Business Context
A data science team runs a 'model factory' process that generates hundreds of individual forecast tables daily. To manage this output, they need an efficient, batch-oriented method to tag each new table with its corresponding model metadata (Model ID, Algorithm, Target Variable) from a central control table.
Data Preparation
Create a control table containing the attributes for multiple forecast tables. Then, create several empty forecast tables to simulate the output of a model factory.
Copied!
| 1 | DATA casuser.model_metadata; |
| 2 | LENGTH _tableName_ $ 32 _caslib_ $ 8 _attrName_ $ 20 _attrValue_ $ 50; |
| 3 | INFILE DATALINES dsd; |
| 4 | INPUT _tableName_ $ _caslib_ $ _attrName_ $ _attrValue_ $; |
| 5 | DATALINES; |
| 6 | FORECAST_STORE101, CASUSER, ModelID, F-S101-20251126 |
| 7 | FORECAST_STORE101, CASUSER, Algorithm, ARIMA |
| 8 | FORECAST_STORE101, CASUSER, Target, Sales |
| 9 | FORECAST_STORE205, CASUSER, ModelID, F-S205-20251126 |
| 10 | FORECAST_STORE205, CASUSER, Algorithm, ESM |
| 11 | FORECAST_STORE205, CASUSER, Target, Inventory |
| 12 | FORECAST_STORE330, CASUSER, ModelID, F-S330-20251126 |
| 13 | FORECAST_STORE330, CASUSER, Algorithm, ARIMA |
| 14 | FORECAST_STORE330, CASUSER, Target, Sales |
| 15 | ; |
| 16 | RUN; |
| 17 | |
| 18 | DATA casuser.FORECAST_STORE101; x=1; RUN; |
| 19 | DATA casuser.FORECAST_STORE205; x=1; RUN; |
| 20 | DATA casuser.FORECAST_STORE330; x=1; RUN; |
Étapes de réalisation
1
Load the metadata control table and the empty forecast tables into CAS.
Copied!
| 1 | PROC CAS; |
| 2 | TABLE.loadTable / caslib='CASUSER' path='model_metadata.sashdat' casOut={name='MODEL_METADATA', replace=true}; |
| 3 | TABLE.loadTable / caslib='CASUSER' path='FORECAST_STORE101.sashdat' casOut={name='FORECAST_STORE101', replace=true}; |
| 4 | TABLE.loadTable / caslib='CASUSER' path='FORECAST_STORE205.sashdat' casOut={name='FORECAST_STORE205', replace=true}; |
| 5 | TABLE.loadTable / caslib='CASUSER' path='FORECAST_STORE330.sashdat' casOut={name='FORECAST_STORE330', replace=true}; |
| 6 | RUN; |
2
Use a single 'attribute' action call with the 'table' parameter to add attributes to all forecast tables from the metadata control table.
Copied!
| 1 | PROC CAS; |
| 2 | TABLE.attribute / |
| 3 | task='ADD' |
| 4 | caslib='CASUSER' |
| 5 | TABLE='MODEL_METADATA'; |
| 6 | RUN; |
3
Verify the attributes for one of the tables by exporting them.
Copied!
| 1 | PROC CAS; |
| 2 | TABLE.attribute / |
| 3 | task='EXPORT' |
| 4 | name='FORECAST_STORE205' |
| 5 | caslib='CASUSER' |
| 6 | TABLE='VERIFY_ATTRIBUTES'; |
| 7 | |
| 8 | TABLE.fetch / TABLE='VERIFY_ATTRIBUTES'; |
| 9 | RUN; |
| 10 | QUIT; |
Expected Result
The action should successfully add attributes to all three forecast tables in a single call. The final fetch on 'VERIFY_ATTRIBUTES' should display the three attributes ('ModelID', 'Algorithm', 'Target') for the 'FORECAST_STORE205' table, confirming the batch operation was successful.