# I thought ruby didn't optimize tailcalls?

**URL:** <https://www.ruby-forum.com/t/i-thought-ruby-didnt-optimize-tailcalls/148511>\
**Category:** Ruby\
**Created:** [September 13, 2008, 10:10pm UTC](https://www.ruby-forum.com/t/i-thought-ruby-didnt-optimize-tailcalls/148511 "2008-09-13T22:10:40Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![cnb](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/c/9dc877/32.png) [@cnb](https://www.ruby-forum.com/u/cnb)\
**Post date:** [September 13, 2008, 10:10pm UTC](https://www.ruby-forum.com/t/i-thought-ruby-didnt-optimize-tailcalls/148511/1 "2008-09-13T22:10:40Z")

</div>

Im using ruby 1.8.6 mri. and it seems to optimize tailcalls which I  
have been told it doesn’t. what else could be going on here?

###################

def factail(n, acc=1)  
if n \> 0 then factail(n-1, n\*acc)  
else acc end  
end

irb(main):163:0\> factail(10000)  
factail(10000)  
284625968091705451890641321 and so on…

###################

def facto(n)  
if n \> 0 then n \* facto(n-1)  
else 1 end  
end

irb(main):164:0\> facto(10000)  
facto(10000)  
SystemStackError: stack level too deep  
from (irb):155:in `facto' from (irb):155:in`facto’  
from (irb):158  
from :0  
irb(main):165:0\>

########################

def fact(n)  
def f(n, acc)  
if n \> 0  
then f(n-1, n\*acc)  
else acc  
end  
end  
f(n, 1)  
end

irb(main):183:0\> fact(10000)  
fact(10000)  
2846259680917054518 …

also works.

---

<div class="post-metadata">

**Author:** ![cnb](https://www.ruby-forum.com/letter_avatar_proxy/v4/letter/c/9dc877/32.png) [@cnb](https://www.ruby-forum.com/u/cnb)\
**Post date:** [September 15, 2008, 9:34am UTC](https://www.ruby-forum.com/t/i-thought-ruby-didnt-optimize-tailcalls/148511/2 "2008-09-15T09:34:07Z")

</div>

cnb wrote:

> Im using ruby 1.8.6 mri. and it seems to optimize tailcalls which I  
> have been told it doesn’t. what else could be going on here?

With that program, factail(2000) works for me but factail(5000) bombs  
out with “stack too deep”. So I guess you just have a larger default  
stack size than me.

Factorials generate very large Bignums which will slow things down  
substantially for large numbers. I suggest you try something simpler:

def counttail(n, acc=0)  
if n \> 0 then counttail(n-1, acc+1)  
else acc end  
end

Then you can try counttail(10\_000), counttail(100\_000),  
counttail(1\_000\_000) etc until you can show your stack overflowing.

If counttail(10\_000\_000) still works, then post your exact version of  
MRI and where you got it from. Perhaps you have some sort of patched  
version (although I’ve not come across such a patch myself)

Regards,

Brian.