Film

Goal

• Reflect upon decisions made on my last project Films (submitted in July of 2019) and make improvements on my findings.
• Be more pragmatic with my approach to solving problems in code.
• Spend less than 24 hours designing and implementing (18 hours spent in total).

Reflective findings when revisiting the previous project

• It was an overly complicated project which made the codebase extremely hard to navigate.
• The networking logic would make redundant calls whenever a cell came into view as there was no caching logic for OMDB calls.
• The procedural image loading on the cell was buggy due to improper delegate management.
• The nesting of collection views within table view cells was extremely heavy and resulted in good, but not great, scroll performance.

Improvements made over previous project

• Reduced lines of code from 3622 to 1556 (57% decrease).
  • Project migrated from using Storyboard to now being purely programmatic.
  • Adjusting for autolayout, setup, and styling code which was previously managed in Storyboards (250 lines), the number of lines is further reduced to 1306 (65% decrease).
  • $ find . \( -iname \*.swift \) -exec wc -l '{}' \+
• Sorting logic has been replaced with a dynamic debounced search query which has a custom cache backing to avoid redundant filters.
• Cell dequeuing previously caused issues by not retaining scroll position offset which have since been fixed.
• The detail screen is more informative and slightly cleaner.
• Declared subdomains for HTTP payloads on the nflx and omdb endpoints as to not expose all traffic through the application.
• By utilizing systemColors, this new application supports both light and dark modes.

Cool discoveries

• UICollectionViewCompositionalLayout is awesome. I decided to take a swing at using this new API as it aligned nicely with the requirements with Discover screen design.
  • It only took 13 lines of code to implement the layout for the collection view. Yes, really.
• UICollectionViewDiffableDataSource is another awesome addition to the SDK.
  • Snapshots avoid index out of bound errors and eliminate the need for data source lookups (typically).
  • Out of the box diffing...sign me up! Diffable data source is definitely a keeper.

Discussion

• MVVM design pattern utilized as VIPER became a bit overkill with an application of this size.
• In-memory cache utilized at the service level to avoid reconstructing custom objects.
• Cache used to store searches to avoid re-indexing the entire response for duplicate queries.

Potential improvements

• Offline mode - Core Data could have been utilized to persist data to the disk; however, due to time constraints I ended up not implementing this logic.
• Better logging - A custom logger could have aided in identifying erroneous behaviors; however, considering the time frame I felt if was better if I omitted it.
• Error handling - Error handling is crucial to a great application; however, it is often the bulk of the development effort.
  • Errors in this project were handled by placeholders and fake data.