Developed with Rust 1.55.0 on Windows Basics ====== Formatting via cargo fmt with default options. Linting via cargo clippy --all-targets. Parsing command line arguments - as long as there is at least one argument, for the file, the execution will proceed. Completeness ============ All cases should be covered. Correctness =========== Testing via cargo test. There are unit tests and also integration tests that run using the binary. The integration tests process the various .csv files in the tests directory. Safety and Robustness ===================== Errors are dealt with in one of two ways: returning a non-zero exit code with error on stderr; or ignoring it. Failure to interact with the filesystem results in a std::io::Error printed to stderr and a non-zero exit code. Data errors in the file are ignored. Data errors include things such as: * bad headers * records that cannot be read * records that makes no sense - such as a dispute on a non-existent transaction or a transaction for the wrong client Ignoring bad data like this means that processing a very large file of mostly good data is not halted. Typically other processes can detect errors where all the data is corrupted - such as detecting the output is empty / very short. Efficiency ========== Uses a two-pass approach - trading slower processing time (roughly half the speed) for substantially lower memory consumption. In the first pass through the file, we read which transactions are being disputed. This means that in the second pass through the file, we can discard any transactions we know are not disputed, once they are processed. Overall memory consumption is proportional to the number of clients - and that's limited by the maximum value of a u16. Maintainability =============== Separation of input/output and execution in the tests makes them simple to reason about Code separated into modules, with some visibility defined Code is reasonably DRY - without going over the top on this