Open AbrJA opened 8 months ago
Hi, thank you. I am glad it works for you.
The core idea of the new release is to make possible other kinds of applications. It removes some global context and simplify the explicit usage of the resources and configurations.
The deprecation file https://github.com/sadit/SimilaritySearch.jl/blob/main/src/deprecated.jl only lists some high-level functions that do not use a search context because there are identical functions now accepting the context explicitly.
In previous releases, these contexts were implicit (they use global shared variables) and are now explicitly passed to each function.
With this search context, generating isolated indexes using the same Julia process is more straightforward. They also now control logging, so each index instance can have a different log (e.g., storing in a log file) or use some indexing events (append items) to save index data incrementally.
Distance functions remained untouched, and I have no plans to remove or deprecate them. Maybe some automated tooling shows that they are in deprecation.
Best, Eric
I found my self divagating on SimilaritySearch 0.11 instead of LoopVectorization deprecation for Julia 1.11
https://discourse.julialang.org/t/why-is-loopvectorization-deprecated/109547
I am trying to figure out what to do...
Best, Eric
Hi @sadit, thank you for your replies
I had read that post and it's a shame, LoopVectorization is a great package and widely used. I will continue looking for another option or approach if I find something I will share it with you!
Cheers,
Abraham
Hi there,
Thank you for the package you have done, it has so much potential!
I noticed that one of the dependencies (LoopVectorization) will possibly be deprecated on Julia 1.11. This will have implications on the TurboDistances you have created.
If it happens (I hope not), solving this without affecting performance would be great!