Created by: Aurelius84

• Fix create_cond_node bug that raises error while return_name_ids is empty.

# The repetition code
def bug_test(x):
    x = to_variable(x)
    a = fluid.layers.fill_constant(shape=[1,2], value=9, dtype="int64")
    b = fluid.layers.fill_constant(shape=[3], value=3,  dtype="int64")
    if x.numpy()[0] > 0:
        fluid.layers.assign(x, a)
    else:
        fluid.layers.assign(b, a)
    return a

• Support and/or in dygraph_to_static control_flow_if

# dygraph function
def foo(x, y=None):
    batch_size = fluid.layers.shape(x)
    if batch_size[0] > 1 and y is not None:
        x = x + 1
    if y or batch_size[0] > 1:
        x = x - 1
    return x

# After transforming into static
def foo(x, y=None):
     batch_size = fluid.layers.shape(x)

    def true_fn_0(x):
        x = x + 1
        return x

     def false_fn_0(x):
        return x
 
    bool_tensor_0 = fluid.layers.fill_constant(shape=[1], value=bool(y is not None), dtype='bool')
    logic_and_0 = fluid.layers.logic_and(x=batch_size[0] > 1, y=bool_tensor_0)

    x = fluid.layers.cond(logic_and_0, lambda: true_fn_0(x), false_fn_0(x))

    def true_fn_1(x):
        x = x - 1
        return x

     def false_fn_1(x):
        return x

    bool_tensor_1 = fluid.layers.fill_constant(shape=[1], value=bool(y), dtype='bool')
    logic_or_0 = fluid.layers.logic_or(x=bool_tensor_1, y=batch_size[0] > 1)

   x = fluid.layers.cond(logic_or_0, lambda: true_fn_1(x), false_fn_1(x))

    return x

• Support complicate joined and/or syntax.

# more than two `and`
if x and batch_size[0] > 1 and y is not None and mean_res.numpy()[0] > 0:

# joined `and/or`
if (x and batch_size[0] > 1) or (y is not None and mean_res.numpy()[0] > 0):