Pipeline Executor return no rows

[fbourgine-cpu]

Hi everyone !
I try to use Pipeline Executor but I can't make it work properly.
The sub pipeline works fine and return many rows. I try to output on Google Sheet, and everything works fine.
I substitute with "Copy rows to results" and the results are good in the preview.
But when I try to execute this pipeline in a parent pipeline with Pipeline Executor, the execution is immediate and return nothing.
I tried many things but nothing is working ...

Can you help me understand ?
Thank you ! 🙏

--
Apache Hop 2.14
MacOS 26.1
openjdk 21.0.8 2025-07-15 LTS
OpenJDK Runtime Environment Microsoft-11933201 (build 21.0.8+9-LTS)
OpenJDK 64-Bit Server VM Microsoft-11933201 (build 21.0.8+9-LTS, mixed mode, sharing)

[liosc]

Can you share the two pipelines?

[hansva]

Also take a look at the sample in the samples project under loops/pipeline-executor.hpl to see how it works.

[dave-csc]

Does your sub-pipeline have an input, too (i.e. starts with a Get rows from result)?

Pipeline Executor is meant to execute a pipeline for each row (or group of rows) in a parent pipeline, if you have no input you should probably place the sub-pipeline in a workflow rather than a pipeline.

Alternatively, you can still place it in a parent pipeline, but you need to trigger it with at least one row (the simplest way is to link a Generate rows transform before)

Hope this helps :)

[fbourgine-cpu]

Hi Thank you very much for your help. I tried to execute the sample project and it works fine. So my usage is wrong I guess.. I try to execute 2 branches in parallel and merge data but not looping. So my sub transform has no input. I just want to clarify my code. Here is an example. Fabrice Le ven. 7 nov. 2025 à 08:24, dave-csc ***@***.***> a écrit :

…

Does your sub-pipeline have an input, too (i.e. starts with a *Get rows from result*)? Pipeline Executor is meant to execute a pipeline for each row (or group of rows) in a parent pipeline, if you have no input you should probably place the sub-pipeline in a workflow rather than a pipeline. Alternatively, you can still place it in a parent pipeline, but you need to trigger it with at least one row (the simplest way is to link a *Generate rows* transform before) Hope this helps :) — Reply to this email directly, view it on GitHub <#5965 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BZZ3DEAHURBIKY6TJ64NQX333RCLBAVCNFSM6AAAAACLLWDUFOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTIOBZHE2TANI> . You are receiving this because you authored the thread.Message ID: ***@***.***>

-- <https://www.letelegramme.fr/> *Fabrice BOURGINE* * DIRECTEUR SYSTEMES D'INFORMATION GROUPE * ***@***.*** <http:///cdn-cgi/l/email-protection#680e460a071d1a0f01060d28040d1c0d040d0f1a0905050d460e1a> 02 98 62 20 19 / 06 15 69 34 43 Plan d'accès MORLAIX <https://www.google.com/maps?q=7+VOIE+D'ACCÈS+AU+PORT+-+29600+MORLAIX> [image: LinkedIn] <https://fr.linkedin.com/company/letelegramme> [image: Facebook] <https://www.facebook.com/letelegramme> [image: Twitter X] <https://x.com/LeTelegramme> [image: Instagram] <https://www.instagram.com/letelegramme> [image: Youtube] <https://www.youtube.com/user/LeTelegramme> [image: Campagne mail] ***@***.***>

[hansva]

You could add a generate rows and generate 1 row to make them execute. Though if there is no "looping" aspect then I am not sure using pipeline executors is the best approach. What is the reasoning? Do you want to have a reusable block that has to be used somewhere else?

[dave-csc]

If I get correctly, you need to get the output of two distinct pipelines and merge them somehow in a parent pipeline.

Then you can call each one with a Pipeline Executor, but as mentioned above you need to trigger them with a single row: use a Generate row transform, configure it to create a single row and no schema. Link this transform to (all) the subsequent Pipeline Executor(s), this is enough to trigger it/them (remember to "copy" the row and not to "distribute" it)

Finally, link the Pipeline Executor output(s) to the merging transform

[fbourgine-cpu]

Hi Thank you for your reply I tried but get an issue (JSON..) Error registering execution information sub_pipeline1.hpl.0 - Null key for a Map not allowed in JSON (use a converting NullKeySerializer?) (through reference chain: org.apache.hop.execution.Execution["parameterValues"]) Le ven. 7 nov. 2025 à 10:08, dave-csc ***@***.***> a écrit :

…

If I get correctly, you need to get the output of two distinct pipelines and merge them somehow in a parent pipeline. Then you can call each one with a Pipeline Executor, but as mentioned above you need to trigger them with a single row: use a Generate row transform, configure it to create a single row and no schema. Link this transform to (all) the subsequent Pipeline Executor(s), this is enough to trigger it/them (remember to "copy" the row and not to "distribute" it) Finally, link the Pipeline Executor output(s) to the merging transform — Reply to this email directly, view it on GitHub <#5965 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/BZZ3DEHOM6ED7YU7DGZCYPT33RORXAVCNFSM6AAAAACLLWDUFOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTIOJQGAZTONY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[fbourgine-cpu]

Ok I succeded ! My input parameter was not set in the sub_pipelines See attached Thanks very much ! Best Regards Fabrice Le ven. 7 nov. 2025 à 11:28, Fabrice BOURGINE < ***@***.***> a écrit :

…

Hi Thank you for your reply I tried but get an issue (JSON..) Error registering execution information sub_pipeline1.hpl.0 - Null key for a Map not allowed in JSON (use a converting NullKeySerializer?) (through reference chain: org.apache.hop.execution.Execution["parameterValues"]) Le ven. 7 nov. 2025 à 10:08, dave-csc ***@***.***> a écrit : > If I get correctly, you need to get the output of two distinct pipelines > and merge them somehow in a parent pipeline. > > Then you can call each one with a Pipeline Executor, but as mentioned > above you need to trigger them with a single row: use a Generate row > transform, configure it to create a single row and no schema. Link this > transform to (all) the subsequent Pipeline Executor(s), this is enough to > trigger it/them (remember to "copy" the row and not to "distribute" it) > > Finally, link the Pipeline Executor output(s) to the merging transform > > — > Reply to this email directly, view it on GitHub > <#5965 (reply in thread)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/BZZ3DEHOM6ED7YU7DGZCYPT33RORXAVCNFSM6AAAAACLLWDUFOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTIOJQGAZTONY> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> >