Hi, I am trying to generate a reflection-based ui so I can edit flatbuffers. I don't care if the performance isnt ideal when im BUILDING them in this case, as its a GUI tool, and I understand as strings and other optional tables are added by the user, that the underlying vector has to be resized.

Ive gone through ReflectionTest in test.cpp, and am using a resizing vector with a PIV just like the test.
I can set a struct member just fine, but I cant set an optional string, even with the AddFlatBuffer function.

It goes something like this:

	virtual void text_changed( const String& p_str ) override
	{
		// fb
		// "fb" is simply a wrapper object that holds a vector data, and the "piv" to the root inside of it
		// it can be coerced to both Table* and Table* via operator() casts for ease of use
		const flatbuffers::String* str = flatbuffers::GetFieldS( *fb, *field );
		// There's already a string in the table, modify it
		if ( str ) {
			flatbuffers::SetString(
				*schema,
				std::string( p_str.ascii() ),
				flatbuffers::GetFieldS( *fb, *field ),
				&fb->data );
			
			print_line("MODIFIED FB STRING");
			return;
		}
		// we need to add a new one
		flatbuffers::FlatBufferBuilder s_fbb;
		s_fbb.Finish( s_fbb.CreateString( (const char*)p_str.ascii() ) );
		auto string_ptr = flatbuffers::AddFlatBuffer( fb->data, s_fbb.GetBufferPointer(), s_fbb.GetSize() );
		
		bool ret_val = flatbuffers::SetFieldT( *fb, *field, string_ptr );
		
		print_line("ADDED FB STRING");
	}

In this case, ret_val is always false. It seems to fail in SetFieldT >>> table->SetPointer() >>> GetOptionalFieldOffset(field) which returns 0, causing it to fail.

I notice the test example uses a vector of strings, and manually resizes the vector to add in a new string offset, which is different. I do not see a test case that simply adds in a string by itself.

Is there a different way of doing this? Or is it a bug?