[ 
https://issues.apache.org/jira/browse/STORM-1961?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15515234#comment-15515234
 ] 

Roshan Naik edited comment on STORM-1961 at 9/23/16 3:16 AM:
-------------------------------------------------------------

# Strong typing in the APIs is really great!
# Would be good to do away with StreamBuilder and write it more directly and 
concisely :
{quote}
   Stream<String> x  = new Stream(...).flatmap().blah()
{quote}
# Can we avoid the build() call there 
{quote}
StormSubmitter.submitTopologyWithProgressBar("test", new Config(), 
builder.build());
{quote}
and simplify it to
{quote}
StormSubmitter.submitTopologyWithProgressBar("test", new Config(), stream );
{quote}
and have the build() or whatever else needs to happen, get invoked internally 
within submitTopology ?
# Good to have overloaded version taking arrays in flatMap(T[] ) and elsewhere 
to natively support arrays... so that conversion via Arrays.asList is  not 
needed.
# The doc Needs to have a more concrete definition for Stream concept.  
#* Is it just the data stream produced by the first operator/spout ? or is it 
the  whole pipeline of operators ? 
#* is if different from what we call 'topology' in storm ?  When you say ... 
Stream<T>  .. what is T ? Is the type of value produced by the terminal 
operator ? or that of the first ? 
#* What if there is a branch/split and each terminal operator creates different 
types ? 
#* Can a stream pick up data from two different sources ? for example from 
kafka and hdfs.
# The diagram in the doc shows fields and shuffle groupings. Not clear from the 
examples as to who the various gropings will be supported in the API. 
# Would like to see API examples in doc as how to the grouping and parallelism 
hints will be expressed in code.
# Would this API provide a mechanism to use the existing set of Storm spouts 
and terminal bolts (like KafkaSpout, HdfsSpout, HbaseBolt, etc) ? Or do we need 
to have new implementations ?


was (Author: roshan_naik):
# Strong typing in the APIs is really great!
# Would be good to do away with StreamBuilder and write it more directly and 
concisely :
{quote}
   Stream<String> x  = new Stream(...).flatmap().blah()
{quote}
# Can we avoid the build() call there 
{quote}
StormSubmitter.submitTopologyWithProgressBar("test", new Config(), 
builder.build());
{quote}
and simplify it to
{quote}
StormSubmitter.submitTopologyWithProgressBar("test", new Config(), stream );
{quote}
and have the build() or whatever else needs to happen, get invoked internally 
within submitTopology ?
# Good to have overloaded version taking arrays in flatMap(T[] ) and elsewhere 
to natively support arrays... so that conversion via Arrays.asList is  not 
needed.
# The doc Needs to have a more concrete definition for Stream concept.  
#* Is it just the data stream produced by the first operator/spout ? or is it 
the  whole pipeline of operators ? 
#* is if different from what we call 'topology' in storm ?  When you say ... 
Stream<T>  .. what is T ? Is the type of value produced by the terminal 
operator ? or that of the first ? #*What if there is a branch/split and each 
terminal operator creates different types ? 
#* Can a stream pick up data from two different sources ? for example from 
kafka and hdfs.
# The diagram in the doc shows fields and shuffle groupings. Not clear from the 
examples as to who the various gropings will be supported in the API. 
# Would like to see API examples in doc as how to the grouping and parallelism 
hints will be expressed in code.
# Would this API provide a mechanism to use the existing set of Storm spouts 
and terminal bolts (like KafkaSpout, HdfsSpout, HbaseBolt, etc) ? Or do we need 
to have new implementations ?

> Come up with streams api for storm core use cases
> -------------------------------------------------
>
>                 Key: STORM-1961
>                 URL: https://issues.apache.org/jira/browse/STORM-1961
>             Project: Apache Storm
>          Issue Type: Sub-task
>            Reporter: Arun Mahadevan
>            Assignee: Arun Mahadevan
>         Attachments: UnifiedStreamapiforStorm.pdf
>
>




--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to