1.2 ➡️ 1.3

In versions 0.3.0 and above, workspace 1.3 is required. As a first step, please update your workspace block to the following:

workspace:
  edition: "1.3"
  ...

​

Breaking Changes

​

Providers -> Integrations

SDF can read metadata and data from a variety of sources, including databases, data lakes, and data catalogs. In the past, we referred to databases as providers, then had confusing methods for configuring data lakes and data catalogs. Now, all of these external relationships are managed in a single, unified integrations block.

This replaces the providers block in the workspace configuration.

If you have a workspace with a provider block like this:

workspace:
  ...
  providers:
    - type: snowflake
      credential: snowflake-creds
      sources: 
        - pattern: apress.*.*

It should now look like this:

workspace:
  ...
  integrations:
    - provider: snowflake
      type: database
      credential: snowflake-creds
      sources: 
        - pattern: apress.*.*

This makes integrations much more extensible as we grow towards more data and metadata sources for local execution. For an in-depth guide on how to configure a certain integration or a description about our new integrations philosophy, see the integration documentation.

​

Goodbye Compute

Previously, the compute property was required in the defaults block to tell SDF where to run the query, i.e. local or remote. This is now inferred from the integrations block, rendering it unnecessary. It should now be removed from the defaults block.

workspace:
  ...
  defaults:
    ...
    catalog: xyz
    compute: remote

Should now look like this:

workspace:
  ...
  defaults:
    ...
    catalog: xyz

​

Information Schema V2

The information schema has been updated to include more metadata about your tables, and has been re-architected to enable performant queries on larger workspaces. If you have any checks or reports, these will need to be rewritten to use the new information schema.

For example, if you have a check that looks like this:

SELECT 
    t.table_ref
FROM 
    sdf.information_schema.tables t
JOIN 
    sdf.information_schema.columns c ON t.table_ref = c.table_ref
WHERE 
    c.classifiers LIKE '%PII%' AND t.schema_name = 'external'
GROUP BY t.table_ref;

It should now look like this:

SELECT 
    t.table_id
FROM 
    sdf.information_schema.tables t
JOIN 
    sdf.information_schema.columns c ON t.table_id = c.table_id
WHERE 
    CONTAINS_ARRAY_VARCHAR(c.classifiers, 'PII') AND t.schema_name = 'public'
GROUP BY t.table_id;

As you can see, new functions are now available for local execution, including CONTAINS_ARRAY_VARCHAR.

For a full reference on the new information schema, see the information schema documentation.

​

Goodbye .sdfcache, hello sdftarget

Based on some quality user feedback, the .sdfcache directory has been renamed to sdftarget. This is where SDF will store all of its metadata and cache files.

This shouldn’t break anything as is, but if you have any scripts or processes that rely on the .sdfcache directory, you’ll need to update them to use sdftarget instead.

Furthermore, .gitignore files should be updated to ignore sdftarget instead of .sdfcache.

​

Support for trino syntax

We’ve enabled trino to be used as an alias for trino for the workspace’s dialect property.

To see all accepted dialects, see the dialect documentation.

​

seeded: true -> cycle-cut-point: true

Due to confusion with configuring seed models and breaking cycles with the seeded property, we’d renamed this field to cycle-cut-point. This now explicitly marks this table as first table to be processed in a cycle.