Arvados

h1. Metadata

Input and output documents gain parallel "input metadata" and "output metadata" documents.

The input metadata is parallel to the input document, except that it describes metadata about each parameter.  For example, for this input:

<pre>

foo:

  class: File

  location: keep:abc+123/foo.bam

</pre>

the metadata might be:

<pre>

foo:

  xyz:sample: sample1122333

  xyz:sequencer: illumina

</pre>

On workflow start, reverse lookup PDH to UUID and find all tags.  Create input metadata document.

On workflow end, create output metadata document based on "arv:PropagateMetadata".

<pre>

hints:

  arv:PropagateMetadata:

    setMetadata:

      - outputParameter: bar

        valueFrom: $(metadata.foo)

</pre>

Resulting in output:

<pre>

bar:

  class: File

  location: keep:abc+123/result.vcf

</pre>

with output metadata:

<pre>

bar:

  xyz:sample: sample1122333

  xyz:sequencer: illumina

</pre>

Arvados then creates tags to the output collection based on output metadata document.  There are several possible representations (with mapping to RDF triples model)

<pre>

link_class: tag

name: xyz:sample

head_uuid: zzzzz-4zz18-zzzzzzzzzzzzzzz

properties: 

  value: sample1122333

  path: result.vcf

</pre>

* head_uuid+properties[path] -> subject

* name -> predicate

* properties[value] -> object

Alternate representation:

<pre>

link_class: xyz:samplename

name: sample1122333

head_uuid: zzzzz-4zz18-zzzzzzzzzzzzzzz

properties: 

  path: result.vcf

</pre>

* head_uuid+properties[path] -> subject

* link_class -> predicate

* name -> object

<pre>

link_class: xyz:samplename

name: result.vcf

head_uuid: zzzzz-4zz18-zzzzzzzzzzzzzzz

properties: 

  value: sample1122333

</pre>

* head_uuid+name -> subject

* link_class -> predicate

* properties[value] -> object