sortedcontainers VS deno_std

If you like Python, the library sortedcontainers as a clear, well documented, yet short source code that is a joy to read for a non trivial problem:

https://github.com/grantjenks/python-sortedcontainers/blob/m...

See for yourself. Looks like sortedset uses sortedlist under the hood, which itself uses a list of lists under the hood.

You could use http://www.grantjenks.com/docs/sortedcontainers/ instead!

Even the Gentoo package repository manages fine with a two-level hierarchy. There's also a Python library, sortedcontainers, that suggests two-level trees are pretty good at any reasonable human-scale (and beyond), even while fixed-arity trees are asymptotically optimal.

It's a good question and to be accurate, depending on the benchmark, Mongita is about the same speed at SQLite to several-times slower.

There is less happening algorithmically than you would think. Where the tricky slow bits do exist, they have largely fallen into the happy-path of fast data structures in the Python language/stdlib. I also use sortedcontainers for indexes which helped quite a bit (http://www.grantjenks.com/docs/sortedcontainers/).

If you're curious, the benchmark code is in the repo: https://github.com/scottrogowski/mongita/blob/master/benchma...

I’d like to add sortedcontainers. I use it all the time. It basically does what it says on the tin. Other than the SortedList, the fact that the container is sorted only comes into play when you iterate over it or perform a bisect left/right.

For reference (for the issues you mentioned): 1. This issue was opened almost immediately to solve the weird .only function not working https://github.com/denoland/deno_std/issues/2979 2. That looks weird to me, will get back to you on this one since it should work I think 3. Generally polluting the global namespace isn't great, but because we're only polluting the namespace of a module (and we choose what parts to import), I personally find it quite freeing. I entirely understand how that might feel awkward. 4. you CAN specifying only writing to certain directories! --allow-write=/path/to/dir would allow that!

At least for the ones related to trees, it's just a renaming. Below is a link to the PR. When I initially implemented these trees, I chose the names BSTree and RBTree to keep the names short. I'm guessing the person that proposed renaming them did so to make it more obvious what they are.

https://github.com/denoland/deno_std/pull/2400

The standard library is separate from the runtime. It wouldn't break backward compatibility if you were to update. For example, if you were importing RBTree and upgraded Deno to the latest release, it would keep working just fine. You would only really need to switch to using RedBlackTree instead if there was a change made to it that you wanted.

I think the only time you would need to update your standard module imports to be able to use newer versions of the Deno runtime if the standard module were depending on runtime APIs that have a breaking change.

It is like Demo deprecating fs.exists().[1]

[1]https://github.com/denoland/deno_std/discussions/2102

This was fun to read through.

I would need to profile the code, but the startup time being bad for Deno seems like maybe a combination of the code in here being unoptimized:

https://github.com/denoland/deno_std/blob/0ce558fec1a1beeda3...

(Ex. Lots of temporaries)

And usage of the readFileSync+TextDecoder API instead of readTextFile (which is also a docs issue since it's suggests the first one). It seems the code loads the 100MB into memory, then converts to another 100MB of utf8, then parses with that inefficient csv decoder. The rust and go versions look to be doing stream/incremental processing instead.

But it there's some talk to reconsider it

Sorry to hear that. I want to provide expect API in deno_std in the future: https://github.com/denoland/deno_std/issues/1779

I just migrated the module to Deno and rewrote the test cases using the Deno test runner. Also contributed a bug fix to the test runner that I encountered during the migration. An npm version is still available and automatically generated from the Deno code via a small bash script (rewriting imports, adding an index.ts, etc.).

Much of Node.js is written in C, yet it's still called Node.js.

Deno has some JavaScript/TypeScript in it. On GitHub https://github.com/denoland/deno is 22.8% JavaScript and 13.2% TypeScript, and https://github.com/denoland/deno_std is 68.2% JavaScript and 31.6% TypeScript.

So to me it's misleading about the name, but not about what Deno is written in.