Search…
Data diff in CI
dbt integration with Datafold

Prerequisites

Make sure Datafold GitHub integration is set up.

Tag primary keys in dbt model

Datafold needs to know which column is the primary key of the table to perform the diff. When Datafold cannot determine the primary key of the two tables to diff, and will produce an error:
When setting up the CI integration, one of the steps is proving the primary-key tag:
This tag we can use in the dbt metadata to let Datafold know which column can be used to perform the diff. Datafold supports composite primary keys, meaning that you can assign multiple columns that make up the primary key together. There are three ways of doing this, which we'll discuss next:

Metadata

The first one is setting the tag in the dbt metadata. We set the primary key tag to primary-keyso we use this in the metadata.
Table metadata can also be used to specify per-model diff options. In the example below diff is configured to compare only rows matching user_id > 2350. The expression in the filter is an SQL expression and can be anything you could put into where clause when selecting from the tables.
1
models:
2
- name: users
3
meta:
4
datafold:
5
datadiff:
6
filter: "user_id > 2350"
7
8
columns:
9
- name: user_id
10
meta:
11
primary-key: true
12
# In the case of a compound Primary Key, you can assign a second column (or more)
13
- name: version_id
14
meta:
15
primary-key: true
Copied!

Tags

If the primary key is not found in the metadata, it will go through the tags.
1
models:
2
- name: users
3
columns:
4
- name: user_id
5
tags:
6
- primary-key
7
# In the case of a compound Primary Key, you can assign a second column (or more)
8
- name: version_id
9
tags:
10
- primary-key
Copied!

Inferred

If the primary key isn't provided explicitly, Datafold will try to assume a pk from dbt's uniqueness tests. If you have a single column uniqueness test defined, it will use this column as the PK:
1
models:
2
- name: users
3
columns:
4
- name: user_id
5
tests:
6
- unique
Copied!
Also, model level uniqueness tests are used for inferring the PK:
1
models:
2
- name: sales
3
columns:
4
- name: col1
5
- name: col2
6
...
7
tests:
8
- unique:
9
column_name: "col1 || col2"
10
11
- name: sales
12
columns:
13
- name: col1
14
- name: col2
15
...
16
tests:
17
- unique:
18
column_name: "CONCAT(col1, col2)"
Copied!
Finally, we also support unique_combination_of_columns from the dbt_utils package:
1
models:
2
- name: users
3
columns:
4
- name: order_no
5
- name: order_line
6
...
7
tests:
8
- dbt_utils.unique_combination_of_columns:
9
combination_of_columns:
10
- order_no
11
- order_line
Copied!
Keep in mind that this is a failover mechanism. If you change the uniqueness test, this will also impact the way Datafold performs the diff.

Checking primary key annotations

You can check what models in your dbt repo already have primary key annotations, and which need more attention. You'll need to install Datafold SDK and configure access parameters:
1
$ pip3 install 'datafold-sdk>=0.0.6'
2
3
# skip this step if you are using app.datafold.com
4
$ export DATAFOLD_HOST=https://<hostname>
5
6
# get your API key in Datafold UI -> Edit Profile -> API Key
7
$ export DATAFOLD_APIKEY=RSSQrpfddSEEK8WVtc0zd27f9nsdhPU3AxZ
8
Copied!
After that you need to compile manifest.json and you are ready to do the check:
1
# Lookup your CI configuration id in URL when you go to Settings -> CI settings -> <name>:
2
# https://app.datafold.com/settings/ci_integrations/14
3
4
$ datafold dbt check-primary-keys --ci-config-id 14 manifest.json
5
meta dbt_snowflake.service_calls INCIDENT_NUMBER models/service_calls.sql models/schema.yml
6
meta dbt_snowflake.supply_of_ones ID models/supply_of_ones.sql models/schema.yml
7
none dbt_snowflake.fokko.boom models/fokko/boom.sql
8
none dbt_snowflake.new_service_calls models/new_service_calls.sql models/schema.yml
9
tags dbt_snowflake.ephemeral_supply_of_twos ID models/ephemeral_supply_of_twos.sql models/schema.yml
10
uniqueness dbt_snowflake.new_service_calls_concat2 CAL_YEAR, INCIDENT_NUMBER models/new_service_calls_concat2.sql models/schema.yml
11
uniqueness dbt_snowflake.supply_of_twos ID models/supply_of_twos.sql models/schema.yml
12
Copied!
The first column shows how the key was inferred:
  • none - Datafold was unable to find any PKs,
  • uniqueness - primary keys were derived from uniqueness tests,
  • tags - PKs were specified with column-level tags,
  • meta - column-level metadata was used,
  • meta_table - table-level metadata.
Out of those, none and possibly uniqueness require further actions.
The other fields in the printout are:
  • fully qualified name of dbt model,
  • list of primary keys,
  • sql file that contains model definition,
  • "patch" yml file that has dbt configuration of the model.

dbt metadata synchronization

Datafold integrates very well with dbt, and also has the ability to ingest the metadata provided by dbt automatically. dbt models has metadata that can be synchronized from the production branch into the Datafold catalog. When a table has metadata being synchronized using dbt, user editing is no longer permitted for that entire table. This is to ensure that there is a single source of truth.
Metadata can be applied both on a table and column level:
1
models:
2
- name: users
3
description: "Description of the table"
4
meta:
6
foo: bar
7
tags:
8
- pii
9
- abc
10
columns:
11
- name: user_id
12
tags:
13
- pk
14
- id
15
meta:
16
pk: true
17
- name: email
18
description: "The user's email"
19
tags:
20
- pii
21
meta:
22
type: email
23
Copied!
There are two special meta types:
  • owner: Used to specify the owner of the table and applies the owner of the table in the catalog view
  • <pk_tag>: The tag/name that is configured to identify primary columns is not synchronized into the meta-information, but it is synchronized as a tag if it exists.
So for the above table:
  • description is synchronized into the description field of the table in the catalog.
  • The owner of the table is set to the user identified by the [email protected] field. This user must exist in Datafold with that email.
  • The foo meta information is added to the description field with the value bar
  • The tags pii and bar are applied to the table as tags.
For the columns above:
  • The column user_id has two tags applied: pk and id
  • The metadata for user_id is ignored, because it reflects the primary key tag.
  • The email column has the description applied.
  • The email column has the tag pii applied
  • The email column has extra metadata information in the description field: type with the value email.
Metadata synchronization occurs in one of two methods:
  • The meta_schedule is set for the dbt cloud integration. This will run according to the specified cron schedule, find the most recent dbt cloud production run, and synchronize the metadata from there.
  • It can also be configured to synchronize metadata whenever a push to production happens.
Last modified 1mo ago