ApprodxQuantile does not seem to be working with delta live tables (DLT)

[Trodenn]

HI,

I am tying to use the approxQuantile() function and populate a list that I made, yet somehow, whenever I try to run the code it's as if the list is empty and there are no values in it.

Code is written as below:

@dlt.table(name = "customer_order_silver_v2")
def capping_unitPrice_Qt():
    df =  dlt.read("customer_order_silver")
    boundary_unit = [0,0]
    boundary_qty = [0,0]
    boundary_unit = df.select(col("UnitPrice")).approxQuantile('UnitPrice',[0.05,0.95], 0.25)
 
    boundary_qty = df.select(col("Quantity")).approxQuantile('Quantity',[0.05,0.95], 0.25)
 
 
    df = df.withColumn('UnitPrice', F.when(col('UnitPrice') > boundary_unit[1], boundary_unit[1])
                                       .when(col('UnitPrice') < boundary_unit[0], boundary_unit[0])
                                       .otherwise(col('UnitPrice')))
    
    df = df.withColumn('Quantity', F.when(col('Quantity') > boundary_qty[1], boundary_qty[1])
                                       .when(col('Quantity') < boundary_qty[0], boundary_qty[0])
                                       .otherwise(col('Quantity')))
                                          
    return df

The output that I get when running is below:

 

Am I missing something somewhere? any advice or ideas are welcomed.

[Hubert-Dudek · Accepted solution]

Maybe try to use (and the first test in the separate notebook) standard df = spark.read.table("customer_order_silver") to calculate approxQuantile.

Of course, you need to set that customer_order_silver has a target location in the catalog, so read using regular spark.read will work.

[Trodenn]

I see what you are suggesting, if I were to run it in the same notebook but in a different cell that is not a @dlt.table, will it work? I need to determine the quantiles and then use that to make changes to the table so that is why.

To read a delta live table do I just use spark.read.table("customer_order_silver")?

[Hubert-Dudek]

It will work inside def capping_unitPrice_Qt() I am using precisely the same approach.

To read a delta live table do I just use spark.read.table("customer_order_silver")?

Yes, if the table is registered in metastore. Usually, you prefix it with a database/schema name (so database.customer_order_silver). It is specified in DLT setting what is the name of the database.

[Trodenn]

what if this is not a database but another delta live table? do correct me if its the same thing. I really just started learning this tool and spark

[Trodenn]

So I tried running the code inside the dlt function, it tells me that I cannot find the table now. Do I need to do anything to make it kknow where the table is? like add the path to it?