Weird (off by one?) behaviour of a trie tree I'm sketching out (still underperformant)

Thank you everybody for the suggestions. I’ve managed to make a working draft with an immutable trie. I’ll focus on optimizing it in the next days. Thanks especially to @gasche for the proposed signature, which is what I’ve used. If you want to try it I’ve pushed to the completion branch of gobba. Do you know any good micro benchmarking library that I can use to test my language? Something that produces already plotted charts or easily plottable data.