bob poekert@bobpoekert·Feb 20people like unit tests because they're an excuse to do work that you don't have to think too hard about, but they give you no guarantees whatsoever1331314
bob poekert@bobpoekert·Feb 20integration and generative testing get closer to giving you guarantees and require writing way less code but they require actually understanding your problem domain1590
bob poekert@bobpoekert·Feb 20really what you want is to sample from a statistical estimator fit to real world log data but afaik nobody does this349
bob poekert@bobpoekert·Feb 20also unit testing frameworks give you a coverage number so you can have number go up in meetings350
bob poekert@bobpoekert·Feb 20you can get number go up doing more rigorous modeling, but programmers aren't used to thinking in terms of state spaces and transfer functions1130
bob poekert@bobpoekert·Feb 20your sytem model that you test against has an error (some loss function, mse or something) and your system has a failure rate wrt the model116
bob poekert@bobpoekert·Feb 20maybe if we call this "adversarial testing" or something people will think it's ~AI~ and want to do it3430
Tom Schouten@tom_zwizwaReplying to @bobpoekertsomething like support vector machines: find the tests that actually matter9:15 PM · Feb 20, 2022·Twitter Web App