RODBC issue casting NA's to Float for SQL Server

Question

I'm trying to load a dataframe to a SQL Server 2014 table and am getting:

[Microsoft][SQL Server Native Client 11.0]Invalid character value for cast specification

I'm pretty sure I've narrowed it down to the fact that I'm trying to cast the numeric columns which have NA's to float.

dtype = c(Brand ="nvarchar(100)", Date_Loaded="date", Period="int", col1="float")

sqlSave(con, df, "dbo.test", rownames=FALSE, append=TRUE, verbose=TRUE, varTypes=dtype)

Here is some sample data:

| Brand  | Date_Loaded | Period | col1        |
|--------|-------------|--------|-------------|
| Cereal |   9/18/2015 |    1   |    NA       |
| Cereal |   9/18/2015 |    2   |    NA       |
| Cereal |   9/18/2015 |    3   |    NA       |
| Cereal |   9/18/2015 |    4   |    0.127418 |
| Cereal |   9/18/2015 |    5   |    0.167645 |
| Cereal |   9/18/2015 |    6   |    0.144336 |

I've tried playing with nastring=NULL, NA, '' with no luck. My option now is to just stage it as varchar (which works) and clean it up in SQL but that seems too hacky.

Answer 1

Crappy work around is to just bring everything in as varchar(20) and then put a view on the table that checks for NA and casts to float.

dtype = c(Brand ="nvarchar(100)", Date_Loaded="date", Period="int",col1="varchar(20)")

sqlSave(con, df, "dbo.test", rownames=FALSE, append=TRUE, varTypes=dtype)

create view dbo.test_vw as
SELECT
[Brand]
,[Date_Loaded]
,[Period]
,(case ISNUMERIC(col1) WHEN 1 then cast(col1 as float) else null end) as col1
FROM
dbo.test;