[ 
https://issues.apache.org/jira/browse/S2GRAPH-229?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16529440#comment-16529440
 ] 

Daewon Jeong commented on S2GRAPH-229:
--------------------------------------

[~steamshon] Thank you for your feedback.

I first created the 'RxStep' interface as you suggested.
{code}
trait RxStep [-A, + B] extends (A => Observable [B])
{code}
The reason inherited the function(A => Observable[B]) is that I actually want 
the steps we are using to naturally connect over the 'Observable' interface.

 

Finally, I wrote a simple test code that replaces the existing 's2graph query' 
using the above interface.

I will write more about PR.

> 'Step' abstraction for combinable queries
> -----------------------------------------
>
>                 Key: S2GRAPH-229
>                 URL: https://issues.apache.org/jira/browse/S2GRAPH-229
>             Project: S2Graph
>          Issue Type: Task
>          Components: s2core
>            Reporter: Daewon Jeong
>            Priority: Minor
>
> Graph Query is a combination of queries of each step.
> For example, 
>   1. you can take a vertex, 
>    2. then bring the edge back with that vertex, 
>    3. get the vertex from the edge, 
>    4. and bring the edge from that vertex.
> The query that is being provided now is not tied to each of the above steps, 
> but is tightly coupled throughout the query.
> Since functions such as filter, transform(group_by, merge), and 
> postprocess(conversion to json/xml) are tightly coupled to the entire query 
> function, 
>  it is difficult to test each function separately. 
> To solve this problem, 
>    1. each step is created as an independent function and 
>    2. the function of assembling each step is separated separately.
> Each step is separated as above, the following functions can be implemented 
> without additional code.
>   ex) In the case of deleteAll, fetch each edge and make delete request 
> again. 
> Implement Prototyping
>     1. For query efficiency, each step implements the function to return 
> Observable.
>     2. Use Observable(RxScala)'s flatMap to connect each step.
> Below is the pseudocode.
> 1. Bringing the edges from the vertex, collect only edges whose score is 0.5 
> and serialize them to Json
> {code:java}
>  val vs: Step[Int, S2VertexLike] = VertexStep()
>  val es: Step[S2VertexLike, S2EdgeLike] = EdgeStep(dir = "OUT")
>  val js: Step[S2EdgeLike, Json] = JsonSerializeStep(limit: 10)
> val q: Observable[Json] = 
>  vs.apply(1) 
>  .flatMap(es)
>  .filter(_.property("score") > 0.5)
>  .flatMap(js)
> {code}
> What to think about.
>     1. The ability to link each step is subject to Observable.
>  - This makes it impossible to draw the connection graph of the Step instead 
> of making the function easier to implement.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to