should be strictly read only / immutable here.

99bb2f1

Nice javadoc here!

Could add a helper to ProtoJson to avoid duplicating logic in the Value implementation classes:

  static String asString(Value<?> value) {
    StringBuilder sb = new StringBuilder();
    append(sb, value);
    return sb.toString();
  }

Then every asString() implementation just looks like:

@Override
public String asString() {
  return ProtoJson.asString(this);
}

Oh wait, I see. Its the difference between the top level representation. I still think you could centralize the logic all in ProtoJson by adding a int level parameter to append and changing the logic based on whether the level was 0 or non-0. Something like the following, demonstrated on DOUBLE but applicable to all the types:

  static void append(StringBuilder sb, Value<?> value, int level) {
    switch (value.getType()) {
      case DOUBLE:
        appendDouble(sb, (Double) value.getValue(), /*wrapWithQuotes*/ level > 0);
        break;
      // other cases omitted for brevity
    }
  }

  private static void appendDouble(StringBuilder sb, double value, boolean wrapWithQuotes) {
    if (wrapWithQuotes) {
      sb.append('"');
    }
    if (Double.isNaN(value)) {
      sb.append("NaN");
    } else if (Double.isInfinite(value)) {
      sb.append(value > 0 ? "Infinity" : "-Infinity");
    } else {
      sb.append(value);
    }
    if (wrapWithQuotes) {
      sb.append('"');
    }
  }

seems more complicated to me, but can apply this pattern if you prefer

Nah you can ignore this if you like the way it is. I don't like the duplicate logic, but my suggestion is crufty. Maybe I'll go back and see if there's a cleaner way to DRY it up post merge.