incubator-blur-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Aaron McCurry <amccu...@gmail.com>
Subject Re: BlurConfiguration
Date Thu, 20 Sep 2012 03:03:57 GMT
So, what if we add a separate method called override that allowed you
override the values in the blur-site / blur-default files?  And then you
could call that after the base constructor is called?

Also that would just a be a shortcut for calling config.set(name,value)
which you can do now (I think).  Would either of these solve the issue?

I think that an improvement to MiniCluster would be to random choose ports
for the Blur thrift servers to bind to automatically.  This is how the HDFS
mini cluster works.

Aaron

On Wed, Sep 19, 2012 at 10:55 PM, Aaron McCurry <amccurry@gmail.com> wrote:

> What would you suggest we do even if it's not backwards compatible?
>
>
> On Wed, Sep 19, 2012 at 10:21 PM, Tim Williams <williamstw@gmail.com>wrote:
>
>> On Wed, Sep 19, 2012 at 10:12 PM, Tim Williams <williamstw@gmail.com>
>> wrote:
>> > I'm working BLUR-23[1]  and running into an issue with configs.
>> > BlurConfiguration currently initializes all defaults in the
>> > constructor.  This makes it impossible to merge a few user overrides
>> > on top of the defaults.  If the user wants to pass a
>> > BlurConfiguration, they currently have to own the whole thing.  The
>> > only clean way I can see to move to allowing this while maintaining
>> > compatibility is to create a new constructor allowing passing in
>> > overrides.. (e.g. BlurConfiguration(BlurConfiguration overrides).
>> > Another solution might be to just add a "merge(BlurConfiguration
>> > overrides)" method, I suppose.
>>
>> Sorry, i take it back.  Because we initialize the defaults in the
>> default constructor (which the caller would have used) - I don't see a
>> way of allowing overrides in the existing paradigm through either
>> methods or constructors - in a backwards compatible way, at least.
>> Thoughts?
>>
>> --tim
>>
>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message