Question: Workflow Improvement Requests (Long)
0
Assaf Gordon • 320 wrote:
Dear all,
Recently, users (of our local galaxy server) started using workflows,
and are very pleased. However, as workflows get more complicated, it
gets harder to track the input and output of the workflows.
I'd like to share an example, to illustrate the problems that we
encounter.
The workflow (pictured in the attached 'workflow.jpg') takes 4 input
datasets, and produces 4 output datasets.
The first problem is that there's no way to differentiate between the
input datasets (They appear simply as "Step 1: Input dataset", "Step
2:
Input Dataset", etc). Since each dataset has a specific role, I've had
to print the workflow and give the users instructions as to which
dataset (in their history) goes into what dataset. (see attached
'crosstab_workflow_input_datasets.jpg').
The second problem is that whenever I change something in the workflow
and save it - the order of the dataset change!
So what was once dataset 1, can now be dataset 2,3 or 4.
Users have no way of knowing this... (keen users might notice the the
description of the first tool changed from "Output dataset 'output'
from
step 2" to "Output dataset' output' from step 4" - but this is very
obscure...).
The third problem is that once the workflow completes, the resulting
dataset have cryptic names such as "Join two queries on Data 10 and
Data
2". Since "Data 10" is "Awk on Data 8" and data-8 is "Generic
Annotations on Data 7 and Data 1" and data-7 is "Intersect data 1 and
data 6" - it gets a bit hard to know what's going on. (see attached
'crosstab_history.png').
For the meantime, I've simply gave written instructions on what each
dataset means (see attached
'crosstab_workflow_dataset_explnanations.jpg).
If I may suggest a feature - it would be great if I could name a
dataset
inside the workflow. Instead of naming it "Input dataset" I could give
it a descriptive name, so even if the order of the input datasets
changes, users will know which dataset goes into which input.
Regarding the output dataset names, the 'label' option in the tools'
XML
is a good start, but still creates very long, hard-to-understand
names.
Another great feature would be the possibility to add an 'output
label'
for each step in the workflow.
Regardless of the above, I'd like to say (once again) that Galaxy is a
great tool, and workflows are really cool - we have several long
workflows which do wonderful things.
Thanks for reading so far,
Gordon.
ADD COMMENT
• link
•
modified 10.0 years ago
by
James Taylor • 70
•
written
10.0 years ago by
Assaf Gordon • 320
