Case insensitive keys

[mrxrsd]

Is there any option to make cache keys case-insensitive? I did it here using a decorator, but I was wondering if there is a better way to go, if there is any interceptor, or if I can do it through a plugin?

tks

[jodydonetti]

Hi @mrxrsd , nothing so far, nobody asked for this as of today, but I can see if something can be done.

I did it here

Is there some code that should've been posted as an example?

Questions: are you interested in L1 only or L1+L2?

Thanks

[mrxrsd]

Hi Jody,

Sorry if I wasn't clear, I didn't intend to send any code. I meant that I implemented it here in my project using the decorator pattern, but I was wondering if there was a more 'native' way to do it instead of using a pattern.

By the way, I saw this other thread, which in the end was something similar to what I did here, but here I created a singleton decorator that does a toLower on the keys before calling the fusion cache.

#360

My case is simpler because I don't rely much on the context. If I had a key transformer Func<original_key, transformed_key>, it would solve the problem.

services.AddFusionCache().WithKeyTransformer((key) => key?.toLower());

I was thinking, a solution to unify it could be for the 'WithCachePrefix' to internally create a function (key) => _prefix + key;

Regarding your question, it would be for L1 and L2, yes. In my case, they were using a legacy cache where the keys are case-insensitive. Since I'm refactoring using the fusion cache, I thought it was more prudent to maintain the same behavior.

However, this approach would only solve my problem. For the multitenant case, the transformer function would need to have some extra context, maybe something even via DI.

[jodydonetti]

Hi @mrxrsd

My case is simpler because I don't rely much on the context. If I had a key transformer Func<original_key, transformed_key>, it would solve the problem.

services.AddFusionCache().WithKeyTransformer((key) => key?.toLower());

In the past I thought about such a thing, but the main issue I can see is that while the current way is declarative (meaning: the input is a value, and the logic is applied by FusionCache itself), the transformer is procedural (meaning: the input is the logic).

This in turn means that what we currently have is, in theory, reversable: from a processed cache key (meaning: with the prefix applied) FusionCache may be able to get back the original key, while with the transformer that is not the case. There are scenarios I'm currently exploring where something like that may be useful: I'm not there/sure yet, but still.

Regarding your question, it would be for L1 and L2, yes.

Ok, I was asking because instead of doing a ToLower()/ToLowerInvariant() (which allocates etc, and is something you can already do before passing the key to FusionCache) I was thinking abut something like a case-sensitivity option, which under the cover may use some comparer or similar. The problem there was what would happen in an L2, where I have no control over that.

However, this approach would only solve my problem. For the multitenant case, the transformer function would need to have some extra context, maybe something even via DI.

Yep, that is another problem to tackle, definitely not easy to cover all edge cases.

Thanks!

[stebet]

As additional input into this discussion, I've been working on a backplane using NATS and it does pose some problems since there are some characters like :, !, > and * that are reserved and can not be used.

So to make that work we either need a way to configure/override these special values and separators or something like the transformer function mentioned above.