Post #1963483

This might not be the mainissue after all, as a new string implementation with logarithmic cost of all operations does not help. Apparently one needs to remember to benchmark and profile the code. It seems that currently my level-1 parser has cubic cost in the length of the input 🙃