Multiple calls to current_user per action

Question

Multiple calls to current_user per action

jwoertink opened this issue 5 years ago · comments

This is a minor thing, but something to think about. Here's a log showing which method is being called on a basic install with Authentic:

web    | GET /
web    |  ▸ Handled by Home::Index
web    |  ▸ Ran protect_from_forgery
web    |  ▸ Ran test_backdoor
web    |  ▸ current_user?
web    |  ▸ current_user?
web    |  ▸ Rendered Lucky::WelcomePage
web    |  ▸ Sent 200 (1.56ms)
web    |
web    | GET /sign_up
web    |  ▸ Handled by SignUps::New
web    |  ▸ Ran protect_from_forgery
web    |  ▸ Ran test_backdoor
web    |  ▸ current_user?
web    |  ▸ Ran redirect_signed_in_users
web    |  ▸ current_user?
web    |  ▸ Rendered SignUps::NewPage
web    |  ▸ Sent 200 (939.0µs)
web    |
web    | POST /sign_ups
web    |  ▸ Handled by SignUps::Create
web    |  ▸ Ran protect_from_forgery
web    |  ▸ Ran test_backdoor
web    |  ▸ current_user?
web    |  ▸ Ran redirect_signed_in_users
web    |  ▸ Sent 302 (75.63ms)
web    |
web    | GET /me
web    |  ▸ Handled by Me::Show
web    |  ▸ Ran protect_from_forgery
web    |  ▸ Ran test_backdoor
web    |  ▸ current_user?
web    |  ▸ find_current_user
web    |  ▸ Ran require_sign_in
web    |  ▸ current_user?
web    |  ▸ find_current_user
web    |  ▸ Rendered Me::ShowPage
web    |  ▸ Sent 200 (2.08ms)

Notice once we get to the /me route, it's calling the find_current_user methods twice. We can't memoize that method because it takes an argument, but if we can find a way to memoize something in here, that could reduce the load time just a tad more.

Paul Smith · Answer 1 · Sun Sep 15 2019 11:55:43 GMT+0800 (China Standard Time)

Maybe we memoize in the auth mixins for `current_user` since those don’t take any args! Nice find BTW

…

Vlad Zarakovsky · Answer 2 · Wed Oct 09 2019 17:52:36 GMT+0800 (China Standard Time)

Looks like this one can be closed?

I've generated full app with Lucky 0.18.0 and don't see the duplicates in logs.

Jeremy Woertink · Answer 3 · Wed Oct 09 2019 23:28:21 GMT+0800 (China Standard Time)

@vlazar did you generate the app with the auth stuff? I don't think we've made any changes that would have affected this, but I'll take a look

Paul Smith · Answer 4 · Wed Oct 09 2019 23:30:13 GMT+0800 (China Standard Time)

I think the logging was added by the issue reporter. But there are still multiple SQL calls. I think we need to add memoize in the generated code to avoid calling the db every time `current_user` is called

…

On Oct 9, 2019, 11:28 AM -0400, Jeremy Woertink ***@***.***>, wrote: @vlazar did you generate the app with the auth stuff? I don't think we've made any changes that would have affected this, but I'll take a look — You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

Paul Smith · Answer 5 · Wed Oct 09 2019 23:31:13 GMT+0800 (China Standard Time)

https://github.com/luckyframework/lucky_cli/blob/a0ac6e52bcb0fd65605019905dd06fdb3e6582fa/src/api_authentication_app_skeleton/src/actions/mixins/api/auth/helpers.cr#L2

and https://github.com/luckyframework/lucky_cli/blob/a0ac6e52bcb0fd65605019905dd06fdb3e6582fa/src/browser_app_skeleton/src/actions/browser_action.cr.ecr#L28

need memoize

Vlad Zarakovsky · Answer 6 · Thu Oct 10 2019 00:08:07 GMT+0800 (China Standard Time)

Will memoize work in this case with User? type?

The Lucky guides states that

There are a few caveats to the memoize macro. It will not memoize false, or nil return values. If your method returns a “falsey” value, then it will be ran each time.

Paul Smith · Answer 7 · Thu Oct 10 2019 00:09:27 GMT+0800 (China Standard Time)

Good point! Haha. I guess I should know my own framework a bit better :P We'll have to figure something else out then. Maybe we need to bite but bullet and add some query cache to Avram. That'd fix this automatically which would be very cool

Jeremy Woertink · Answer 8 · Thu Oct 10 2019 00:24:22 GMT+0800 (China Standard Time)

It's definitely time for some query cache!

Vlad Zarakovsky · Answer 9 · Thu Oct 10 2019 00:32:44 GMT+0800 (China Standard Time)

So I assume building memoize macro which allows for nil and false results to be memoized is impossible in Crystal?

Paul Smith · Answer 10 · Thu Oct 10 2019 00:33:38 GMT+0800 (China Standard Time)

It is possible, but we haven't implemented it yet, and it can be error-prone

Vlad Zarakovsky · Answer 11 · Thu Oct 10 2019 00:44:16 GMT+0800 (China Standard Time)

@paulcsmith If I understood you correctly having memoize macro allowing nil and false might cause more trouble and not worth it?

This is not a good first issue then :)

Jeremy Woertink · Answer 12 · Thu Oct 10 2019 00:46:14 GMT+0800 (China Standard Time)

Yeah, I didn't add in nil or false support because it causes a lot of other issues. Same with passing arguments. There's some trade-offs, but our current memoize support works pretty great. If you're looking for some more "good first issues", I'll go through and start marking some stuff 😄 Thanks for all the help!

Vlad Zarakovsky · Answer 13 · Thu Oct 10 2019 00:57:54 GMT+0800 (China Standard Time)

@jwoertink Yeah, I'm looking around currently, going random directions to learn some Crystal and Lucky. Maybe I should also start building some app in Lucky.

Matthew McGarvey · Answer 14 · Tue Jun 09 2020 04:54:52 GMT+0800 (China Standard Time)

Just an update: we can now cache nil and false results with the memoize macro luckyframework/lucky@9be023f

Paul Smith · Answer 15 · Wed Jun 10 2020 00:38:35 GMT+0800 (China Standard Time)

@matthewmcgarvey YES! Thanks for doing that. I'll update my PR and merge that once green