Provide initial implementation

[DmitryTsepelev]

More details on Cult.

We need to implement a module that patches controller actions to measure database and HTTP response payload:

 class ReportController < ApplicationController
   include IoToResponsePayloadRatio::Controller

   def monthly_transaction_sum
     sum = Transactions.where(date: Date.current.all_month).sum(&:amount)
     render json: { sum: sum }
   end
 end

It can help find possible memory bloats earlier. These measurements can be sent to Rails logs or published as notification events. For instance, log could look like this:

Completed 200 OK in 1232ms (ActiveRecord: 400ms, DB Payload: 10kB, Body: 1kB, Allocations: 213223)

[prog-supdex]

Hello, @DmitryTsepelev!
I am now working on this and have several questions)

What Do I need to measure? Allocated memory of executed query and builded ActiveRecord::Relation or just calculate fetched data from database using bytesize (PG::Result#tuples or ActiveRecord::Result#rows)?

Just, if we execute

Transaction.select(:amount).sum(&:amount)

we will get all instances of Transaction, not only field amount.

and

Transaction.pluck(:amount).sum

will get less memory, than first, but same data fetched

Currently, i am measuring spent memory for this and also calculates bytes of fetched data. And I don`t know, which way is right

Second question. Should I calculate response body only when using render json? Or Have I also think about rendering view case?

[DmitryTsepelev]

Hi @prog-supdex!

will get less memory, than first, but same data fetched

That's right, and Transaction.sum(:amount) will get the same data fetched with even less amount of memory needed, and that's what I'd like to measure 🙂My assumption was that when we load a lot of data from the source but return way less data to the client there is a chance we can move these calculations outside our app, and the gem can help detect such things.

Should I calculate response body only when using render json?

My idea was to start with something simple, so JSON would be enough. The task is a huge enough already 🙂You can also omit some data sources and focus on one or two (e.g., only DB).