not_inside constraint unexpectedly reducing output set significantly

[apaiva]

Here's an example ( on version pyvsc 0.8.6.6319294550) a 'not_inside' constraint is reducing a rand enum's set down to almost nothing, where I'd expect it to only exclude the specified values. You can see in the output below that the first set of randomizations (without the not_inside constraint) give a nice distribution; however the second set 'op2' result in only 'C'.

Output

op1:L op1:M op1:I op1:B op1:M op1:C op1:M op1:D op1:J op1:L op2:C op2:C op2:C op2:C op2:C op2:C op2:C op2:C op2:C op2:C

Code Example

`import vsc from enum import IntEnum, auto

class TestIntEnum(IntEnum): A = 0 # B = 1 # C = 2 # D = 3 # E = 11 # F = 5 # G = 6 # H = 7 # I = 15 # J = 10 # K = 14 # L = 8 # M = 9 # N = 4 # O = 128 # P = 129 # Q = 0x82 # R = 0x8A # S = 0xff #

@vsc.randobj(srcinfo=False) class RandOp:

def __init__(self):
    self.r_op = vsc.rand_enum_t(TestIntEnum)

@vsc.constraint
def c_op(self):
    self.r_op.inside(vsc.rangelist(
        TestIntEnum.A,
        TestIntEnum.B,
        TestIntEnum.C,
        TestIntEnum.D,
        TestIntEnum.E,
        TestIntEnum.F,
        TestIntEnum.G,
        TestIntEnum.H,
        TestIntEnum.I,
        TestIntEnum.J,
        TestIntEnum.K,
        TestIntEnum.L,
        TestIntEnum.M ))

@vsc.randobj() class RandOp2(RandOp):

@vsc.constraint
def c_op2(self):
    self.r_op.not_inside(vsc.rangelist(
        TestIntEnum.A,
        TestIntEnum.B))

if name == 'main': for _ in range(10): rand_op = RandOp() rand_op.randomize() print(f"op1:{rand_op.r_op.name:s}")

for _ in range(10):
   rand_op = RandOp2()
   rand_op.randomize()
   print(f"op2:{rand_op.r_op.name:s}")`

[mballance]

Good catch on this one, @apaiva, and many thanks for the easy-to-reproduce testcase. not_inside is represented as !inside internally, and the randomization code was incorrectly treating !inside the same as inside. For now, PyVSC simply provides no special handling for !inside. This results in much better distribution than you were seeing previously. However, there is a visible 'spike' in the histogram. Let me give some thought to how to improve this. I've released 0.8.7 with the improvements. Feel free to give it a try!