It'd be nice to have a caching system for output of the providers; this might speed up re-running the spider after changes in page objects or callbacks.
The main issue is how to compute the cache key, as kwargs might have different semantics for different providers.
So, one option is to have some interface for providers which would allow them to tune how to do caching.
Another option is to do nothing, and handle it on the provider level, or on lower levels.
This is for discussion.
It'd be nice to have a caching system for output of the providers; this might speed up re-running the spider after changes in page objects or callbacks.
It seems this is the place where cache could happen: https://github.com/scrapinghub/scrapy-poet/blob/f7ad036f62699513e10d5749910042a5268153fc/scrapy_poet/injection.py#L152
The main issue is how to compute the cache key, as kwargs might have different semantics for different providers. So, one option is to have some interface for providers which would allow them to tune how to do caching. Another option is to do nothing, and handle it on the provider level, or on lower levels.