Rack::CommonLogger does not receive Logger Instance. is this bug?

Rack::CommonLogger does not receive Logger Instance. is this bug?

[sanemat]

hi. I'm sanemat.

I use rest-client-components and rack-common-logger.
It does not work, and I find problem in Rack::CommonLogger.

Logger Instance does not have +write+ method.
But Rack::CommonLogger's constructer require +write+ method in second
parameter.

------
NoMethodError - undefined method `write' for #<Logger:0xb736acec>:
 /usr/lib/ruby/gems/1.8/gems/rack-1.1.0/lib/rack/commonlogger.rb:31:in
`log'
 /usr/lib/ruby/gems/1.8/gems/rack-1.1.0/lib/rack/commonlogger.rb:20:in
`call'
------

modify way I think:
 a. change Logger instance
 b. change commonlogger

or anyway.

so i try 'a' way like below, and it works.

http://github.com/choonkeat/tweetstreamproxy/commit/ec82b0ac04b4f20ba33bf4f25dfb3f60e122d44c
  Logger.class_eval { alias :write :"<<" } unless
Logger.instance_methods.include? "write"

or 'b' way. like this.
http://github.com/sanemat/rack/commit/865ab9655aa2aa226a9d21b8dfd39e001c9bdeb0
change method again from +write+ to +<<+. this changed one years ago.
http://github.com/rack/rack/commit/761c6246e18b92752353c0f412976ded4a417dc5

which is better?
I confuse this.
i may be wrong but few people use Logger?

thank you.

rack 1.1.0
$ ruby -v
ruby 1.8.7 (2009-12-24 patchlevel 248) [i686-linux]
$ uname -a
Linux fri 2.6.32.10-90.fc12.i686 #1 SMP Tue Mar 23 10:21:29 UTC 2010
i686 i686 i386 GNU/Linux
--
sanemat

Re: Rack::CommonLogger does not receive Logger Instance. is this bug?

[James Tucker]

On 5 May 2010, at 13:03, sanemat wrote:
> Logger Instance does not have +write+ method.
> But Rack::CommonLogger's constructer require +write+ method in second
> parameter.

class MyLogger << Logger
  alias write <<
end

use Rack::CommonLogger, MyLogger.new

> modify way I think:
> a. change Logger instance
> b. change commonlogger
> 
> or anyway.
> 
> so i try 'a' way like below, and it works.
> 
> http://github.com/choonkeat/tweetstreamproxy/commit/ec82b0ac04b4f20ba33bf4f25dfb3f60e122d44c
>  Logger.class_eval { alias :write :"<<" } unless
> Logger.instance_methods.include? "write"

This is horrible. I do wish people would subclass first rather than monkey patch first (glaring evil stares at rails here too as there's no need to monkey patch Logger, like, ever).

> or 'b' way. like this.
> http://github.com/sanemat/rack/commit/865ab9655aa2aa226a9d21b8dfd39e001c9bdeb0
> change method again from +write+ to +<<+. this changed one years ago.
> http://github.com/rack/rack/commit/761c6246e18b92752353c0f412976ded4a417dc5

I see the implementation used to be compatible with Logger, that change did indeed break it. It uses write to expect something IO-ish, not something Logger-ish, but I think it would be a valid change to use << instead, however, I cannot remember if we specify this as a requirement for env['rack.errors']. Certainly we should decide one way, however, I have no real problem with the way it presently is, the implementation being simpler than it used to be.

> which is better?
> I confuse this.
> i may be wrong but few people use Logger?

Generally, no, people use env['rack.errors'], which is IO-ish.

Re: Rack::CommonLogger does not receive Logger Instance. is this bug?

[Paul O'Keeffe]

[-- Attachment #1: Type: text/plain, Size: 257 bytes --]


>
> Calling that second parameter to the initialize method 'logger' tripped me 
> up too. Having not seen earlier implementations, I assumed this to be a 
> ::Logger rather than an IO stream.Perhaps the parameter should be renamed 
> to 'io' or 'stream'?


[-- Attachment #2: Type: text/html, Size: 370 bytes --]