SQL Server 2012 Concatenating Strings from Multiple Rows - sql-server

[Edit] Due to time constraints I gave up on using a CTE and created a function that returns the concatenated string:
CREATE FUNCTION fn_GetCategoryNamesAsString
(
#lawID INT
)
RETURNS NVARCHAR(MAX)
AS
BEGIN
DECLARE #categoryNames NVARCHAR(MAX)
SET #categoryNames = ''
DECLARE #categoryID INT
DECLARE CUR CURSOR LOCAL FORWARD_ONLY READ_ONLY FOR
SELECT t1.LawCategoryID FROM [GWS].[dbo].[GWSMasterLawsLawCategories] t1 WHERE t1.LawID = #lawID
OPEN CUR
FETCH FROM CUR INTO #categoryID
WHILE ##FETCH_STATUS = 0
BEGIN
SET #categoryNames = #categoryNames + (SELECT t2.Name
FROM [GWS].[dbo].GWSMasterLawCategories t2
WHERE t2.LawCategoryID = #categoryID) + ', '
FETCH NEXT FROM CUR INTO #categoryID
END
CLOSE CUR
DEALLOCATE CUR
RETURN #categoryNames
END
GO
This does the job but I don't really like it. If anyone has a better solution I'd love to know.
[End edit]
I have seen several questions the deal roughly with the same topic but none cover the inclusion of null values.
I am writing a query that should return the full contents of one table with a couple of columns added with relevant data from other tables. These columns can include 0 - n values.
Null values need to be stored as an empty string and sets that do have the extra data should display it separated by commas.
Some approaches delivered all the names strung together, some only returned the values separately, some no values at all and, most often, the recursion went to deep (which means I fouled up as the dataset is small).
This is my current approach:
DECLARE #categoryNames NVARCHAR(MAX);
SET #categoryNames = '';
WITH sources (sourcesLawSourceID, sourcesName) AS (
SELECT DISTINCT [LawSourceID], [name]
FROM [GWS].[dbo].[GWSMasterLawSources]
),
categories AS(
SELECT GWSCategories.LawCategoryID AS categoryID, GWSLawCategories.LawID AS lawID,
categoryNames = #categoryNames
--CAST(LEFT( GWSCategories.name, CHARINDEX(',', GWSCategories.name + ',') -1) AS NVARCHAR(MAX)) categoryName,
--STUFF(GWSCategories.name, 1, CHARINDEX(',', GWSCategories.name + ','), '') categoryNames
FROM [GWS].[dbo].[GWSMasterLawCategories] GWSCategories
JOIN [GWS].[dbo].[GWSMasterLawsLawCategories] GWSLawCategories
ON GWSCategories.LawCategoryID = GWSLawCategories.LawCategoryID
UNION ALL
SELECT categories.categoryID, categories.lawID,
CAST(LEFT( #categoryNames, CHARINDEX(',', #categoryNames + ',') -1) AS NVARCHAR(MAX)) + GWSCategories.Name
FROM categories
JOIN [GWS].[dbo].[GWSMasterLawCategories] GWSCategories
ON categories.categoryID = GWSCategories.LawCategoryID
WHERE #categoryNames > ''
)
SELECT DISTINCT GWSMaster.[LawID]
,[Name]
,sources.sourcesName LawSourceName
,(SELECT STUFF((SELECT DISTINCT ', ' + RTRIM(LTRIM(categories.CategoryNames))
FROM categories
FOR XML PATH ('')), 1, 1, '')) Categories
FROM [GWS].[dbo].[GWSMasterLaws] GWSMaster
JOIN sources
ON sources.sourcesLawSourceID = GWSMaster.LawSourceID
JOIN categories
ON categories.lawID = GWSMaster.LawID
This leaves the category name field completely empty.
If I can give any more information or I have missed a question that answers my problem please let me know.

