You’ll need to have a user parameter in your child workspace that you can use to limit the amount of data it’s going to process, then fill that parameter dynamically through the WorkspaceRunner.

Without knowing a bit more about what you’re trying to do and how you could potentially divide the data up it’ll be hard to give you more detailed advice.

Dear ​@redgeographics ,

Thank you very much for your reply.

In this case, and to begin, I just want to keep fex line form my ESRI Shapefile, and use a WorkspaceRunner to run a LineCombiner.

I’ve, User parameter in my parent workbench, where I’ve my source shapefile :

And after, in my child workbench, I’ve two User parameter :

1/ The same source from my parent workbench where I’ve the Sapefile

2/ Gesojson writter, for the ouptut of my workbench (it is this file waht I read in the FeatureReader)

Okay, so is there a logical way that you can divide those lines? Since you’re grouping on code_ouvra in your LineCombiner that is probably best, do you know how many unique values there are for that?

Assuming it’s not too many:

1. Change your child workspace to filter out a single value for code_ouvra and then use the LineCombiner on those. Make that value a User Parameter
2. Change your parent workspace to have a Sampler after reading your data, set that to group by code_ouvra and only pass the first feature. Essentially this gives you all the unique values for code_ouvra.
3. Have the WorkspaceRunner run your child worksplace and plug the code_ouvra attribute into that User Parameter. 
4. You should end up with one file per code_ouvra, so you’ll need to combine them in the end, I think your FeatureReader does that.

There’s still a bit too much overhead for my liking: the child workspace does read all of the features before selecting them, so this may not be an improvement.