Exponents with sign should be recognized as decimal numbers.

[faze-geek]

Python 3.9 -

>>> a = 1e2
>>> a
100.0
>>> a = 1e-2
>>> a
0.01

Pocketpy -

>>> a = 1e2
>>> a
100.0
>>> a = 1e-2
  File "<stdin>", line 1
    a = 1e-2
          ^
SyntaxError: invalid number literal
>>> a = 1e+2
  File "<stdin>", line 1
    a = 1e+2
          ^
SyntaxError: invalid number literal

[faze-geek]

This is not really an error in code but an extension of the partial implementation of exponentiation that is already supported in pocketpy. Hence, I didn't open it up as a bug.

[gimmeursocks]

@blueloveTH i dont know if i should make a pull request as i dont know if this behaviour is intended or not but i modified lexer::eat_number

if it passes by a -ve or a +ve character it skips it and parses the rest

       if(text[text.size()-1] == 'e'){
            if(*i == '+' || *i == '-'){
                i++;
                while(kValidChars.count(*i)) i++;
                text = std::string_view(token_start, i - token_start);
                this->curr_char = i;
            }
        }

        // try float
        double float_out;
        char* p_end;
        try{
            float_out = std::strtod(text.data(), &p_end);
        }catch(...){
            SyntaxError("invalid number literal");
        }

it passes all tests but as i said im reluctant to make a pr for this.

[blueloveTH]

I know about this. But I am reluctant to fix it.

[carlkl]

Could this be noted in https://pocketpy.dev/features/differences/, as this kind of literals are quite common?

[faze-geek]

As mentioned in the documentation pocketpy is a project also being heavily used for scientific computing. In that context it is crucial to have special numbers/constants like (Inf, eps, e, pi etc.) and their various formats.
Hence I'm a bit confused on the reluctance to support this literal.

[carlkl]

Maybe a C++ runtime math parser like https://beltoforion.de/en/muparserx/ could be useful as an addition?
EDIT: muparsex needs STL unfortunately.