Jacques,
I made some progress on this today. I have a generic DrillbitService class
which is part of the exec module. This is implemented by HttpEngine. For
the cp scanning bit I was hoping we could use DrillbitConfig itself. I want
to do something like -
*
*
* config.getInstanceOf("drill.exec.http.implementation",
DrillbitService.class); // The key "**drill.exec.http.implementation" is
mapped to the actual class*
I think this should work as long as the class is found in the cp. But in
the build order http-engine gets built after java-exec and hence I get a
ClassNotFound while running the tests. The compile time issue is gone but
the runtime issue is still there. Any ideas on how to workaround this?
*
*Hari
On Tue, Jul 30, 2013 at 5:26 PM, Srihari Srinivasan <
[email protected]> wrote:
> Hi
>
> I won't be able to join the stand up today. My update for the week - I'll
> continue to work on getting the HttpEngine to integrate with the Drillbit
> class based on the approach suggested by Jacques.
>
> -- Hari
>
>
> On Sun, Jul 28, 2013 at 10:55 PM, Srihari Srinivasan <
> [email protected]> wrote:
>
>> This should be possible. I'll try and see how it goes!
>>
>>
>> On Sat, Jul 27, 2013 at 7:09 PM, Jacques Nadeau <[email protected]>wrote:
>>
>>> I suggest you define a generic Drill bit Service interface that receives
>>> the drill configuration and is started/stopped and then use drill module
>>> conf and cp scanning similar to how we incorporate CallProviders or
>>> DrillFuncs. Then switch the dependency direction using the pom snippet I
>>> provided and have your service implement this new interface..
>>> Thoughts?
>>> On Jul 27, 2013 6:32 AM, "Srihari Srinivasan" <[email protected]
>>> >
>>> wrote:
>>>
>>> > The HttpEngine class which encapsulates the Jetty server is
>>> instantiated
>>> > from the Drillbit class to start the web server. So java-exec depends
>>> on
>>> > http-engine.
>>> >
>>> > Happy to have a quick hangout on this if you have the time...
>>> >
>>> > Hari
>>> >
>>> > On Sat, Jul 27, 2013 at 6:42 PM, Jacques Nadeau <[email protected]>
>>> > wrote:
>>> >
>>> > > I didn't realize that. Why do you have that dependency?
>>> > > On Jul 27, 2013 5:04 AM, "Srihari Srinivasan" <
>>> [email protected]
>>> > >
>>> > > wrote:
>>> > >
>>> > > > If I understand correctly what you are suggesting is we add the
>>> snippet
>>> > > > below in the http-engine's pom file. Since I already have java-exec
>>> > > depend
>>> > > > on http-engine won't this make it a case of having cyclic
>>> dependencies
>>> > > that
>>> > > > mvn will complain about?
>>> > > >
>>> > > > Hari
>>> > > >
>>> > > > On Fri, Jul 26, 2013 at 11:07 PM, Jacques Nadeau <
>>> [email protected]>
>>> > > > wrote:
>>> > > >
>>> > > > > Let's get an alpha out before we break out the client.
>>> > > > >
>>> > > > > In the meantime, I downloaded your stuff and it seemed to work
>>> fine
>>> > > > > when it depends on java-exec.
>>> > > > >
>>> > > > > I'd suggest you use the dependency like this to avoid any weird
>>> > > > conflicts:
>>> > > > >
>>> > > > > <dependency>
>>> > > > > <groupId>org.apache.drill.exec</groupId>
>>> > > > > <artifactId>java-exec</artifactId>
>>> > > > > <version>1.0-SNAPSHOT</version>
>>> > > > > <exclusions>
>>> > > > > <exclusion>
>>> > > > > <artifactId>jsp-api-2.1</artifactId>
>>> > > > > <groupId>org.mortbay.jetty</groupId>
>>> > > > > </exclusion>
>>> > > > > <exclusion>
>>> > > > > <artifactId>jetty</artifactId>
>>> > > > > <groupId>org.mortbay.jetty</groupId>
>>> > > > > </exclusion>
>>> > > > > </exclusions>
>>> > > > > </dependency>
>>> > > > >
>>> > > > >
>>> > > > >
>>> > > > > On Fri, Jul 26, 2013 at 9:50 AM, Srihari Srinivasan
>>> > > > > <[email protected]> wrote:
>>> > > > > > Hi All,
>>> > > > > >
>>> > > > > > Any thoughts on the DrillClient modularization/short term
>>> > > alternative?
>>> > > > > >
>>> > > > > > Hari
>>> > > > > >
>>> > > > > >
>>> > > > > > ---------- Forwarded message ----------
>>> > > > > > From: Srihari Srinivasan <[email protected]>
>>> > > > > > Date: Tue, Jul 23, 2013 at 10:33 PM
>>> > > > > > Subject: Re: Meeting today
>>> > > > > > To: [email protected]
>>> > > > > > Cc: [email protected]
>>> > > > > >
>>> > > > > >
>>> > > > > > Hi,
>>> > > > > >
>>> > > > > > Quick note on what we discussed w.r.t the REST API.
>>> > > > > >
>>> > > > > > I am at a point where to submit a query (logical for now) to
>>> the
>>> > > > > Drillbit I
>>> > > > > > need an instance of the DrillClient class from the Query REST
>>> > > resource.
>>> > > > > > DrillClient is currently part of the java-exec module. I've
>>> pulled
>>> > > out
>>> > > > > the
>>> > > > > > jersey/jetty stuff into its own mvn module to keep the concerns
>>> > > > separate.
>>> > > > > >
>>> > > > > > So the really question is - Is now a good time to pull out
>>> > > DrillClient
>>> > > > > and
>>> > > > > > its dependencies into its own client module? That way
>>> http-engine
>>> > can
>>> > > > > just
>>> > > > > > depend on the client lib. Having this will be a huge help as
>>> it can
>>> > > > > > potentially keep a parallel stream of working going.
>>> > > > > >
>>> > > > > > Any other ideas/suggestions?
>>> > > > > >
>>> > > > > > Jason - This is the link to my github
>>> > > > > > branch<https://github.com/srihari/incubator-drill>.
>>> > > > > > Its still a clone of the execwork branch.
>>> > > > > >
>>> > > > > > -- Hari
>>> > > > > >
>>> > > > > > On Fri, Jul 12, 2013 at 9:22 PM, Jacques Nadeau <
>>> > [email protected]>
>>> > > > > wrote:
>>> > > > > >
>>> > > > > >> Sounds good to me.
>>> > > > > >>
>>> > > > > >> On Fri, Jul 12, 2013 at 7:07 AM, Srihari Srinivasan
>>> > > > > >> <[email protected]> wrote:
>>> > > > > >> > HI Folks,
>>> > > > > >> >
>>> > > > > >> > Moving on I am trying to implement the functionality that
>>> > creates
>>> > > a
>>> > > > > query
>>> > > > > >> > via the REST interface. I have a question w.r.t its
>>> > > implementation.
>>> > > > > One
>>> > > > > >> way
>>> > > > > >> > to implement this would be via the DrillClient class by
>>> calling
>>> > > its
>>> > > > > >> > runQuery(QueryType) method as illustrated in
>>> > > > > TestDistributedFragmentRun.
>>> > > > > >> >
>>> > > > > >> > Does this approach make sense? Should we be considering any
>>> > other
>>> > > > > >> > alternatives?
>>> > > > > >> >
>>> > > > > >> > Hari
>>> > > > > >>
>>> > > > >
>>> > > >
>>> > >
>>> >
>>>
>>
>>
>
