uthash VS smol_world

You should check out uthash, which has an extensive test suite and documentation (the user guide), to see how a library like this one that's used in the real world is written, tested, and used: https://github.com/troydhanson/uthash/tree/master/tests

Uthash – C macros for hash tables and more\ (16 comments)

Thank you for this article.

I have been working on trying to design a contiguous hash table data structure that has these two mutual requirements:

a) allows deep clones (due to contiguous memory copyable with memcpy)

b) supports arbitrarily nestable hashmaps

The closest library I know about is smolworld ( https://github.com/snej/smol_world ) but I don't know how easy it can be cloned.

These requirements rules out multiple mallocs: I do one single malloc and expect that to be enough for the entire hashmap.

I'm not sure if these constraints might also force a fixed number of buckets and a fixed capacity.

The clonable property requires interior mutability be limited, because if you memcpy pointers they would cause structure sharing, that I'm trying to avoid, hence a deep clone.

Rationale: these are the use cases I have for such a data structure. The first is cheap copy on write. I also have a left-right concurrency hashmap, but I feel the properties of this data structure are even better. The first is a sharding in multithreading design: do a cheap memcpy for each thread and shard the processing and merge at the end, without any synchronization cost while the threads are working. I know that deep cloning is slow in Java if you do it with loops rather than memcpy. Another use case is efficient serialization.