Threading issues with requests_cache

[amotl]

For my part, I think that it is unlikely that Niquests caused the seg fault. I made sure that Niquests works under the freethreaded build. On the other hand, there is multiple reports of threading issue with requests_cache.
Unless we can have a minimal reproducer, I'd say Niquests is safe.

Originally posted by @Ousret in #179 (comment)

[amotl]

Hi @Ousret. Thanks for your feedback here. I do have a bunch of questions, maybe you can answer some of them?

• Can you refer to some of those threading issues with requests_cache, so we could investigate them?
• Do you have another recommendation instead of using this tandem; just do the caching on the application level instead, using traditional Python's Beaker or Dogpile Cache libraries?
• Do you know about any other suitable integrated caching libraries for Requests or Niquests?

[Ousret]

The main issue I think is requests-cache/requests-cache#1008 or more widely https://github.com/requests-cache/requests-cache/issues?q=is%3Aissue%20state%3Aclosed%20thread%20issue

two solutions emerge:

• contribute to requests_cache to ensure proper thread safety (seems a new alpha release tried to address that, worth a try)
• drop it and build something straight forward and simple, maybe related to filelock (but we should consider an abstract interface, so that filelock can be replaced with something else, like redis for example)

regards,

[amotl]

Thanks for your swift reply, @Ousret. We will try requests-cache 1.3.0a1 first, thank you for pointing this out.