From: Eric Wong <normalperson@...>
Date: 2015-12-11T20:10:17+00:00
Subject: [ruby-core:72060] Re: [Ruby trunk - Feature #11405] [Assigned] [PATCH] hash.c: minor speedups to int/fixnum keys

Yusuke Endoh <mame@ruby-lang.org> wrote:
> > I'm curious, how did you notice this?
> 
> In short, I was studying hash conflicts for another reason.
> 
> The detailed context is off topic, but you might be interested.

<snip>

> Though I'm not sure if this IS a problem, it was anyway fixed by nobu
> at r53031 by explicitly rehashing a cdhash.

Thanks, I was wondering about the reference in the r53031 commit
message last night.
(http://d.hatena.ne.jp/ku-ma-me/20151210)

> But this behavior of st_table is not only for case statement.  I could
> reproduce it by normal Hash:
> 
>     h = {}
>     10000.times do |n|
>       h[n] = true
>     end
>     1_000_000_000.times { h[9999] } # 57.2 sec
>     1_000_000_000.times { h[0] }    # 89.1 sec
> 
> I'm not sure if (and how) we should fix this.  Reducing rehash
> threshold would work, but it will cause frequent rehashing, which may
> lead to another overhead and memory waste.

Perhaps teach developers to rehash explicitly.  Maybe Hash#freeze
can imply rehash, too.  But yes, I worry about memory increase more
than speed nowadays.