reFetch extension for maxAge functionality about memoizee HOT 9 CLOSED

I've proposed a bit different thing. Not autorefresh, but background load after request. For examlpe:

• i like to cache data for 60 seconds.
• if next data requested between 60-70 seconts - cache is almost fine, use it and load new data in background.
• if data not requested for a long time - it's completely expired, and loaded in ordinary way, via cache proxy.

No needs to reload each cache slot, only used ones. Tought, in my use cases, all data will be reloaded anyway. So, no principle difference, wich method will be used.

from memoizee.

@puzrin Thanks for clarification. I get the point, only if data is accessed again and close to expiry date, it's wise to refresh it in background and push forward expiry date (so likely future access won't have to wait).

Indeed if data is requested frequently it's beneficial, and it's more optimized approach than refreshing any cached value when it's expired, very good point, I'll make reFetch to work exactly that way.

The only difference to what you propose is that I would probably do re-fetch on calls made some time before expiry date rather than still serving cached value to those that were slightly later, I think it would be more what you expect. Serving value obtained 10 seconds ago when e.g. maxAge is set to 8 seconds is controversial. Let me know if I still miss something.

from memoizee.

Yeah, -(10-20%) before expire will make the same effect, as +(10-20%) after expire (in real life). No principle difference, but pre-fetching looks more logical, than post-loading.

Probably, this interval could be tuned via options, but i guess, 10% default will be "good for all".

from memoizee.

Thank you! It landed in v0.2.2. I used -33% as default, but it's customizable

from memoizee.

Thanks!

from memoizee.

I looked code a bit. Didn't understood, how do you resolve race conditions? If cache is about to expire, and has multiple async gets, will they cause multiple refresh requests? The same question for start state, when cache is empty.

from memoizee.

@puzrin async logic is separated from maxAge/preFetch logic.

When we're working with asynchronous function, cache hit occurs here: https://github.com/medikoo/memoize/blob/master/lib/ext/async.js#L37

preFetch logic that occurs on cache hit is here:
https://github.com/medikoo/memoize/blob/master/lib/ext/max-age.js#L48
Race condition is prevented by L49 and L51. There's no way that value would be reFetched multiple times.

Another call to memoized function at L55 invokes https://github.com/medikoo/memoize/blob/master/lib/ext/max-age.js#L23-27 again, so preFetch configuration is reinitialized then.

from memoizee.

Got it, thanks. I was not sure about preFetchCache intention. Probably, it worth to rename it to preFetchLock or similar, to reduce possible confusion. I fact, that's not cache, but singleton locks.

from memoizee.

Good hint, actually preFetchCache holds either timeout id or race condition lock, but still you're right it deservers a better name. I'll fix that. Thanks!

from memoizee.