Parse generic numbers (int, hex, real)

[facontidavide]

Follow up of #85

The matching doesn't seem to work as expected if I do this: https://godbolt.org/z/jWG4zoKTP

Choice is ordered and takes the first branch that matches. When it sees 5.3 dsl::integer matches so that is parsed; it then leaves .3 unconsumed. I recommend attempting to match the real number first and ensuring that it does not match integers only: https://godbolt.org/z/dn83sxqdG

For the sake of completeness, this is the final solution I found to address negative integers and hexadecimals as well:

https://godbolt.org/z/Td76TsrKK

The only case that I am not handling correctly is the string 0x100.2 that should fail.

[foonathan]

The only case that I am not handling correctly is the string 0x100.2 that should fail.

I mean, it does fail. It leaves a .2 on the input which will cause subsequent parse operations to fail or exit early. If you add a + dsl::eof to the root production, lexy will complain about the trailing input: https://godbolt.org/z/Y8dafdqzn

You can use dsl::peek_not to get a custom error if it happens: https://godbolt.org/z/cqzjdxbTx

You can even get a bit more fancy by recovering after the error: https://godbolt.org/z/nEbfd94oE

[facontidavide]

thanks a lot!

Something that confuse me a little is this

  static constexpr auto rule = [] {
    auto hex_integer = LEXY_LIT("0x") >> dsl::integer<int, dsl::hex>;
    auto regular_integer = dsl::peek(dsl::lit_c<'-'> / dsl::lit_c<'+'> / dsl::digit<>) >> dsl::p<integer>;
    auto float_error = dsl::peek_not(dsl::period / dsl::lit_c<'e'> / dsl::lit_c<'E'>).error<invalid_float_suffix>;
    return ( hex_integer | regular_integer ) >> float_error;
  }();

I expected it to be

    return ( hex_integer | regular_integer ) + float_error;

That doesn't compile ("choice requires a branch condition").
Of course, there is something about the DSL that I am not grasping yet! 😅

[foonathan]

You're using dsl::p<Integer> in a choice dsl::p<Real> | dsl::p<Integer>, which requires that both rules are branch rules. dsl::p is a branch rule if and only if the rule of the Production is a branch rule. a + b is not a branch rule, hence the error. a >> b is a branch rule, with a the condition.

https://lexy.foonathan.net/learn/branching/
https://lexy.foonathan.net/reference/dsl/branch/#doc
https://lexy.foonathan.net/reference/dsl/choice/#choice

[facontidavide]

thanks!

[foonathan]

BTW, as of 0370d94 the recovery behavior of dsl::peek_not has been changed for this use case.
The following code checks if it has one or more characters that make up a float suffix, and if so, consumes them and issues an error covering them: https://godbolt.org/z/YbWxhE1dv

[facontidavide]

thanks. I will update my parser accordingly