It turned out that the fix applied last week on the backend relating to items not being properly decremented upon removal was incomplete, and there were still edge cases. Thankfully, this was uncovered during further testing and integration this week, way in time before the final demo. This bug has now been fixed (and verified through the same rigorous testing process) through a complete rewrite of the item quantity computation logic. In short, instead of returning a single result after one pass through the transactions, it now makes 2 passes in order to properly account for item returns. In terms of time complexity, both approaches have a complexity of O(n), so there should be no significant difference in backend latency as a result of the change. In exchange, not only is the bug fixed, but the logic and code is also much more clearer and readable, significantly helping in any future efforts to expand the backend.
I have also made the backend more robust against incorrect quantity data, e.g. “removing” more apples than there are in the fridge. In essence, the backend now computes as if there is a floor of 0, instead of going into negative numbers, which is not logical. We passed over this aspect of the project initially as we prioritized producing new features that are suitable for the demo, but the time is now right for this change to be implemented as the product matures.
My main priority this weekend is to produce and complete the final presentation for our project. I recorded several video demos that I intend to incorporate into the slides, and collected a series of performance and benchmarking data to compare against our use case requirements (good news: we’re meeting them!)
Coming up, the largest priority is to complete the wrap up of the project as the semester comes to a close, in the form of the final presentation, poster, and final demo. We also intend to expand user friendliness and productiveness even further by allowing the user to select multiple fruits in one confirmation, based off our discovery that the CV is capable to predicting all 3 different types of fruits when all 3 of them are placed on the platform.
We’re comfortably ahead of schedule! The project including the hardware is now all in place, integration tests are nearly complete, and the semester is coming to a close. This will mark a great end to our CMU careers!