google cloud dataflow - PipeLine with multiple Transformations -

i trying understand lifecycle of transforms within pipeline.

i have pipleline several transforms.

pipeline p = pipeline.create(options); p.apply(textio.read.named("readlines").from(inputfile))             .apply(new readdata())             .apply(new match())             .apply(new record())             .apply(bigqueryio.write                     .to(tableref)                     .withschema(getschema())                     .withcreatedisposition(bigqueryio.write.createdisposition.create_if_needed)                     .withwritedisposition(bigqueryio.write.writedisposition.write_truncate)); 

inside each of these transformations single dofn. entire batch node processing need complete before moving next transformation?

what observing @ least directpipelinerunner entire dataset read before match transformation run.

with directpipelinerunner, transforms executed entirely serially observed. when running dataflowpipelinerunner without --streaming set, many transforms can fused , run simultaneously. --streaming, data continually stream through entire pipeline, , transforms active.