HTTP monitor

TLDR:
    make check
    make test
    cat input.csv | make HIGH_TRAFFIC_THRESHOLD=10 DISPLAY_STATS=true run_stdin

Disclaimer:
    I definitely spent more than 3 hours on this.
    By the 3 hour mark I was maybe done with the stats part and some tooling, but there were
    no tests and no alert functionality. I think I spent about 5 hours with this task in total unfortunately.

Approach:
    The input is almost sorted by time, so we can write all the logic as if the input was sorted
    and sort incoming batches. If the "skew" would be bigger than it is in the
    sample file, this approach wouldn't work. See config.py to for the biggest combination that works.

    Doing the batch sort simplifies rest of the code a lot, since we can now accumulate all the data for given
    second and only when we encounter the next second take some potential action.

    The stats part is quite straightforward, we just store all the data for the given time period and then
    turn it into strings. It would be possible to do this per-second on the fly to save memory, but the approach
    when everything is stored is easier. The array is reset after stats are printed.

    For the alerting part I've decided to store the per-second request counts in a sliding window of sorts.
    The approach I took only works if there are no gaps in the timestamps. It would be easy to extend though,
    by adding zeros to the window whenever we see a new timestamp which isn't prev_timestamp + 1.

Files:
    main.py: pre-sorting functions and reading the input.
    config.py: configuration values, some are changeable using env variables.
    log_processor.py: contains 3 classes, Entry (stores the log lines) and LogProcessor, which is the main class.
                      LogProcessor expects sorted input and works on line-by-line basis. It returns a tuple
                      with possible stats and alert actions as well as the strings to print.
                      The enum with actions is the third class.

    tests.py: really simple test for the alerting logic.
    Makefile: build, test or run the project.
    Dockerfile: really simple dockerfile, doesn't install anything.
    test.Dockerfile: dockerfile for tests, installs pytest.
    docker-compose.yml: docker compose file with just one container.
    input.csv: the test file that was provided.

TODOs:
    More tests, only had time to write a simple test for the alerting logic, the stats functionality is
    not tested at all.

    More docstrings, really light on these, hopefully most of the code is self-explanatory but the public
    methods and the LogProcessor need more documentation for sure.

    I wrote the stats logic first and it shows, the alerting part of the code behaves a bit differently.

    The naming is a bit incosistent.

    Partial stats are not printed at the end.

    Some of the checks that decide whether this is the first entry are a bit ugly.

    There is almost no input checking.

How to run:
    You need make and docker installed. To start the program do "make run_stdin" and pipe the input to it.
    See the top of this readme.

    mypy and flake8 checks can be done by "make check".

    Tests can be executed by "make test".

    After you're done with testing, do "make cleanup" to delete the created docker images.

Author:
    [email protected]