06Feb

How do you know what to do when you’re down in the weeds?

Posted by Elf Sternberg as Rust

As I’ve been reading Rust code and learning how to write it, I’ve become frustrated with one detail that’s been driving me a little crazy from time to time.

How do other developers know when some side-issue management is necessary? In my specific case, it’s about Drop.

In two different contexts, I’ve now seen highly specialized Drop implementations that, quite honestly, I don’t understand how the developer knew they were going to be necessary.

The first is in Beingessner’s Too Many Lists. It’s obvious, in hindsight, that a list with ten thousand entries would fit just fine on the heap, but the stack would get blown up by the recursive drop, so an iterative, hand-written drop was necessary. Which is great, but how did Beingessner know it was necessary without experiencing the blowup? I’ve written code for decades, much of it in C or C++, and I’ve never experienced a circumstance where I had to care about blowing the stack with a recursive unroll. Maybe I’ve never worked with tens of thousands of objects, but that seems unlikely given that I work in a big data field.

The second is in Jerome Froe’s LRU Cache. I was writing an LRU cache for Rust with some specific needs (basically, my ejection rule was based on an arbitrary scale provided by an object with a functor, whereas Jerome’s uses a standard volume-based rule. I struggled for days with my drop rule until I finally cheated and looked at his and… I don’t get it. I mean I get what it is and how it works, but I’m still bothered by… how did he know how to do that? The mem::forget trick is nifty: those head and tail values point to cache items that are going to be dropped automatically, so we need to not drop them. But I still don’t know how Froe knew that.

I know it seems… I dunno, petty, I guess. But I always feel as if there’s a detail missing that I’m not getting, which is really odd because I’m considered a “thorough” developer according to my peers. But maybe only within a safe space, and Rust these days is a very new, and very awkward, territory for me to be crossing.

Comment Form

Subscribe to Feed

Categories

Calendar

February 2018
M T W T F S S
« Jan   Mar »
 1234
567891011
12131415161718
19202122232425
262728