Errors when using Lambdify on sympy.Sum

[MostlyHarmless420]

I have encountered a couple of errors when using lambdify and Sum:
When summing over a symbol, it brings up an error:

using Sympy

x = symbols("x")

lambdify(sympy.Sum(x, (x,0,10)))() #>UndefVarError: x not defined

Also when using euler_maclaurin summing over certain functions it produces an error:

x, y = symbols("x y")
lambdify(sympy.Sum(x*y, (x,0,10)).euler_maclaurin(n=3)[1])(3.) # Works fine
lambdify(sympy.Sum(sympy.exp(x), (x,0,10)).euler_maclaurin(n=3)[1])() #Works fine
lambdify(sympy.Sum(sympy.exp(x*y), (x,0,10)).euler_maclaurin(n=3)[1])(3.) #MethodError: no method matching &(::Bool, ::Sym)

[jverzani]

Okay, thanks for reporting. I made some adjustments in #296 that address these last cases.

The first did lead to a useful change, but is still problematic: well, Sum(x, (x,0,10)) doesn't have any free variables; but that isn't it, reallySum(x, (x,0,10)) does not have a Julia counterpart (but Sum(x, (x,0,10)).doit() does), so the converted expression refers to sympy objects

convert(Expr, sympy.Sum(x, (x,0,10)))

That Sum is a problem, but so is x. You can bypass these manually by creating the function in the calling module, along the lines of:

Sum = sympy.Sum
x = symbols("x")
ex = Sum(x, (x,0,10))
ux = convert(Expr, ex)
fn = eval(Expr(:function, Expr(:call, gensym()), ux))
fn()  # a SymPy object though

[MostlyHarmless420]

Thanks for looking into it.
Just tested the latest #master and, while not receiving an error any more, Lambdify isn't properly substituting for certain functions and is returning a Sym type.

x, y = symbols("x y")
lambdify(sympy.Sum(sympy.exp(x*y), (x,0,10)).euler_maclaurin(n=3)[1])(3.)  #Returns: 7699986595795.14 + exp(10*y)/y - 1/y

[jverzani]

Hmm, there are two errors, one I can work around, one I'm stumped with:

x, y = symbols("x y")
ex = sympy.Sum(sympy.exp(x*y), (x,0,10)).euler_maclaurin(n=3)[1]

ux = convert(Expr, ex)

## ux = :(1 / 2 + (1 / 2) * exp(10y) + -1//12y + -1//30240 * y ^ 5 + 1//720 * y ^ 3 + -1//720 * y ^ 3 * exp(10y) + 1//12 * y * exp(10y) + 1//30240 * y ^ 5 * exp(10y) + ifelse(&(y > -Inf, y < Inf, y !== 0), exp(10*y)/y - 1/y, ifelse(true, 10, nothing)))

## The  `&(...)` is an issue we can work around
ALL(xs...) = all(xs)
ux = SymPy.walk_expression(ex, fns=Dict("And" => :ALL))

## then
fn = eval(Expr(:function, Expr(:call, gensym(), :y), ux))

fn(3.0) # has a `y`

## But, directly this works (avoiding) eval...

fn = y -> (1 / 2 + (1 / 2) * exp(10y) + -1//12y + -1//30240 * y ^ 5 + 1//720 * y ^ 3 + -1//720 * y ^ 3 * exp(10y) + 1//12 * y * exp(10y) + 1//30240 * y ^ 5 * exp(10y) + ifelse(ALL(y > -Inf, y < Inf, y !== 0), exp(10*y)/y - 1/y, ifelse(true, 10, nothing)))
fn(3.0)