Github user fmcquillan99 commented on the issue:

    https://github.com/apache/madlib/pull/291
  
    In cols2vec,
    
    For this table:
    ```
    CREATE TABLE golf (
        id integer NOT NULL,
        "OUTLOOK" text,
        temperature double precision,
        humidity double precision,
        "Temp_Humidity" double precision[],
        clouds_airquality text[],
        windy boolean,
        class text,
        observation_weight double precision
    );
    ```
    
    this fails:
    ```
    SELECT madlib.cols2vec(
        'golf',
        'cols2vec_result',
        'id, temperature'
    );
    ```
    because `id` is INT and `temperature` is FLOAT.
    
    It forces the user to do:
    ```
    SELECT madlib.cols2vec(
        'golf',
        'cols2vec_result',
        'id::FLOAT, temperature'
    );
    ```
    but this is inconvenient especially if you have a big 
    table and are using '*' to get all columns into the feature
    vector and they are a mix of numeric types.  
    
    Also a mix of VARCHAR and TEXT fails in a similar way
    but should not.
    
    Use PostgreSQL precendence rules to fix this please.
    
    



---

Reply via email to