@regehr "find the seams" is much harder advice to execute on when dealing with code that may not even have seams because it was not constructed with any intent but rather is a gooey blob of gunk secreted from a statistics engine
glhf needleworker, you shit your bed now you can lie in it
The more shit changes, the more it stays the same.
https://archive.org/details/working-effectively-with-legacy-code
@kstatz12 @regehr that always made me laugh - I spent two days tracking down all the places in the legacy code that had to be changed to accommodate this new feature, refactoring half a dozen near identical versions that should have been a single function, coding up the feature, modifying the UX and adding an API call or two. And you can say LGTM in under three minutes.
The sad thing is I understand. My request to review blew away your twenty minutes mental preparation to get into the mind space to do your own work.
I'm increasingly of the opinion that if AI is used, it should be to review code changes, never to write code. Leave the creating to the humans, offload the tedious unwanted but necessary checking to the machines.
@theorangetheme @rayckeith @regehr feels a lot like the COBOL programmers who commanded premium rates in the lead up to Y2K, doesn't it?
"Cleanup on aisle Spagetti."
John, are we there yet?
@regehr Aside from *gestures vaguely*... as a lead engineer I'm _also_ lost staring at hundreds of lines of Python.
So much of my work is about achieving more, in smaller chunks, with less code, in a way humans can understand.
