Update relationships

[andrewelamb]

Problem:

The Data Model documentation wasn't quite accurate.
The Parent column wasn't described.
The Validation Rules column wasn't listed as required.
These columns were required by the DataModelRelationships class, but aren't for JSON Schema creation:

• DependsOn Component
• Properties
• Source

Solution:

The documentation has been updated.

These columns are no longer required:

• DependsOn Component
• Properties
• Source

Note: I tried to make ValidationRules not required, but doing so broke a lot of tests. There may be a bug here. This will need to be resolved in SYNPY-1693

[BryanFauble]

1 comment but lgtm

[linglp]

Could you confirm whether this step is required? If I remember correctly, the Parent column is only used for data visualization, so it shouldn’t be required.

[linglp]

oh never mind. I tried without Parent column and got an error:

 File "/Users/lpeng/code/synapsePythonClient/synapseclient/extensions/curator/schema_generation.py", line 606, in check_schema_definition
    raise ValueError(
    ...<2 lines>...
    )
ValueError: Schema extension headers: {'DependsOn', 'Format', 'Description', 'Source', 'DependsOn Component', 'Pattern', 'Valid Values', 'Maximum', 'Required', 'Minimum', 'Attribute', 'Properties', 'columnType', 'Validation Rules'} do not match required schema headers: ['Attribute', 'Description', 'Valid Values', 'DependsOn', 'Required', 'Parent', 'Validation Rules']
(

[andrewelamb]

It's required in the sense that it's how we differentiate between actual data types and other things that have attributes in the dependsOn column but are only there for conditional dependencies. If you go down to the Conditional Dependencies section example, Patient is an actual data type, but Cancer isn't. They both have
dependsOn values. So the way to tell them apart is by setting the Parent value. This is used by generate_json_schema when the user doesn't submit a data type, which means to create a JSON Schema for every data type.

It's also required in the sense that the DataModelrelationships class throws an error if the column isn't present in the CSV.

[andrewelamb]

I think for Valid Values, the parent colunm also needs to be filled

It doesn't for the purpose of creating JSON Schema (I just tried).

[linglp]

Okay got it. Thank you for explaining.. Maybe we should just change Parent column to DataType and just fill it with True or False in the future.

[andrewelamb]

That would probably be the best way to do it if we were starting from scratch :)

[linglp]

LGTM! Thank you for your hard work. I left some minor comments.