grenad VS bumpalo

I tried new inserting algorithms on the sorter to reduce the amount of time spent sorting but didn’t find anything better than that, would you have any idea? I gave up trying better inserting algorithms by declaring that doing one final in-memory sort was more effective that doing a lot of comparison to try inserting keys in the right order…

I have finalized the latest release of grenad which introduce a lookup struct called a Cursor. Grenad is a library that provides tools to sort, merge, write, and read immutable key-value pairs.

I worked on my grenad library: a library used to sort, merge, write, and read key-value pairs, used in MeiliSearch for the new indexing system. I added a lot of documentation to the crate and done a lot of improvement by reducing the amount of copied memory when merging key-value pairs.

I was wondering if you could maybe point me to any blog post regarding the segment merging? I am using my grenad library, which is a simplification of my MTBL Rust port. This is a key-value store with some useful sorter helping functions, using a merge function, some compression parameters... Works great but could maybe be improved.

I also wanted to develop an immutable key-value store with some interesting properties like compression and multi-database (like for LMDB), it is much easier to develop than LMDB, I just need to base this on something like my grenad lib plus a wrapper to support multi-databases.

Long story short, heap allocation is painfully slow. Any sort of malloc will always be slower than a custom pool or a bump allocator, because it has a lot more context to deal with.

Rust makes it especially hard to use custom allocators, see bumpalo for example [0]. To be fair, progress is being made in this area [1].

Theoretically one can use a "handle table" as a replacement for pools, you can find relevant discussion at [2].

[0] https://github.com/fitzgen/bumpalo

There are ways to accomplish this as well. Different allocator libraries exist for this kind of scenario, namely bumpallo which allocates a larger block of memory from the kernel, and allocates quickly thereafter. That would amortize the cost of memory allocations in the way I think you're after?

Meaning, storing a lot of things in the same block of allocated memory? Vec is a thing, you know. There's also a bump allocator library.

There are data structures like slotmap, and relatively low-level crates like bumpalo. This is not to say that either fits your use case, just that you definitely have access to the necessary parts to fit what you describe.

Sometimes you can put everything in a bump allocator, then when you're done, free the entire bump allocator in one go. https://docs.rs/bumpalo/

This is useful with crates like bumpalo which give you bump-allocation arenas whose lifetimes are tied to the objects they allocate.

TSC doesn't need to "stick around", right? Just a run-once and the program is over?

In those cases, https://github.com/fitzgen/bumpalo works amazingly as an arena. You can pretty much forget about reference counting and have direct references everywhere in your graph. The disadvantage is that it's hard to modify your tree without leaving memory around.

We use it extensively in http://github.com/dioxusLabs/dioxus and don't need to worry about Rc anywhere in the graph/diffing code.

Probably bumpalo, but then its Box will have a lifetime parameter - bumpalo::boxed::Box<'a, dyn MyTrait>

Strictly speaking, Rust doesn't need this as a built-in language feature, because its design allows it to be implemented as a third-party library: https://docs.rs/bumpalo

The biggest problem is that there's some awkwardness around RAII; I'm not sure whether that could have been avoided with a different approach.

Of course, ideally you'd want it to be compatible with the standard-library APIs that allocate. This is implemented, but is not yet at the point where they're sure they won't want to make backwards-incompatible changes to it, so you can only use it on nightly. https://doc.rust-lang.org/stable/std/alloc/trait.Allocator.h...

Or are you suggesting that the choice of allocator should be dynamically scoped, so that allocations that occur while the bump allocator is alive automatically use it even if they're in code that doesn't know about it? I think it's not possible for that to be memory-safe; all allocations using the bump allocator need to know about its lifetime, so that they can be sure not to outlive it, which would cause use-after-free bugs. I'm assuming that Odin just makes the programmer responsible for this, and if they get it wrong then memory corruption might occur; for a memory-safe language like Rust, that's not acceptable.