Is there a way for me to program a hashtable that can intake multiple values of the same key?

[Vizonex]

Hi, I wanted to create this issue because I'm struggling to yank out the RawTable structure code so I can use it for a different purpose.
You see I'm a maintainer of the multidict project for python and for years we have used C as it's backbone to speedup it's slower sections but as time went on, multiple memory related bugs have been found and it's safety has been wildly questionable. Currently I'm trying to experiment with the possibility of programming a rewrite for the project in rust that can take in multiple of the same key as long as they are all the same key without using values that store a vector for example: <key, Vec<values>> which I see as slow and unreliable. I am a bit newer to rust so sorry if I'm having trouble understanding certain concepts. I'm just trying to gain access to the lower code code so I can find ways to save the maintainers a bit of unwanted maintenance costs if that makes any sense to anybody.

[cuviper]

This should be possible with HashTable, using insert_unique (even though it's not) and iter_hash/iter_hash_mut filtered by item equality to do lookups. It may not perform well in some cases though, especially if you remove items and cause a lot of tombstones.

Can you share more about what you've tried and where you're getting stuck?

[clarfonthey]

It's also worth pointing out that you could probably pretty easily get away with using something similar to indexmap that only does two allocations; just make a sort of "linked list" using list indices so that the hash table stores the index of the first item, but you can get more. This is probably the best way to handle multiple insertion in the way Python does it as well, since Python retains insertion order.

[cuviper]

I think you probably meant to reference indexmap, like here:

https://github.com/indexmap-rs/indexmap/blob/main/src%2Finner.rs#L29