Fun exercise! I found that the Ruby in this post resembles modern C# (probably because Ruby is a gold standard in readability, and in the last 5 years or so C# has been making big strides in readability). I've recreated the Ruby code mostly line-for-line in C# and posted it here: https://gist.github.com/waf/5c6a04899e8250cb9a89406b978c9bcc
It mapped from Ruby to C# really well. The only concept I had to add was a `BoardCell` base type for the `Mine` and `Empty` types so we could return either `Mine` or `Empty` in a typesafe way. Everything else matched conceptually 1-1.
It's 118 lines total, with 18 lines of whitespace, so 100 lines exactly. I figure the original Ruby code was also not counting whitespace so it's a fair comparison. I tried to not use any "code golf tricks" that would minimize line count. It's a full program; it could be copy/pasted into a modern .NET 8 project and run without any additional setup.
Note, I posted this same comment on the original blog post, but I've posted it here too as it might be interesting for HN folks.
I've been adding C# code completion functionality to my REPL tool, and ended up reverting to the text-davinci model.
The codex (discontinued?) and text-davinci models gave much better results than GPT3.5-turbo, specifically for code completion scenarios. The latest models seem to produce invalid code, mostly having trouble at the boundaries where they start the completion.
My suspicion is that these latter models focus more on conversation semantics than code completion, and completing code "conversationally" vs completing code in a syntactically valid way has differences.
For example, if the last line of code to be completed is a comment, the model will happily continue to write code on the same line as the comment. Not an issue in a conversation model as there is a natural break in a conversation, but when integrating with tooling it's challenging.
Most likely the issue is that I'm not yet effective at prompt engineering, but I had no issues iterating on prompts for the earlier models. I'm loving the DaVinci model and it's working really well -- I just hope it's not discontinued too soon in favor of later models.
It mapped from Ruby to C# really well. The only concept I had to add was a `BoardCell` base type for the `Mine` and `Empty` types so we could return either `Mine` or `Empty` in a typesafe way. Everything else matched conceptually 1-1.
It's 118 lines total, with 18 lines of whitespace, so 100 lines exactly. I figure the original Ruby code was also not counting whitespace so it's a fair comparison. I tried to not use any "code golf tricks" that would minimize line count. It's a full program; it could be copy/pasted into a modern .NET 8 project and run without any additional setup.
Note, I posted this same comment on the original blog post, but I've posted it here too as it might be interesting for HN folks.