Skip to content

Modes

NEAT supports two modes of data modeling: additive and rebuild. Depending on the selected mode, the behavior of NEAT when deploying changes to a data model in Cognite Data Fusion (CDF) will vary. NEAT classifies changes to the data model into three severity levels:

  • SAFE: allowed changes that do not impact existing data or functionality
  • ⚠️ WARNING: allowed changes where caution is advised, as they may have implications
  • BREAKING: disallowed changes, which would not be applied to CDF

In the following sections, we describe the characteristics of each mode and provide detailed tables outlining the allowed and disallowed changes for each severity level.

Changes can be done only in the data model space

It is important to know that NEAT will only consider changes to the data model components (aka DMS schema components) in the same space as the data model itself. Any change outside of the data model space will be ignored. The reason for this is to avoid accidental changes to data models that have different owners or are managed by different teams.

Additive Mode#

Key characteristics of the additive mode are:

  • Forward compatibility: old versions of applications and services can work with the updated data model without requiring changes.
  • Incremental changes: data model changes are applied incrementally, allowing for gradual evolution of the data model.
  • Suitable for production environments: this mode is ideal for production systems where stability and forward compatibility are crucial.

The following table summarizes the allowed and disallowed changes in additive mode:

Severity Operation Change Description Example
SAFE Create Data model -
SAFE Create View Entire views, new properties, filters
SAFE Create Container Entire containers, new properties, indexes and constraints
SAFE Create New enums to existing enum properties -
SAFE Update Data model metadata Name, description, version
SAFE Update Data model's view order -
SAFE Update View metadata Name, description, version
SAFE Update Container metadata Name, description
SAFE Update Property metadata Name, description
SAFE Update Enums metadata Name, description
WARNING Update View filters -
WARNING Update Edge type -
WARNING Update Source property for reverse connection Also known as through in CDF API
WARNING Remove Container constraints -
WARNING Remove Container indexes -
WARNING Update Mutability of container properties -
WARNING Update Auto-increment of container properties -
WARNING Update Default value of container properties -
WARNING Update Max count of listable properties -
WARNING Remove Container property unit Removing meter from a property
WARNING Add Container property unit Adding meter to a property that did not have a unit before
BREAKING Delete Existing data model -
BREAKING Delete View reference from the data model -
BREAKING Delete Existing view -
BREAKING Add Implements to an existing view -
BREAKING Delete Property from an existing view -
BREAKING Update Mapping of view (property) to container (property) -
BREAKING Update Direct/reverse connection value type Also known as source in CDF API
BREAKING Update Edge source of properties Setting new view as source of properties
BREAKING Update Edge direction -
BREAKING Update Connection type From edge to direct connection
BREAKING Update What container is used for Switching usedFor from node to edge
BREAKING Update Value type of container property From text to integer
BREAKING Update Container constraint definitions uniqueness constraint property bySpace cannot be changed
BREAKING Update Container index definitions Adding or removing properties to which the index applies
BREAKING Update Nullability of container properties -
BREAKING Update Container property unit Changing from meter to kilometer
BREAKING Update maxTextSize of text properties -
BREAKING Delete Enum values from existing enum properties -

Rebuild Mode#

Unlike additive mode, rebuild mode allows every change to be applied to the data model, as it is assumed that the entire data model will be rebuilt from scratch. This mode is suitable where there is no existing data or applications depending on the current data model, such as in development or testing environments, since every component of the data model will be first deleted and then recreated. In case when containers that are to be deleted contain data, deployment in the rebuild mode will fail, and NEAT will not accidentally remove data. However, if you are intentionally looking to delete existing data, you will have to set drop_data=True when deploying the data model via command neat.physical_data_model.write.cdf(dry_run = False, drop_data = True).