Skip to main content
Implemented

Access to Data Model Instances in transformations

Related products:Transformations and RAWData Modeling
  • April 16, 2025
  • 3 replies
  • 74 views

Hi team,

 

We’d like to suggest an improvement for the Transformations feature.

Currently, in Transformations, we only have access to staging tables and not to the instances of the data model that have already been created. This leads to some redundancy and inefficiencies in transformation logic.

 

Use case:

Let’s say I load Well objects in a first transformation, applying a set of filters (e.g., valid wells, correct country, etc.). Then, in a second transformation, I want to load Wellbore objects. However, to ensure consistency, I want to join with only the wells that were already loaded into the data model — not reapply the same filtering logic I used earlier.

Currently, since I can’t access the already loaded Well instances in the data model, I have to duplicate the filtering logic or rely on external references, which is error-prone and inefficient.

 

Suggestion:

Allow read access to instances of the data model from within transformations, so we can:

  • Reference already created objects (e.g., Well) directly when creating new ones (e.g., Wellbore).
  • Avoid duplication of logic.
  • Ensure better consistency across related business objects.

Let me know if this is technically feasible or already planned.

Thanks a lot!

3 replies

Peter  Arwanitis
Practitioner
Forum|alt.badge.img

Hi Oussama,

You wrote:

> Currently, in Transformations, we *only* have access to staging tables and not to the instances of the data model that have already been created.

 

You're aware that you have a list of functions to read all types of CDF resource-types from within a SQL statement?

Not only classic but data-model centric too, docs starting here:

https://docs.cognite.com/cdf/integration/guides/transformation/write_sql_queries/#cdf_nodes

 

Please let me know, if this helps anyway or you meant something else?

Best regards

Peter

(=PA=)


Alex
Practitioner
  • Practitioner
  • April 23, 2025

@Oussama ALLALI - would that help based on what we discussed yesterday, or are you looking for more a “recommendation” ?


Hello ​@Peter Arwanitis thank you for your answer, ​@Alex indeed, it's really helpful.