proc sql left join error - column could not be found

[csetzkorn]

I am getting this error:

Column SomeColumn could not be found in the table/view identified with the correlation name TRANS.

Using this code

proc sql;
CREATE TABLE WORK.Test AS
select
Trans.SomeColumn 
from WORK.Joiner AS J
left outer join work.parameterestimatestrans AS Trans
ON J._NAME_ = Trans._NAME_;
run;

I know that SomeColumn is not in the table aliased Trans. However, a left out join should simply return a null (i.e. . in SAS). So I am a bit confused about the non sql sas behaviour. Any ideas?

[Kurt_Bremser]

The error is exactly what should happen. You explicitly named a column in the select that does not exist, so your select statement is useless.

SQL does not automatically create variables with missing values, as the data step does.

[csetzkorn]

Would you mind creating the datastep equivalent please?

[ballardw]

---

@csetzkorn wrote:
Would you mind creating the datastep equivalent please?

---

It might help if you describe what the output should be. From the code posted it only appears that you want a data set with single variable and all values missing.

[Reeza]

Here, I fixed it for you, with a numeric and character missing. 

proc sql;
CREATE TABLE WORK.Test AS
select
" " as SomeColumn,
. as NumSomeColumn
from WORK.Joiner AS J
left outer join work.parameterestimatestrans AS Trans
ON J._NAME_ = Trans._NAME_;
run;

[csetzkorn]

Thanks getting closer. The challenge I have that work.parameterestimatestrans may contain NumSomeColumn. I only want to set NumSomeColumn to 'null'/. if work.parameterestimatestrans does not exist.

[csetzkorn]

I found a solution to my problem. Using:

data work.parameterestimatestrans;
    set work.parameterestimatestrans;
	NumSomeColumn = coalesce(NumSomeColumn,.);
run;

Will impute the variable should it be missing and the left join picks this up.

[Kurt_Bremser]

What do you mean by

---

@csetzkorn wrote:
 if work.parameterestimatestrans does not exist.

---

- the dataset does not exist

- or the dataset exists, but there is no matching record in it

- or the column does not exist in the dataset

?

[csetzkorn]

I meant if NumSomeColumn does not exist in work.parameterestimatestrans

[Reeza]

If it doesn't exist the COALESCE doesn't do anything. 

It assigns a missing value. 

[Kurt_Bremser]

Since the OP has finally stated his/her true intentions, I can say that the data step with coalesce() is the correct solution.

The OP wants to reach a state where the variable will be present, no matter what is in the original dataset.

If the variable is not there, it will be created as missing, if it is there, the values will be preserved.

[LinusH]

There is no exact equivalent between SQL and the data step. Isn't it easier to fix the existing query?
If you want to create a column with missing values specify it as a constant in the select clause.