Business Context
A data science team maintains a 'recommendationEngine' as a CAS action set. This engine is updated frequently. A CI/CD pipeline must be able to redeploy the action set and overwrite the existing definition table in-place to ensure dependent applications always have the latest metadata.
Data Preparation
Define the initial version (v1) of the recommendation engine action set.
Copied!
| 1 | PROC CAS; |
| 2 | ACTION BUILTINS.defineActionSet / |
| 3 | name='recommendationEngine' |
| 4 | definition={{ |
| 5 | ACTION='getRecommendations', |
| 6 | parms={{name='userId', type='string'}}, |
| 7 | code='print("v1: Getting recommendations for user " .. _parms_["userId"]);' |
| 8 | }}; |
| 9 | RUN; |
| 10 | QUIT; |
Étapes de réalisation
1
Initial deployment: Convert the v1 action set to a table named 'RECOMMENDER_METADATA'.
Copied!
| 1 | PROC CAS; |
| 2 | ACTION BUILTINS.actionSetToTable / |
| 3 | actionSet='recommendationEngine' |
| 4 | casOut={name='RECOMMENDER_METADATA', caslib='casuser'}; |
| 5 | RUN; |
| 6 | QUIT; |
2
Simulate an update: Define a new version (v2) of the action set with the same name but modified code.
Copied!
| 1 | PROC CAS; |
| 2 | ACTION BUILTINS.defineActionSet / |
| 3 | name='recommendationEngine' |
| 4 | definition={{ |
| 5 | ACTION='getRecommendations', |
| 6 | parms={{name='userId', type='string'}, {name='maxResults', type='int', defaultValue=10}}, |
| 7 | code='print("v2: Getting top " .. _parms_["maxResults"] .. " recommendations for user " .. _parms_["userId"]);' |
| 8 | }}; |
| 9 | RUN; |
| 10 | QUIT; |
3
Automated redeployment: Convert the updated (v2) action set, overwriting the existing table using 'replace=true'.
Copied!
| 1 | PROC CAS; |
| 2 | ACTION BUILTINS.actionSetToTable / |
| 3 | actionSet='recommendationEngine' |
| 4 | casOut={name='RECOMMENDER_METADATA', caslib='casuser', replace=true}; |
| 5 | RUN; |
| 6 | QUIT; |
4
Verification: Fetch the table content to confirm it now reflects the v2 definition, including the new 'maxResults' parameter.
Copied!
| 1 | PROC CAS; |
| 2 | ACTION TABLE.fetch / |
| 3 | TABLE='RECOMMENDER_METADATA'; |
| 4 | RUN; |
| 5 | QUIT; |
Expected Result
The action runs successfully in both instances. After step 3, the 'RECOMMENDER_METADATA' table is overwritten. The final fetch in step 4 shows the v2 definition of the 'getRecommendations' action, which now includes the 'maxResults' parameter, confirming the lifecycle update was successful.