I'm curious what people are doing for memoization in #golang. I've looked around and haven't found great libraries for this which makes me wonder if I'm pursuing the wrong solution to a problem.
Caching the return values of functions based on the params has been useful to reduce load on downstream services, make things a bit faster on average and even add some level of consistency in functions that can be highly variable (which is an odd use case but nonethelass useful). But maybe there's a different pattern/idiom that's used in the Go ecosystem?
In applications where I’ve needed this, I’ve taken a manual approach. Structure the function to return a single Result struct (including an error field), develop a convention for mapping function inputs to a string, then add reads & writes to a
map[string]*Result
which allows me to return cached values as a shortcut. No idea if it’s the most efficient way, I’ve not actually considered finding a package to handle the process.Edit: For more advanced behavior, what are your thoughts on the official memoize package?
I suppose I should be clearer on the features I want. I'd want to be able to store my cache in memcached or redis and I want the cached data to expire. So for one call, I might want to keep it for five minutes, but another one can stick around for 24 hours.
The memorize package falls down there.
Pet peeve: don't use string keys. It invites key collision errors. Use the fact Go supports structs as keys in maps. Safer and more efficient.