So I just got Claude Code set up but honestly I have no clue what I am doing. I have this massive pile of legacy Java code for a tiny nonprofit project I'm volunteering for and it needs to be refactored by next Friday. I keep hearing about skills or specific ways to prompt it for big codebases but I'm such a newbie at this.
My logic was just to let it read everything and pray but that feels wrong. Should I be focusing on specific refactoring patterns or is there a way to tell it to just handle the old Spring stuff first? Sorry if this is basic but I'm totally drowning in these old files...
Honestly, been there. Legacy Java is basically a digital archeology project. I've tried many ways to automate this over the years, and jumping in blind usually breaks everything. Trying to meet a Friday deadline with a massive codebase is stressful, but you can definitely pull it off if you stop praying and start scoping. Here is what I would do based on my experience:
Totally agree! To save money, use Anthropic Claude 3.5 Sonnet API over OpenAI GPT-4o API.
tried dumping a whole monolith into the context window last month and honestly it was a total mess. wasn't as good as i expected... it just got confused by the old XML beans. Heres what actually helped me:
Bookmarked, thanks!