Invalid Characters in Symbols for Saving and Loading Circuits

[KripaPanchagnula-Quantinuum]

When trying to save and load symbolic circuits with

with open("circ.json", "w") as f:
    json.dump(circuit, f)

with open("circ.json", "r") as f:
   c = Circuit.from_dict(f)

or the pickle equivalent, if the circuit symbols contain certain non-alphanumeric characters this errors with

RuntimeError: syntax error

Some protected characters include + and -, which make sense as they may be interpreted as mathematical operations. However, {}() are also not allowed, but the curly braces are necessary for TeX rendering of the symbols, if superscripts and subscripts contain more than 1 character.

For example, from InQuanto ansatze, FermionSpaceAnsatztUPS makes extensive use of curly braces and parentheses in the symbols as they are chemically meaningful, and when the circuit is used within an algorithm and protocol it's perfectly happy to do its thing, but for saving/loading it is not happy. Contrast with FermionSpaceAnsatzUCCSD whose symbols are names s{n} and d{m} by enumerating through the single and double excitations but this circuit can be saved and loaded.

Do we know a priori what non-alphanumeric characters are protected so that they can be avoided? Although I would be sad to not allow () if that happens so be it, but I think if we support TeX rendering of symbols we should try to allow {}.

[KripaPanchagnula-Quantinuum]

Trying a bunch of special characters with

def try_symbols() -> None:
    nonalphanumeric = "!£$%^*()_-+={}[]:;'@#~\<>,./?"

    invalid_syntax, unknown_tokens, non_erroring = [], [], []

    for s in nonalphanumeric:
        print(f"Running {s=}")
        c = Circuit(1)
        sym = Symbol(f"Rz{s}1")
        c.Rz(sym, 0)

        with open("symbol_circ.json", "w") as f:
            json.dump(c.to_dict(), f)

        try:
            with open("symbol_circ.json", "r") as fr:
                read_circ = Circuit.from_dict(json.load(fr))

        except RuntimeError as e:
            if str(e)[0] == "s": # syntax error
                invalid_syntax.append(s)
            elif str(e)[0] == "U": # Unknown token
                unknown_tokens.append(s)
            print(f"{e=}\nInvalid symbol {s=}\n")

        else:
            non_erroring.append(s)

    print(f"{invalid_syntax=}")
    print(f"{unknown_tokens=}")
    print(f"{non_erroring=}")

gives the following

invalid_syntax=['(', ')', '~', ',', '.']
unknown_tokens=['!', '$', '%', '=', '{', '}', '[', ']', ':', ';', "'", '#', '\\', '?']
non_erroring=['£', '^', '*', '_', '-', '+', '@', '<', '>', '/']

The characters & and | result in seg faults, presumably they're protected with special meanings in terminals. If the trailing 1 is not included in the symbol name, only £ and _ are non erroring.