didenko/omc_sim

Tested cases (from email):

Done	Test case description
☑	code should fail on integer prices
☑	varying symbols
☑	overflows
☑	underflows
☑	long string sequences
☑	floating order size
☑	wrong field composition
☑	error output
☑	specific process termination
☑	crash exceptions
☑	detect errors and crashes as unexpected behavior
☑	detect errors as expected reaction to a wrong input

Structure:

• bin directory has the exchange_sim executable
• tests directory has the test invocations
• test_data directory has the relevant invocations
• operational directory has the test logic in it

The test data in test_data directory comtains grouped interaction scripts, where fields are (all required):

Field	Description
Order	string
Execution	string if expected, None if not
Error	string if expected, None if not

The test logic is as following:

Test invocations start the ./bin/exchange_sim process usin the exchange_sim test fixture. The fixture also starts two background threads which transfer the exchange_sim output to buffered queues.

After that the invocations feedthe use cases into the batch processor, which is decorated with the check_process function. The check_process function is responsible for all process-related issues, like IOErrors, crashes, and final cleanup (partially).

The batch function feeds the orders to simulator via stdin and after each order checks an expected outcome - both in trades and error feeds.

As errors were not specified in the spec, the tests only check for a presence of an error. Later error detection can be improved, including pattern matching.

Other improvements may be to merge test invocations in one method - currently not done as pytest Does not have a "mark failed and continue" behavior, which means that the batch of tests will stop after the first failure.