Better error message when parsing Float32 fails

[jonas-schulze]

Recently, I was trying to scale a variable f and subtract x from it, 2*f - x. Without * and spaces, that leads to a malformed expression error:

julia> mirror(f, x) = x > f ? (2f - x) : x
mirror (generic function with 1 method)

julia> mirror(f, x) = x > f ? (2*f-x) : x
mirror (generic function with 1 method)

julia> mirror(f, x) = x > f ? (2f-x) : x
ERROR: syntax: malformed expression
Stacktrace:
 [1] top-level scope
   @ none:1

It took me a while to figure out that Julia was trying to parse this as a Float32 where f seperates mantissa from exponent, e.g. 2f-3. A better error message would have helped me a lot, e.g. cannot parse expression as Float32. Then, the docstring of Float32 would reveal the use of f that I didn't have in mind.

Would it be reasonable to parse 2f-X as 2f - X if X is anything but a number?
This would make 2f-x behave as expected.

---

Side note: if there is a variable f around, this leads to some interesting situations:

julia> f = 10
10

julia> 2f + 1
21

julia> 2*f+1
21

julia> 2f+1
20.0f0

julia> 2f+1f # (2f+1)*f
200.0f0

[ViralBShah]

Much nicer with the new Julia parser thanks to @c42f.

julia> mirror(f, x) = x > f ? (2f-x) : x
       ERROR: syntax: malformed expression
       Stacktrace:
        [1] top-level scope
          @ none:1
ERROR: ParseError:
# Error @ REPL[3]:1:25
mirror(f, x) = x > f ? (2f-x) : x
#                       └─┘ ── invalid numeric constant
# Error @ REPL[3]:2:25
mirror(f, x) = x > f ? (2f-x) : x
ERROR: syntax: malformed expression
#                       └─────────┘ ── extra tokens after end of expression
# Error @ REPL[3]:3:12
ERROR: syntax: malformed expression
#          ┌
Stacktrace:
 [1] top-level scope
┘ ── line break after `:` in range expression
# Error @ REPL[3]:4:5
Stacktrace:
 [1] top-level scope
#   └──────────────┘ ── extra tokens after end of expression
# Error @ REPL[3]:5:5
 [1] top-level scope
   @ none:1
#   ╙ ── whitespace is not allowed here
Stacktrace:
 [1] top-level scope
   @ none:1

[jonas-schulze]

Nice. 🚀

Will this be part of Julia 1.10?

[oscardssmith]

already in 1.10