https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127690
--- Comment #5 from Oleksandr Kulkov <adamant.pwn at gmail dot com> --- Hi Jonathan, thanks for keeping me CC’d! I still think pb_ds is a great concept and would be sorry to see it go, but I understand your frustration with maintaining it with an unclear target audience, and no dedicated maintainers particularly interested in it. I can go over the issues you flagged with an AI and try resolving them to the best of my ability, but GCC's current AI policy would likely force me to spend way more time doing this than I can currently afford on a volunteer project. Putting it on GitHub would make it easier to contribute to, but at the same time it would completely erase the main reason the audience I'm familiar with (competitive programmers) cares about it in the first place (out-of-the-box availability on online judges). I remember that it's not a particularly interesting use case to you, which is fair, but it's the only use case I know, so I'm afraid it's likely the project will die completely if it's put aside. > If there are standard library replacements for the pb_ds functionality > I suggest to retire pb_ds, possibly by first making its install opt-in Not for all of it. One especially useful feature of pb_ds is tree_order_statistics_node_update, which adds order_of_key() and find_by_order() functions to the regular std::set. It is not covered by the standard library, and it probably can't be, because it would require to keep subtree size metadata in std::set, which would significantly increase its size. Other parts of pb_ds may not have standard library coverage either, but I can't readily tell how significant they are. I saw a blog that gp_hash_table performs "orders of magnitude" better than std::unordered_set: https://codeforces.com/blog/entry/60737. At the same time, there are also parts of pb_ds that are much less performant than they could be, which also means that if it's made a standalone project, it might not be able to compete with state of the art data structures, unless someone puts considerable effort making it so (the only way I see it gaining traction as a standalone repo).
