Fredrik Kihlander

I'm still not convinced that I want to support being able to hook in any "vector like" type into dl. For example, would that force me to implement constructors/destructors on...

One reason why I am skeptical to adding support for "your own array type" is that I have seen, in a system really similar to this, this kind of thing...

* How would the user know if it need to free the array or if it lives in the initial loaded chunk. Would it always malloc the arrays? * How...

As I wrote above I think we could make all array:s in fl dynamic on request. We could reserve a bit in count or lowest bit in pointer to indicate...

Where free_all() would keep an array of allocated ptr:s that are still alive.

All of this can also be built on top of dl but maybe the API becomes nicer to use if it's built in, not sure.

The biggest issue, I guess, with your above approach is how you plan to handle sub-pointers. I.e. arrays in structs stored in arrays.

I absolutely see your point and your predicament but I can't seem to figure out a way to do this in a good way. There is just too many "valid"...

Honestly I do not see how a dynarray special case would work either. In the end it would be 10+ special cases :(

I'm personally not that big of a python user these days, and by that not that knowledgeable about pip and what is needed for that. pymmh3 was designed as a...