null values in the WRROC metadata

[simleo]

• I built and installed the plugin with make install

• I updated nextflow to version 25.04.6

• I added workflow.output.mode = 'copy' to nextflow.config at the top level (so that the RO-Crate gets a copy of the files rather than symlinks)

• I ran:

$ nextflow run nf-prov-test/ -plugins nf-prov@1.5.0

 N E X T F L O W   ~  version 25.04.6

Launching `nf-prov-test/main.nf` [prickly_cantor] DSL2 - revision: a16de98cd6

WARN: WORKFLOW OUTPUT DEFINITION IS A PREVIEW FEATURE - SYNTAX AND FUNCTIONALITY CAN CHANGE IN FUTURE RELEASES
executor >  local (7)
[20/8b0b9a] ECHO_SCRIPT (r1) [100%] 3 of 3 ✔
[33/6d7f4f] ECHO_EXEC (r3)   [100%] 3 of 3 ✔
[dc/fd4755] WC_SAMPLE (1)    [100%] 1 of 1 ✔
Staging foreign file: https://openstack.cebitec.uni-bielefeld.de:8080/swift/v1/denbi-mg-course/read1.fq.gz
Staging foreign file: https://openstack.cebitec.uni-bielefeld.de:8080/swift/v1/denbi-mg-course/read2.fq.gz
WARN: Could not determine pipeline repository URL for Workflow Run RO-Crate -- launch the pipeline with canonical URL (e.g. `nextflow run nextflow-io/hello`) to ensure that the pipeline repository URL is recorded in the crate
WARN: Could not identify process definition for `ECHO_SCRIPT` -- resulting RO-Crate may be invalid (hint: define each process in a separate module script to fix this issue)
WARN: Could not identify process definition for `WC_SAMPLE` -- resulting RO-Crate may be invalid (hint: define each process in a separate module script to fix this issue)

In results/ro-crate-metadata.json, some CreateActions have an instrument pointing to null (instead of pointing to the relevant SoftwareApplication); similarly, some HowToSteps have a null workExample. Due to this, the crate is invalid. This can be checked using the validator:

rocrate-validator -y validate -l REQUIRED -p provenance-run-crate -v -w 79 -o err.txt ~/repos/nf-prov/results

Also, there is no trace of the executions of ECHO_EXEC. I'm attaching the ro-crate-metadata.json file.

ro-crate-metadata.json

[fbartusch]

@bentsherman I will check on it.

[fbartusch]

The underlying problem is this line:

final processDef = defs.size() == 1 ? defs.first() : null

The nf-prov-test pipeline defines three processes in the main workflow file.
This means that defs.size() != 1, and therefore instrument is null.

A possible fix would be to replace that line with:

final processDef = defs.find { it.name == processor.name }

With this change, the generated ro-crate-metadata.json looks like:

[...]
"name": "ECHO_SCRIPT (r2)",
 "instrument": {
  "@id": "#module/ECHO_SCRIPT"
},
[...]

That’s an improvement. However, I’m not completely satisfied that the instrument is named #module/ECHO_SCRIPT, since in this case it’s actually a Nextflow Process, not a Module.

I suspect I only tested the plugin with nf-core pipelines, which use modules exclusively rather than standalone processes.