shiyanhui / router Goto Github PK

Hi!
Since, the same static string is used in the test, I think that PyPy test is a fake due to JIT. The same may be related to Go/C++ tests as time delta is too big.
If we change test as following, we get pretty the same time on CPython.

def JIT_mimic(r,m, static = [0,0,0]):
    if r is static[0] and m is static[1]:
        return static[2]
    static[0] = r
    static[1] = m
    ret = static[2] = tree.get(r, m)
    return ret
    
def testPerformance():
    def f():
        for i in range(10000000):
            JIT_mimic("/user/Lime/male/25", "GET")
            #tree.get("/user/Lime/male/25", "GET")

    print(timeit(f, number=1))

For objectivity, we need to use an array of strings (100 or 1000) with the same content (the main thing is that these should be different objects * ) and iterate them in closed loop.

* different object means arr[0] is arr[1] == False

And even with different objects PyPy can cheat us due to strings are immutable in python, so PyPy can use the same memory (i.e. the same pointer) for different objects and can still track that we pass the same args.