performance utilities

[robnagler]

This one is useful to compute memory used by a process (on Linux only):

def mem():
    with open('/proc/self/status') as f:
        m = f.read().split('VmRSS:')[1].split('\n')[0][:-3]
        return f"{int(m.strip()) // 1024:-3d}M"

This one is good for logging time of some arbitrary code in a with:

    @contextlib.contextmanager
    def time_and_log(self, name):
        s = time.perf_counter()
        yield
        r = time.perf_counter() - s
        self.log(
            "{} seconds={:.2}", name, r, pkdebug_frame=inspect.currentframe().f_back
        )

pykern.perf or some other module name should be created to support these operations. Maybe even a "how much does this operation use in DRS".