trafficserver-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Leif Hedstrom <zw...@apache.org>
Subject Re: Remap/regex_remap consolidation
Date Wed, 26 Mar 2014 21:01:52 GMT


> On Mar 26, 2014, at 12:31 PM, Thomas Jackson <jacksontj.89@gmail.com> wrote:
> 
> The biggest performance gain I see of separate regexes is that I can
> execute the unique domain regexes which should resolve to a list of path
> regexes. This should be pretty big performance wise as a lot of rules in
> large remaps have the same domain which may or may not be a regex.

Agreed. That also retains some if the old regex_remap performance characteristics.

-- Leif 

>> On Mar 26, 2014 6:20 AM, "Leif Hedstrom" <zwoop@apache.org> wrote:
>> 
>> 
>> 
>>> On Mar 26, 2014, at 2:02 AM, Brian Geffon <briang@apache.org> wrote:
>>> 
>>> Thomas, I somewhat agree: my guess would be the additional regexes would
>>> likely cancel any performance gain there.
>>> 
>>> Does anyone else have feedback or comments?
>> 
>> 
>> The other argument for this is that with separate regexes, you don't have
>> to create the full URL string representation. I don't know if the core has
>> any optimizations here, but for a plugin that is an expensive operation.
>> With separate regexes for host and path, this is a non-issue.
>> 
>> Maybe you can do separate regexes but let expansions cross over? $h[1] for
>> a host group etc.
>> 
>> Pros and cons :)
>> 
>> -- Leif
>> 
>>> 
>>> Brian
>>> 
>>>> On Tuesday, March 25, 2014, Thomas Jackson <jacksontj.89@gmail.com>
>> wrote:
>>>> 
>>>> Another consideration for having >1 regex (which may or may not be
>>>> premature optimization) is that if you have seperate regexes we can
>> create
>>>> hash maps similar to how maps work (a hashmap of domain_regex -> list
of
>>>> path regexes) which would make overall remap performance faster/better.
>>>> 
>>>> 
>>>> On Tue, Mar 25, 2014 at 9:50 PM, Brian Geffon <briang@apache.org
>> <javascript:_e(%7B%7D,'cvml','briang@apache.org');>
>>>>> wrote:
>>>> 
>>>>> Right.
>>>>> 
>>>>> 
>>>>> On Tuesday, March 25, 2014, Leif Hedstrom <zwoop@apache.org
>> <javascript:_e(%7B%7D,'cvml','zwoop@apache.org');>>
>>>>> wrote:
>>>>> 
>>>>>> 
>>>>>>> On Mar 25, 2014, at 7:51 PM, Brian Geffon <briangeffon@gmail.com>
>> wrote:
>>>>>>> 
>>>>>>> What Thomas called Question #1 -- 1 Regex.
>>>>>> 
>>>>>> Makes sense to have them combined. Assuming groups etc. works, that
>>>>>> allows you to do e.g.
>>>>>> 
>>>>>>  regex_map http://(.*)\.ogre\.com/([^/]+)/(.*)  http://$2/$1/$3
>>>>>> 
>>>>>> 
>>>>>> or some such. i.e. take parts from the path match and use as the
host,
>>>>>> and vice versa. Right? :)
>>>>>> 
>>>>>> Cheers,
>>>>>> 
>>>>>> -- Leif
>> 

Mime
View raw message