Related

SQL Server 2012 - Manipulate string data for stored procedure

Can you give me some pointers (or point in the right direction on what search terms for google)? In a stored procedure I have a parameter #TAG (string). I receive '(14038314,14040071)' (for example) from another application that cannot be altered. In the stored procedure, I need to split apart '(14038314,14040071)' to put quotes around each string value, rebuild it, strip out the outer quotes,strip out the parens and pass it to #TAG in the query below so that it looks like the line commented out below?
SELECT
V.NAME AS VARIETY, TAGID
FROM
mfinv.dbo.onhand h
INNER JOIN
mfinv.dbo.onhand_tags t on h.onhand_id = t.onhand_id
INNER JOIN
mfinv.dbo.onhand_tag_details d on t.onhand_tag_id = d.onhand_tag_id
INNER JOIN
mfinv.dbo.FM_IC_PS_VARIETY V ON V.VARIETYIDX = d.VARIETYIDX
LEFT JOIN
mfinv.dbo.FM_IC_TAG TG ON TG.TAGIDX = t.TAGIDX
WHERE
h.onhand_id = (SELECT onhand_id FROM mfinv.dbo.onhand
WHERE onhand_id = IDENT_CURRENT('mfinv.dbo.onhand'))
AND TG.ID IN (#TAG)
--AND TG.ID IN ('14038314','14040071')
You can Use Dynamic SQL Like This
DECLARE #TAG Nvarchar(MAX)='14038314,14040071'
set #TAG=''''+REPLACE(#TAG,',',''',''')+''''
--select #TAG
DECLARE #SQL NVARCHAR(MAX)=N'
Select V.NAME AS VARIETY, TAGID
FROM mfinv.dbo.onhand h
INNER JOIN mfinv.dbo.onhand_tags t on h.onhand_id = t.onhand_id
INNER JOIN mfinv.dbo.onhand_tag_details d on t.onhand_tag_id = d.onhand_tag_id
INNER JOIN mfinv.dbo.FM_IC_PS_VARIETY V ON V.VARIETYIDX = d.VARIETYIDX
LEFT JOIN mfinv.dbo.FM_IC_TAG TG ON TG.TAGIDX = t.TAGIDX
WHERE h.onhand_id = (SELECT onhand_id FROM mfinv.dbo.onhand
WHERE onhand_id = IDENT_CURRENT (''mfinv.dbo.onhand''))
AND TG.ID IN ('+#TAG+')'
PRINT #SQL
EXEC (#SQL)
Here's what I did. Thank you all for responding. Thanks to dasblinkenlight for answering "How to replace first and last character of column in sql server?". Thanks to SQLMenace for answering "How Do I Split a Delimited String in SQL Server Without Creating a Function?".
Here's how I removed parenthesis:
#Tag nvarchar(256)
SET #Tag = SUBSTRING(#Tag, 2, LEN(#Tag)-2)
Here's how I split and rebuilt #Tag:
AND TG.ID in
(
SELECT SUBSTRING(',' + #Tag + ',', Number + 1,
CHARINDEX(',', ',' + #Tag + ',', Number + 1) - Number -1)AS VALUE
FROM master..spt_values
WHERE Type = 'P'
AND Number <= LEN(',' + #Tag + ',') - 1
AND SUBSTRING(',' + #Tag + ',', Number, 1) = ','
)

