← Back to context

Comment by citizenpaul

4 days ago

The only thing I've found that LLM speeds up my work is a sort of advanced find replace.

A prompt like " I want to make this change in the code where any logic deals with XXX. To be/do XXX instead/additionally/somelogicchange/whatever"

It has been pretty decent at these types of changes and saves time of poking though and finding all the places I would have updated manually in a way that find/replace never could. Though I've never tried this on a huge code base.

16 comments

citizenpaul

Reply
zahlman  4 days ago

> A prompt like " I want to make this change in the code where any logic deals with XXX. To be/do XXX instead/additionally/somelogicchange/whatever"

If I reached a point where I would find this helpful, I would take this as a sign that I have structured the code wrongly.

  • baq  3 days ago

    You would be right about the code but probably wrong about the you. I’ve done such requests to clean up code written over the years by dozens of other people copying patterns around because ship was king… until it wasn’t. (They worked quite well, btw.)

    • zahlman  3 days ago

      If you're generalizing "you" to all developers, perhaps.

      I know how I'd respond based on personal experience.

  • rtpg  3 days ago

    sometimes you want a cutpoint for a refactor and only that refactor. And turns out that there is no nice abstraction that is useful beyond that refactor.

  • citizenpaul  3 days ago

    I knew someone would make this comment. I almost added a "I'm probably not leet enough to avoid these situations" disclaimer. It seemed a bit pointlessly self deprecating.

    You don't always get to choose the state of or the way a system you work in/with is designed. In this case I was working in a limited scripting language that I have no choice about.

    Keep that nose turned up. I'm sure you are leet10xninja. Maybe work on your reading comprehension before you dump on someone though as I already specified that I greatly simplified for comment sake.

    • zahlman  18 hours ago

      No slight was intended. I've learned a lot of techniques; I don't consider this a matter of elitism. I have generally been fortunate to have control over the architecture of my projects; when I encounter something like this in someone else's code, I can at least raise my concern. Sometimes the code is that way because doing it the other obvious way would lead to some other inconvenience. Hence "a sign", not proof. It's worth investigating signs.

skydhash  4 days ago

I supposed you haven’t tried emacs grep mode or vim quickfix? If the change is mechanical, you create a macro and be done in seconds. If it’s not, you still got the high level overview and quick navigation.

  • kfajdsl  4 days ago

    Finding and jumping to all the places is usually easy, but non trivial changes often require some understanding of the code beyond just line based regex replace. I could probably spend some time recording a macro that handles all the edge cases, or use some kind of AST based search and replace, but cursor agent does it just fine in the background.

    • skydhash  4 days ago

      Code structure is simple. Semantics is where it get tough. So if you have a good understanding of the code (and even when you don't), the overview you get from one of those tools (and the added interactivity) is nice for confirming (understanding) the needed actions that needs to be done.

      > cursor agent does it just fine in the background

      That's for a very broad definition of fine. And you still need to review the diff and check the surrounding context of each chunk. I don't see the improvement in metrics like productivity and cognitive load. Especially if you need to do serveral rounds.

      3 replies →

  • citizenpaul  3 days ago

    I'm decent at that kind of stuff. However thats not really what I'm talking about. For instance today I needed two logic flows. One for data flowing in one direction. Then a basically but not quite reversed version of the same logic for when the data comes back. I was able to write the first version then tell the LLM

    "Now duplicate this code but invert the logic for data flowing in the opposite direction."

    I'm simplifying this whole example obviously but that was the basic task I was working on. It was able to spit out in a few seconds what would have taken me probably more than an hour and at least one tedium headache break. I'm not aware of any pre LLM way to do something like that.

    Or a little while back I was implementing a basic login/auth for a website. I was experimenting with high output token LLM's (i'm not sure that's the technical term) and asked it to make a very comprehensive login handler. I had to stop it somewhere in the triple digits of cases and functions. Perhaps not a great "pro" example of LLM but even though it was a hilariously over complex setup it did give me some ideas I hadn't thought about. I didn't use any of the code though.

    Its far from the magic LLM sellers want us to believe but it can save time same as various emac/vim tricks can to devs that want to learn them.