Repository navigation
Optimize URI#hostname and URI#hostname= - #12
Conversation
| # | ||
| def hostname=(v) | ||
| v = "[#{v}]" if /\A\[.*\]\z/ !~ v && /:/ =~ v | ||
| v = "[#{v}]" if v&.index(':') && !v.start_with?('[') && !v.end_with?(']') |
There was a problem hiding this comment.
/\A\[.*\]\z/ !~ v and !v.start_with?('[') && !v.end_with?(']') differ.
It should be !(v.start_with?('[') && v.end_with?(']')).
There was a problem hiding this comment.
What an overlook, shame on me, this is 101. I will amend the fix. Also, I think I should start with start_with in the statement because index call is more costly and it is more likely to not match in the future as the world progresses towards IPv6.
b3a035f to
36c3fd7
Compare
|
Amended the change with force-push. |
|
Could you please check if there are tests for those methods? |
Yes, I can actually amend one more change - an extra edge case and a case when square brackets are already present. Lemme do that. |
36c3fd7 to
599d2b9
Compare
|
@lzap It wasn't me who meant |
|
Oh I am sorry, fixed! |
|
Sorry for the delay, this looks good, will merge shortly. |
4 participants
Hello,
URI#hostname extends URI#host with IPv6 support. In URI, IPv6 address must have square brackets (e.g. http://[2001:db8::1]), URI#hostname strips these characters out while URI#hostname= adds them if missing. There are three regular expressions to perform these tasks which can dramatically slow down performance. I am attaching a two-line patch and here is a benchmark: https://gist.github.com/lzap/24cbecb47daf29111350e41a24250922
In short, URI#hostname is 86-89% and hostname= is 55-154% faster according to my measurements.
For the record, I created a Ruby MRI RM ticket https://bugs.ruby-lang.org/issues/17219 as I had no idea URI is a separate gem hosted on github.