evalf does not correctly interpret the number of digits

[mzaffalon]

evalf seems to convert the argument to Float64 and then extend its precision. I would expect evalf to behave like the last command in the example below. Is my expectation incorrect?

julia> Sym(2.45).evalf(70)
2.450000000000000177635683940025046467781066894531250000000000000000000
julia> BigFloat(2.45, 200) # turn a Float64 into a BigFloat
2.45000000000000017763568394002504646778106689453125
julia> BigFloat("2.45", 200) # parse "2.45" as a BigFloat
2.4500000000000000000000000000000000000000000000000000000000005

This is with [24249f21] SymPy v1.0.28.

Probably related to #369.

[jverzani]

I was going to suggest sympify("2.45") to avoid the conversion to a float within julia before SymPy has a chance to do something, but this has the same issue. Scanning the docs, I see you can use Float passing a string (similar to the big constructor):

sympy.Float("2.45", 100)

[mzaffalon]

It looks like the problem remains:

julia> BigFloat(N(sympy.Float("2.45"), 256), 200)
2.45000000000000017763568394002504646778106689453125

[jverzani]

You need to specify a precision to Float, as in Float(“2.45”, 200)

[mzaffalon]

Oh, I missed that!