Reputation: 501
Why does my view get created with FIRM_PANEL and FIRM_DATE as DATEs and not an NVARCHAR and a DATE as I expect?
SELECT DISTINCT piv3.CASE_NUMBER, FIRM_PANEL, FIRM_DATE
FROM
(
SELECT DISTINCT piv2.CASE_NUMBER, piv2.DETAIL_CODE,
CASE piv2.DATA_FLAGS
WHEN 'TEXT_FLAG' THEN CONVERT(NVARCHAR, TEXT_VALUE)
WHEN 'DATE_FLAG' THEN CONVERT(DATE, DATE_VALUE)
END AS 'VALUE'
FROM
(
SELECT DISTINCT CASE_NUMBER, DETAIL_CODE, piv.DATA_FLAGS, TEXT_VALUE, CONVERT(DATE, DATE_VALUE) AS DATE_VALUE
FROM table
UNPIVOT
(
YES_OR_NO
FOR DATA_FLAGS IN (DATE_FLAG, TEXT_FLAG)
) piv
) piv2
) as sourcetable
PIVOT
(
MAX(VALUE)
FOR DETAIL_CODE IN (FIRM_PANEL,
FIRM_DATE )
) AS piv3
Let me explain each nested pivot/select. The first SELECT gives you an idea how the data is store raw. I needed a way to show 1 CASE_NUMBER and DETAIL_CODEs as unique columns.
SELECT DISTINCT CASE_NUMBER, DETAIL_CODE, piv.DATA_FLAGS, TEXT_VALUE,
CONVERT(DATE, DATE_VALUE) AS DATE_VALUE
FROM table
I then needed to take DATE_VALUE and TEXT_VALUE and merge into one column called VALUE, hence my UNPIVOT.
SELECT DISTINCT piv2.CASE_NUMBER, piv2.DETAIL_CODE,
CASE piv2.DATA_FLAGS
WHEN 'TEXT_FLAG' THEN CONVERT(NVARCHAR, TEXT_VALUE)
WHEN 'DATE_FLAG' THEN CONVERT(NVARCHAR, DATE_VALUE)
END AS 'VALUE'
FROM
(
SELECT DISTINCT CASE_NUMBER, DETAIL_CODE, piv.DATA_FLAGS, TEXT_VALUE, CONVERT(DATE, DATE_VALUE) AS DATE_VALUE
FROM table
UNPIVOT
(
YES_OR_NO
FOR DATA_FLAGS IN (DATE_FLAG, TEXT_FLAG)
) piv
) piv2
Then my final SQL (top) finally PIVOTs again and removes nulls and interprets if the column was a text field then use NVARCHAR and if date use DATE.
However the only way I can SELECT from the view without errors is making both NVARCHAR or else I'll receive "Conversion failed when converting date and/or time from character string" due to both columns being a DATE field when they shouldn't be. However I do not understand why the view is created with the two DATE columns. My current workaround is setting as string and just doing the conversion when selecting from the view.
Sorry for the wordiness.
Upvotes: 0
Views: 103
Reputation: 13949
I think an-d was on the right path.. You should definitely try case aggregates.
SELECT CASE_NUMBER,
MAX( CASE WHEN DETAIL_CODE = 'FIRM_PANEL' THEN TEXT_VALUE END) AS FIRM_PANEL,
MAX( CASE WHEN DETAIL_CODE = 'FIRM_DATE' THEN CONVERT(DATE, DATE_VALUE) END) AS FIRM_DATE
FROM table
GROUP BY CASE_NUMBER
You can add more where clauses to the case expression if you need to, but the above may work.
Upvotes: 1
Reputation: 11
I don't quite understand the reasons for your use of pivots, but the problem seems to be that you're trying to cast the same column VALUE
as both DATE
and NVARCHAR
here -
...
CASE piv2.DATA_FLAGS
WHEN 'TEXT_FLAG' THEN CONVERT(NVARCHAR, TEXT_VALUE)
WHEN 'DATE_FLAG' THEN CONVERT(DATE, DATE_VALUE)
END AS 'VALUE'
...
You can't control when SQL Server performs the CONVERT
operation and it might do it before checking the DATA_FLAGS
. Your query should work if you try to move the CONVERT
outside to the first select -
SELECT DISTINCT piv3.CASE_NUMBER, CONVERT(NVARCHAR, FIRM_PANEL), CONVERT(DATE, FIRM_DATE)
FROM
(
SELECT DISTINCT piv2.CASE_NUMBER, piv2.DETAIL_CODE,
CASE piv2.DATA_FLAGS
WHEN 'TEXT_FLAG' THEN TEXT_VALUE
WHEN 'DATE_FLAG' THEN DATE_VALUE
END AS 'VALUE'
...
If you're open to not using pivots, here's a simple alternative to get the same result -
SELECT CASE_NUMBER
, MAX(CASE WHEN DATA_FLAGS = 'TEXT_FLAG' THEN TEXT_VALUE END) FIRM_DATE
, MAX(CASE WHEN DATA_FLAGS = 'DATE_FLAG' THEN CONVERT(DATE, DATE_VALUE) END) FIRM_PANEL
FROM #in
GROUP BY CASE_NUMBER
This gives the same result as you're looking for with the columns in their expected data types. Note that it assumes you will have only one record with TEXT_FLAG
and DATE_FLAG
each for every CASE_NUMBER
.
Upvotes: 0