builtin,py: use M__neg__ for IntMin

This CL special-cases IntMin to use M__neg__ instead of going
through a conversion to BigInt.

Co-authored-by: Drew O'Meara <drew@plan-systems.org>

Author: drew-512

builtin/tests/builtin.py

@@ -6,8 +6,19 @@
 
 doc="abs"
 assert abs(0) == 0
+assert abs(-0) == 0
+assert abs(0.0) == 0.0
+assert abs(-0.0) == 0.0
 assert abs(10) == 10
 assert abs(-10) == 10
+assert abs(12.3) == 12.3
+assert abs(-12.3) == 12.3
+assert abs(1 << 63) == 1 << 63
+assert abs(-1 << 63) == 1 << 63
+assert abs(-(1 << 63)) == 1 << 63
+assert abs(1 << 66) == 1 << 66
+assert abs(-1 << 66) == 1 << 66
+assert abs(-(1 << 66)) == 1 << 66
 
 doc="all"
 assert all((0,0,0)) == False

py/int.go

@@ -254,8 +254,7 @@ func (a Int) M__pos__() (Object, error) {
 
 func (a Int) M__abs__() (Object, error) {
 	if a == IntMin {
-		abig, _ := ConvertToBigInt(a)
-		return abig.M__abs__()
+		return a.M__neg__()
 	}
 	if a < 0 {
 		return -a, nil