search

common-workflow-language / cwlviewer

A web application to view and share Common Workflow Language workflows
https://view.commonwl.org/
Apache License 2.0
42 stars 27 forks source link

Defer Research Object Bundle creation ; or drop altogether? #305

Open mr-c opened 3 years ago

mr-c commented 3 years ago

The creation of the Research Object Bundle is the source of much complexity and delays, but is likely not used by the majority of CWL viewer users.

The CWL viewer should be refactored so that the creation of the RO Bundle does not delay the visualization of the provided workflow. This could even mean that it is not created at all unless requested by the user via an additional click.

mr-c commented 3 years ago

Possibly stop making the research object bundle. @stain is there a reason to keep that code?

mr-c commented 3 years ago

The WorkflowHub.eu supports the successor to Research Object Bundles, RO-Crate and will even create RO-Crates when one registers a workflow there.

So I propose that we completely remove the "Download as Research Object Bundle" link and all code that supports that and isn't used for another feature.

image

The Permalink should stay; I think that is still an important feature.

mr-c commented 3 years ago

@stain writes

I don't think the RO Bundle complicates code too much as it is mainly run as side jobs. The WorkflowHub has not got the same content (yet) in the RO-Crate e.g. the TTL conversion and the packed workflow. The RO Bundle is imperfect though, e.g. it does not include ../tools/foo.cwl in the archive.

In WorkflowHub this would not be there fully before we get the GitHub integration proper by autumn 2021.

If you add a registration mechanism to WFHub then it could help to make DOIs etc

For cWL viewer I think adding visualization of tools would be nice, secondly some kind of grouping of workflows in different commits or paths of same repo.. there are now 25.000 "workflows".

cwltool can run straight off the raw.githubblabla links, but it can be a bit slow because it does not cache. Also wont be so easy in non-github repos also wont work with git submodules

now it is running in the cloud then continued shallow clone may use more traffic than to check each repo once... but you have less working disk space (before we had 1 TB)

@obasekiosa, @tetron So lets not drop the Research Object Bundle feature for now