ruby catch-throw and efficiency

Question

catch in Ruby is meant to jump out of deeply nested code. In Java e.g. it is possible to achieve the same with Java's try-catch meant for handling exceptions, it is however considered poor solution and is also very inefficient. In Ruby for handling exceptions we have begin-raise-rescue and I assume it is also to expensive to use it for other tasks.

Is Ruby's catch-throw really a more efficient solution then begin-raise-rescue or are there any other reasons to use it to break nested blocks instead of begin-raise-rescue?

Bentinck

If you post some ruby examples of the control structures you are asking about, it may be more clear what you mean. – Strophic

Answer 1

In addition to being the "correct" way to get out of control structures, catch-throw is also significantly faster(10 times as fast in my testing). Check out this gist for my code and results.

Epigraph

Thanks! The benchmark really clears the issue (in my case it is almost 15 times!). Should catch-throw be considered a structural style of programing -- a better practice then despised goto? – Bentinck

It's your design requires catch-throw, there is probably a better design. That being said, there are definitely edge cases where catch throw make a lot of sense, especially in gem develop where you don't know how the end user will use it. – Epigraph

Thanks for all explanations! – Bentinck

Answer 2

Josh's answer is correct. I want to add more information about catch-throw and raise-rescue.

catch-throw is used for flow control whereas raise-rescue is used for exception/error handling. The different is: backtrace is not needed for catch-throw (flow control). Trust me, the main reason causes raise-rescue runs slow than catch-throw 10 times in Josh's gist is raise-rescue takes a lot time to create backtrace object.

If you want to raise without backtrace, use syntax:

raise <type>, <message>, <backtrace>

Checkout my gist. raise without backtrace is much faster than raise with backtrace.

April 2016 Update:

I've updated my gist:

• Fixed "break" test
• Added benchmark tests results for newer ruby version 2.1.8, 2.2.4, 2.3.0

Emmeram