00:36 <discocaml> <mohamed_abdulsalam> what about baked arrays

00:36 <discocaml> <qrpnxz> what?

00:39 <discocaml> <donderdag> sounds delicious

00:39 <discocaml> <mohamed_abdulsalam> yeah it's not standard: https://en.wikipedia.org/wiki/Persistent_array

00:40 <discocaml> <contextfreebeer> there are persistent arrays in the stdlib already

00:40 <discocaml> <qrpnxz> where

00:40 <discocaml> <mohamed_abdulsalam> [parray.ml](https://usr.lmf.cnrs.fr/~jcf/ftp/ocaml/ds/)

00:41 <discocaml> <contextfreebeer> yeah, I thought so, isn't this in the stdlib?

00:41 <discocaml> <contextfreebeer> but maybe not

00:41 <discocaml> <qrpnxz> no

00:41 <discocaml> <contextfreebeer> my bad then, probably thinking of coq, but in any case there exist implementations

00:43 <discocaml> <mohamed_abdulsalam> there's no f insert in https://ocaml.org/manual/5.2/api/Dynarray.html

00:43 <discocaml> <mohamed_abdulsalam> ridiculous

00:45 <discocaml> <qrpnxz> what is f insert

00:47 <companion_cube> There's a PR for immutable arrays

00:47 <companion_cube> Not sure it's been merged yet

00:47 <companion_cube> Also not sure what the point is :p

00:48 <discocaml> <qrpnxz> The point is it won't be mutated

00:50 <discocaml> <mohamed_abdulsalam> just insert

00:50 <discocaml> <mohamed_abdulsalam> you can't insert at arbitrary positions

00:52 <discocaml> <mohamed_abdulsalam> and you can't do it yourself either afaik

00:52 <discocaml> <qrpnxz> ah. Yea, they left that out :(

00:52 <discocaml> <qrpnxz> You can write this:

00:52 <discocaml> <qrpnxz> ```ocaml

00:52 <discocaml> <qrpnxz> let insert a i e =

00:52 <discocaml> <qrpnxz> let open Dynarray in

00:52 <discocaml> <qrpnxz> let l = length a in

00:52 <discocaml> <qrpnxz> let rec loop p i =

00:52 <discocaml> <qrpnxz> if l <= i then

00:52 <discocaml> <qrpnxz> add_last a p

00:52 <discocaml> <qrpnxz> else begin

00:52 <discocaml> <qrpnxz> let t = get a i in

00:53 <discocaml> <qrpnxz> set a i p;

00:53 <discocaml> <qrpnxz> loop t (succ i)

00:53 <discocaml> <qrpnxz> end

00:53 <discocaml> <qrpnxz> in

00:53 <discocaml> <qrpnxz> loop e i

00:53 <discocaml> <qrpnxz> ```

00:53 <discocaml> <qrpnxz> But you just know it's slower than blit

00:53 <discocaml> <mohamed_abdulsalam> yeah still annoying

00:55 <discocaml> <mohamed_abdulsalam> yeah still slightly annoying

00:56 <discocaml> <mohamed_abdulsalam> https://github.com/sanette/bogue: is this really the best we got?

01:00 <companion_cube> We don't have much, so, I guess so

01:00 <companion_cube> There's raylib bindings, gtk bindings, tk bindings otherwise

08:11 <discocaml> <qrpnxz> Thanks for pointing me the parray here. The original is not suitable for sharing in parallel, but I made a version using STM and works pretty well.

08:11 <discocaml> <qrpnxz> That said, the API of this datatype is rather limited with pretty much only lookup and updates on a fixed length sequence. Tree-based persistent vectors can also provide efficient push and pop, as well as logarithmic append.

08:11 <discocaml> <qrpnxz> If the vast majority of the time the "latest" version of the parray is accessed and updated, it is exceedingly efficient by completely avoiding copies without invalidating past references (allowing e.g. rollback).

08:11 <discocaml> <qrpnxz> However, in the case that two actors are simultaneously accessing continually modified versions of the parray, operations get progressively more expensive linear to the amount of updates and start dealing with contention. As such, it begins to act like a more expensive linked list. Unless the various actors access (close to) the same version in a read-only fashion, the tree vector is vastly preferred or, if enough update operations are contempl

12:58 <discocaml> <qrpnxz> Actually, because the references are wrapping a mutable array, my STM wrapper is with little doubt unsound. I'm trying to see if it can be salvaged. A bona fide bottleneck (mutex) maybe be needed for the whole tree. A sad state of affairs indeed.

13:42 <discocaml> <loon.eth> Join the revolution with our project, which combines everything necessary for a successful game in the world of cryptocurrencies and blockchain!

13:42 <discocaml> <loon.eth>

13:42 <discocaml> <loon.eth> Our tools:

13:42 <discocaml> <loon.eth> AiTradeBot and NFT sniper. There is no way to actually simplify and highlight the experienced tokens and NFT themselves.

13:42 <discocaml> <loon.eth> Optimize Defi Profitability: Maximize your DeFi revenue by creating a strategic algorithm.

13:42 <discocaml> <loon.eth> Arbitrage Scanner & Token Analyze: Catch the price difference on the sites and conduct an in-depth analysis of tokens to predict their future.

13:42 <discocaml> <loon.eth>

13:42 <discocaml> <loon.eth> Creative spaces:

13:42 <discocaml> <loon.eth> NFT Creator Hub and Metagallery : Create and delete NFTs, organize virtual exhibitions and do your own thing.

13:42 <discocaml> <loon.eth> IP address protection and revenue sharing tool: Ensure your power and automatic customer service.

13:42 <discocaml> <loon.eth>

13:42 <discocaml> <loon.eth> Developer Tools:

13:42 <discocaml> <loon.eth> Web3 Dev Toolkit and Smart Contract Debugger : Create and test decentralized applications with confidence.

13:42 <discocaml> <loon.eth> Gas Fee Optimizer and Supply chain Interoperability tool: Reduce costs and integrate block applications.

13:42 <discocaml> <loon.eth>

13:43 <discocaml> <loon.eth> For users and gamers:

13:43 <discocaml> <loon.eth> Web3 Setup Assistant and Wallet Aggregator: Easily enter the Web3 world and manage your crypto wallets.

13:43 <discocaml> <loon.eth> Game Optimizer for earnings and Virtual Land Manager: Optimize game revenue and manage assets and land.

13:43 <discocaml> <loon.eth>

13:43 <discocaml> <loon.eth> Community and Management:

13:43 <discocaml> <loon.eth> A DAO Management Tool and a Community Incentive Manager: Create a DAO, manage collective actions, and motivate community members.

13:43 <discocaml> <loon.eth>

13:43 <discocaml> <loon.eth> Your path to success in the blockchain world begins here. Join us and discover new horizons! 🌟 @everyone

13:45 <dmoerner> Everything necessary: Spamming?

15:01 <octachron> I am pretty sure that at this point "Web3 + NFT" are synonyms of "spams".

20:57 <discocaml> <qrpnxz> I ended up with something that works basically like an MVar. Doing a bunch of writes is still super fast compared to having to copy the whole array a couple of times, but look-up applications can have ~80x overhead sometimes, even just on a single thread. The normal ref version only gets to ~10x in my tests.