Add from_encoding parameter to in_tail plugin

[footaku]

Encoding parameter was added in_tail on v.0.12.24 and I used it.
When I reading logs encoded in CP932+LF, fluentd output warn log like a below.

2016-06-01 05:14:40 +0000 [warn]: ... error="invalid byte sequence in UTF-8"

<source>
  @type tail
  path /path/to/log/foo.log
  pos_file /path/to/log/foo.log.pos
  tag foo.var
  encoding utf-8
  format /^ something to read $/
</source>
<match **>
  @type stdout
</match>

[tagomoris]

Specify type of this parameter.

[footaku]

@tagomoris
Thank you for reviewing. I would like to fix them.

[tagomoris]

Use log instead of $log to enable plugin logger (log_level for each plugin sections).
And I think it's ok to say just must be specified with "encoding" parameter.

[tagomoris]

LGTM!

[tagomoris]

@repeatedly Please check just once more.

[repeatedly]

This log is confusing.
If from_encoding must be specified with encoding, we should raise an ConfigError.

In addition, no need two line logs. 2nd line should be merged into 1st.

[repeatedly]

Here is microbenchmark result. I think default should be fast. So f2 implementation is better for almost users.

% ruby b_enc_perf.rb
Rehearsal --------------------------------------------------------
f1 with nil / nil      0.840000   0.000000   0.840000 (  0.842853)
f2 with nil / nil      0.760000   0.000000   0.760000 (  0.769146)
f1 with true / true    0.870000   0.000000   0.870000 (  0.871896)
f2 with true / true    0.920000   0.000000   0.920000 (  0.917469)
f1 with true / false   0.990000   0.000000   0.990000 (  0.989552)
f2 with true / false   0.920000   0.010000   0.930000 (  0.926458)
----------------------------------------------- total: 5.310000sec

                           user     system      total        real
f1 with nil / nil      0.890000   0.000000   0.890000 (  0.890333)
f2 with nil / nil      0.820000   0.000000   0.820000 (  0.822715)
f1 with true / true    0.890000   0.000000   0.890000 (  0.885648)
f2 with true / true    0.940000   0.000000   0.940000 (  0.938831)
f1 with true / false   1.000000   0.000000   1.000000 (  1.003493)
f2 with true / false   0.950000   0.000000   0.950000 (  0.956331)

Code:

require 'benchmark'

class Foo
  def initialize(enc, from_enc)
    @enc = enc
    @from_enc = from_enc
  end

  def f1(num)
    num.times {
      if @enc && @from_enc
        1
      elsif @enc
        2
      end
    }
  end

  def f2(num)
    num.times {
      if @enc
        if @from_enc
          1
        else
          2
        end
      end
    }
  end
end

n = 20000000

Benchmark.bmbm do |x|
  x.report('f1 with nil / nil') {
    foo = Foo.new(nil, nil)
    foo.f1(n)
  }
  x.report('f2 with nil / nil') {
    foo = Foo.new(nil, nil)
    foo.f2(n)
  }

  x.report('f1 with true / true') {
    foo = Foo.new(true, true)
    foo.f1(n)
  }
  x.report('f2 with true / true') {
    foo = Foo.new(true, true)
    foo.f2(n)
  }

  x.report('f1 with true / false') {
    foo = Foo.new(true, false)
    foo.f1(n)
  }
  x.report('f2 with true / false') {
    foo = Foo.new(true, false)
    foo.f2(n)
  }
end

[footaku]

Thank you for microbenchmark result!

[tagomoris]

Merged. Thank you for contribution!