I know this is not necessarily a problem question but more of a style question. I understand that it doesn't have to be pretty for the most part.

In general, context managers are used for mutexing, acquiring resources, and many other things. The context managers usually yield the acquired resource or just provide wrapping functionality.

In general, generators are very capable us directing flow with conditionally yielding and custom iteration.

In my situation, I am writing a pipeline for an automatic probe station with a source and sink, but the probe station may or may not be able to probe multiple pins at the same time. That is, either iterating through each pair or being capable of connecting all pairs at the same time.

For a given pin group there is a list or pin pairs to be stressed. There is also a set of corresponding pins that are tested to determine pass or fail of the pin group given the stress level.

I have implemented a context manager for connecting to pins, yielding for testing, and then disconnecting on clean up.

Given a flag for the moment, I can control whether the loop connects all at once versus iterating through each. In the actual implementation the probe_context_generator will live in a concrete class that will offer only what is needed. The multi flag would only be offered if the concrete class can actually do either

Here is the code.

import contextlib
import random
import time
from itertools import product
from typing import ContextManager, Generator

# Business Logic
def probe_context(p):  # Belongs to device specific probe station class
        print(f'Connecting pins to {p}')
        yield p
        print(f'Disconnecting {p}')

def probe_context_generator(*pin_pairs, multi=False):
    if multi:
        yield probe_context(pin_pairs)
        for pin_pair in pin_pairs:
            yield probe_context(pin_pair)

# Application Code
def probe_for_pulse(multi, p_level, pulse_pin_pairs):
    for probe_pulse_pins_con in probe_context_generator(*pulse_pin_pairs, multi=multi):
        with probe_pulse_pins_con as pulse_pins:
            yield p_level, pulse_pins

def check_leakage(multi, leakage_pin_pairs):
    leakage_results = []
    for probe_leakage_pins_con in probe_context_generator(*leakage_pin_pairs, multi=multi):
        with probe_leakage_pins_con as leakage_pin_pair:
            print(f'Checking leakage on {leakage_pin_pair=}')
            leakage_results.append(random.choice([True, False]))
    return leakage_results

def ready_next_pulse_generator(p_levels, pin_groups: list[PinGroup], multi=False):
    for pin_group in pin_groups:
        print(f"\n!!!!!  Pin Group {pin_group.group_num}  !!!!!")
        for p_level in p_levels:

            # Measure Pass fail
            leakage_results = check_leakage(multi, pin_group.leakage_pin_pairs)
            print("broadcast leakage data to plots and data saving thread")

            if not all(leakage_results):
                print(f"Pin Group {pin_group.group_num} failed {p_level}")

            yield from probe_for_pulse(multi, p_level, pin_group.pulse_pin_pairs)

            print(f"Pin Group {pin_group.group_num} passed {p_level}")

PINS = ["A1", "B1", "C1"]
pulse_levels = [125, 250, 500]

class PinGroup:
    pulse_pin_pairs = [
        (force, sense) for force, sense in product(PINS, PINS) if force != sense
    leakage_pin_pairs = [("Leakage force pin", "Leakage sense pin")]

    def __init__(self, group_num):
        self.group_num = group_num

print("##### Probe Single #####")
for pulse_level, connected_pin in ready_next_pulse_generator(pulse_levels, [PinGroup(i) for i in range(5)]):
    print(f'\tarm scope for {connected_pin}')
    print(f"\tpulse {connected_pin=}, {pulse_level=}")
    print(f"\tread data for {connected_pin}")

print("\n" * 2)

print("##### Probe Multi #####")
for pulse_level, connected_pins in ready_next_pulse_generator(pulse_levels, [PinGroup(i) for i in range(5)], multi=True):
    print(f'\tarm scope for {connected_pins}')
    print(f"\tpulse {connected_pins=}, {pulse_level=}")
    print(f"\tread data for {connected_pins}")

My main question is, I guess, is the probing_context sufficient need for utilizing the contextmanager functionality? Would making the arming of the scope and reading the data also contextmanager increase the chances of this code being used correctly, or does it seem like I'm overusing a cool thing? Subjective, I know. But I haven't seen a lot of things like this. The probing_context, for a robotic probe station, will handle path planning and doing the actual moving.

  • 2
    \$\begingroup\$ Note that the title should be the app's purpose, not the question you're asking. Generic best practices are off-topic. Thanks. \$\endgroup\$
    – ggorlen
    Commented Apr 15, 2023 at 15:51

1 Answer 1


You asked essentially, "is this weird?" or... "am I doing this right?"

You are doing it right. That looks like very nice idiomatic python code to me.

The pattern is:

  1. connect
  2. stimulate the SUT and record measurements
  3. disconnect

with the requirement that disconnect() must happen no matter what. A perfect match for a context manager. Are you overusing a cool thing? No, I don't believe so.

Would making the arming of the scope and reading the data also a contextmanager increase the chances of this code being used correctly

I didn't exactly follow that. It appears to me that step (1) arming is what the caller wants and you've packaged that up nicely, and then step(3) disconnect is the invariant we're trying to preserve, even in the face of misbehaving user code. If the step (2) user code that reads data has its own concerns, then sure, it might use a context manager. But the OP doesn't show me anything that seems to motivate that.

increase the chances of this code being used correctly

I really like the way you're thinking. Arrange for the Right way to be the Path of Least Resistance, and client code will tend to do what you were hoping for.

Along those lines, take some care to export only a limited number of public entry points. In python we're all grownups and it's never impossible to reach in for some _private() function, but you can offer strong hints which will be respected during code reviews.

Take care to write a couple of """docstrings""" that span a few paragraphs, so callers will see how you expect them to call into the Public API of your library.

This is very clear, solid code that achieves its design goals.

I would be happy to delegate or accept maintenance tasks on this codebase.

  • \$\begingroup\$ Wow so wholesome \$\endgroup\$ Commented Apr 16, 2023 at 0:06

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.