[GitHub] nifi pull request #3075: NIFI-5604: Added property to allow empty FlowFile w...

[patricker <gi...@git.apache.org>]

Github user patricker commented on a diff in the pull request:

    https://github.com/apache/nifi/pull/3075#discussion_r233132498
  
    --- Diff: nifi-nar-bundles/nifi-standard-bundle/nifi-standard-processors/src/main/java/org/apache/nifi/processors/standard/GenerateTableFetch.java ---
    @@ -434,48 +448,53 @@ public void onTrigger(final ProcessContext context, final ProcessSessionFactory
                         numberOfFetches = (partitionSize == 0) ? 1 : (rowCount / partitionSize) + (rowCount % partitionSize == 0 ? 0 : 1);
                     }
     
    -                // Generate SQL statements to read "pages" of data
    -                Long limit = partitionSize == 0 ? null : (long) partitionSize;
    -                final String fragmentIdentifier = UUID.randomUUID().toString();
    -                for (long i = 0; i < numberOfFetches; i++) {
    -                    // Add a right bounding for the partitioning column if necessary (only on last partition, meaning we don't need the limit)
    -                    if ((i == numberOfFetches - 1) && useColumnValsForPaging && (maxValueClauses.isEmpty() || customWhereClause != null)) {
    -                        maxValueClauses.add(columnForPartitioning + " <= " + maxValueForPartitioning);
    -                        limit = null;
    -                    }
    +                // If there are no SQL statements to be generated, still output an empty flow file if specified by the user
    +                if (numberOfFetches == 0 && outputEmptyFlowFileOnZeroResults) {
    +                    session.transfer((fileToProcess == null) ? session.create() : session.create(fileToProcess), REL_SUCCESS);
    --- End diff --
    
    @mattyb149 It's not just the `fragment` attributes, I think you should add all the standard attributes. Right now this Flowfile get's routed to success with no attributes at all.
    
    I would suggest refactoring this to include all the standard attributes like `tablename`, `offset` etc... just like a real FlowFile would have. This will make it much easier for downstream processors to use this file like normal. I think the existing logic would cause `offset` to be `null`, which would make sense for this scenario too.


---