Way to enforce `BigFloat` math by default

[tgross35]

julia> big(pi)/2
1.570796326794896619231321691639751442098584699687552910487472296153908203143099

julia> big(pi/2)
1.5707963267948965579989817342720925807952880859375

That is quite an easy mistake to make. Is there a way to do something like set the default numeric type for literals to BigFloat, to help avoid silently wrong math?

Overriding the builtin values with something like pi = big(pi) would also help, but unfortunately this is not allowed.

[jar1]

I bet it would be possible to write a JET analyzer type that checks for floats.

"Unstable API" Analysis · JET.jl

[rfourquet]

That’s the purpose of the SwapLiterals.jl and SafeREPL.jl packages. At the REPL:

julia> using SafeREPL

julia> swapliterals!(Float64 => :BigFloat)  # `BigFloat` function is applied to Float64 literals

julia> 1.2 # identical to `BigFloat(1.2)`
1.1999999999999999555910790149937383830547332763671875

julia> big"1.2"
1.200000000000000000000000000000000000000000000000000000000000000000000000000007

julia> swapliterals!(Float64 => "@big_str")  # apply `@big_str` macro to stringified Float64 literals
1.200000000000000000000000000000000000000000000000000000000000000000000000000007

julia> pi/2 # no Float64 literals, so vanilla julia
1.5707963267948966

julia> pi/2.0
1.570796326794896619231321691639751442098584699687552910487472296153908203143099

julia> swapliterals!(Float64 => :BigFloat, Int => :BigInt) # convert Int literals into BigInt

julia> pi/2
1.570796326794896619231321691639751442098584699687552910487472296153908203143099

To swap literals in code (not in REPL) :

julia> using SwapLiterals; @swapliterals Float64 => "@big_str" 1.2
1.200000000000000000000000000000000000000000000000000000000000000000000000000007

[photor]

that’s not “swap”, that’s "assign"

[nsajko]

Just do pi = BigFloat(MathConstants.pi).

[mkitti]

You may be interested in SafeREPL.jl:

[ANN] SafeREPL: use BigInt by default at the REPL

With Julia 1.5 around the corner, I’m happy to present SafeREPL, a little experimental package which by default interprets Int and Int128 REPL literals as BigInt, and Float64 literals as BigFloat: julia> using SafeREPL julia> factorial(40) 815915283247897734345611269596115894272000000000 julia> sqrt(2.0) 1.414213562373095048801688724209698078569671875376948073176679737990732478462102 julia> c = 299_792_458; c^3 # the cube of the speed of light can even be computed safely! 26944002417373989…

julia> using SafeREPL

julia> π
π = 3.1415926535897...

julia> π/2
1.570796326794896619231321691639751442098584699687552910487472296153908203143099

julia> BigFloat(π)/2
1.570796326794896619231321691639751442098584699687552910487472296153908203143099

[mtfishman]

Or “replace”.