Improve placeholder

[joker1007]

Enable to use placeholder for project, dataset, fetch_schema_table.

Because of it, fetch_schema timing is delayed.

[tagomoris]

It's better to check whether the values from configuration parameters use invalid placeholders or not by calling placeholder_validate!(name, str) in #configure.
That method raises configuration error if ${tag} is used in str but configured buffer chunk key doesn't have tag (it also checks about other keys).

[joker1007]

@tagomoris Thanks!

I wanted to use placeholder_validate!, but I feel that current placeholder_validate! is not easy to use for me.
Because it is seemed that placeholder_validate! requires placeholder syntax in parameter.

For example in config,

<buffer time>
  timekey 1d
</buffer>

project yourproject_id
dataset yourdataset_id
table target_table_%Y%m%d # I want to use placeholder only table param in this config.

in this case, placeholder_validate! regards project and dataset as invalid.
The behavior is not suitable for this plugin.

Of course, I can guard placeholder_validate! if it is not required (ex. use regexp matching).
Is this intended usage?

[tagomoris]

@joker1007 It's intended - Once I tried to design validator API to validate combination of 2 or more configuration parameter values and configured buffer chunk keys, but I found that it's too difficult to both implement and use it.

For this case, we can validate the whole configuration and chunk keys by the code like below:

placeholder_validate!(:project_and_database_and_table, "#{@project}/#{@database}/#{@table}")

It shows error message a bit difficult to understand, but it's much better than occurring problems after Fluentd starts.

[joker1007]

I understand API design. Thanks!!
the snippet is looks good to me.
I will add placeholder_validate!.