DynamicFinderMatch & Co in callbacks chain is a waste

rubyonrails-core
1

If anyone is working on the callback chain (Yehuda is, I think), is there a chance that the callback chain avoids calls to "respond_to?" on class objects in order to skip the DynamicFinder stuff ?

These calls quickly add up. Typical Node.update_attributes (2 models involved) : 53 calls !

Gaspard

2

If anyone is working on the callback chain (Yehuda is, I think), is there a chance that the callback chain avoids calls to "respond_to?" on class objects in order to skip the DynamicFinder stuff ?

These calls quickly add up. Typical Node.update_attributes (2 models involved) : 53 calls !

Do you have a benchmark in mind which shows where this is a problem? Counting method calls is fun if you buy into the reddit meme that calling methods is inherently slow. But even 53 methods will be dwarfed by rendering a single ERB template.

3

You are right. I need to change my assembler thinking mind. Sorry for that.

Gaspard

4

Gaspard Bucher wrote:

You are right. I need to change my assembler thinking mind. Sorry for that.

This could still be a problem, we'd just need to run a full benchmark to see the impact.