HierarchyID: Get all descendants for a list of parents

I have a list of parent ids like this 100, 110, 120, 130 which is dynamic and can change. I want to get all descendants for specified parents in a single set. To get children for a single parent I used such query:
WITH parent AS (
SELECT PersonHierarchyID FROM PersonHierarchy
WHERE PersonID = 100
)
SELECT * FROM PersonHierarchy
WHERE PersonHierarchyID.IsDescendantOf((SELECT * FROM parent)) = 1
Have no idea how to do that for multiple parents. My first try was to write something like several unions, however I'm sure that there should be smarter way of doing this.
SELECT * FROM PersonHierarchy
WHERE PersonHierarchyID.IsDescendantOf(
(SELECT PersonHierarchyID FROM PersonHierarchy WHERE PersonID = 100)
) = 1
UNION ALL
SELECT * FROM PersonHierarchy
WHERE PersonHierarchyID.IsDescendantOf(
(SELECT PersonHierarchyID FROM PersonHierarchy WHERE PersonID = 110)
) = 1
UNION ALL ...
P.S. Also I found such query to select list of ids which might be helpful:
SELECT * FROM (VALUES (100), (110), (120), (130)) AS Parent(ParentID)
To summarize, my goal is to write query which accepts array of parent IDs as a parameter and returns all their descendants in a single set.
You're thinking too hard.
WITH parent AS (
SELECT PersonHierarchyID FROM PersonHierarchy
WHERE PersonID in (<list of parents>)
)
SELECT * FROM PersonHierarchy
WHERE PersonHierarchyID.IsDescendantOf((SELECT * FROM parent)) = 1
I'd write it like this, though:
select child.*
from PersonHierarchy as parent
inner join PersonHierarchy as child
on child.PersonHierarchyID.IsDescendantOf(
parent.PersonHierarchyId
) = 1
where Parent.PersonId in (<list of parents>)
Note: in both cases, this could be slow as it has to evaluate IsDescendantOf for n*m entries (with n being the cardinality of the list of parents and m being the cardinality of the table).
I recently had a similar problem and I solved it by writing a table-valued function that, given a hierarchyId would return all of the parents. Let's look at a solution to your problem using that approach. First, the function:
CREATE FUNCTION [dbo].[GetAllAncestors] (#h HierarchyId, #IncludeSelf bit)
RETURNS TABLE
AS RETURN
WITH cte AS (
SELECT #h AS h, 1 AS IncludeSelf
)
SELECT #h.GetAncestor(n.NumberId) AS Hierarchy
FROM ref.Number AS n
WHERE n.NumberId <= #h.GetLevel()
AND n.NumberId >= 1
UNION ALL
SELECT h
FROM cte
WHERE IncludeSelf = #IncludeSelf
It assumes that you have a Numbers table. They're immensely useful. If you don't have one, look at the accepted answer here. Let's talk about that function for a second. In essence, it says "For the passed in hierarchyId, get the current level. Then get call GetAncestor until you're at the top of the hierarchy.". Note that it optionally returns the passed in hierarchyId. In my case, I wanted to consider a record an ancestor of itself. You may or may not want to.
Moving onto a solution that uses this, we get something like:
select child.*
from PersonHierarchy as child
cross apply [dbo].[GetAllAncestors](child.PersonHierarchyId, 0) as ancestors
inner join PersonHierarchy as parent
on parent.PersonHierarchyId = ancestors.Hierarchy
where parent.PersonId in (<list of parents>)
It may or may not work for you. Try it out and see!
It might be useful for someone. I found way of doing this by self-joining query:
SELECT p2.* FROM PersonHierarchy p1
LEFT JOIN PersonHierarchy p2
ON p2.PersonHierarchyID.IsDescendantOf(p1.PersonHierarchyID) = 1
WHERE
p1.PersonID IN (100, 110, 120, 130)
You can use this query
Select
child.*,
child.[PersonHierarchyID].GetLevel(),
child.[PersonHierarchyID].GetAncestor(1)
From
PersonHierarchy as parents
Inner Join PersonHierarchy as child
On child.[PersonHierarchyID].IsDescendantOf(parents.[PersonHierarchyID] ) = 1
Where
parents.[PersonHierarchyID] = 0x68
Please check, this should work for you. I havent tried to modify your script but just put the query in loop. Hope it helps.
DECLARE #String VARCHAR(MAX) = '100, 110, 120, 130'
DECLARE #SQL VARCHAR(MAX)
SET #String = REPLACE(#String, CHAR(32), '') + ','
WHILE CHARINDEX(',', #String) > 0
BEGIN
DECLARE #ToString INT
DECLARE #StringLength INT
DECLARE #WorkingString VARCHAR(MAX)
DECLARE #WorkingLength INT
SET #ToString = CHARINDEX(',', #String)
SET #StringLength = LEN(#String)
SET #WorkingString = SUBSTRING(#String, 1, #ToString - 1)
SET #String = SUBSTRING(#String, #ToString + 1, #StringLength)
SET #WorkingString = 'SELECT * FROM PersonHierarchy ' + CHAR(13) + CHAR(10)
+ 'WHERE PersonHierarchyID.IsDescendantOf((SELECT PersonHierarchyID FROM PersonHierarchy WHERE PersonID = '
+ #WorkingString + ')) = 1' + CHAR(13) + CHAR(10)
+ CASE WHEN CHARINDEX(',', #String) > 0 THEN 'UNION ALL'+ CHAR(13) + CHAR(10) ELSE '' END
SET #SQL = ISNULL(#SQL,'') + #WorkingString
END
PRINT #SQL
EXEC (#SQL)

How to compare two comma-separated strings, and return TRUE if there is at least 1 match

I have to variables that contain comma-separated strings:
#v1 = 'hello, world, one, two'
#v2 = 'jump, down, yes, one'
I need a function that will return TRUE if there is at least one match. So in the above example, it would return TRUE since the value 'one' is in both strings.
Is this possible in SQL?
Use a split function (many examples here - CLR is going to be your best option in most cases back before SQL Server 2016 - now you should use STRING_SPLIT()).
Once you have a split function, the rest is quite easy. The model would be something like this:
DECLARE #v1 VARCHAR(MAX) = 'hello, world, one, two',
#v2 VARCHAR(MAX) = 'jump, down, yes, one';
SELECT CASE WHEN EXISTS
(
SELECT 1
FROM dbo.Split(#v1) AS a
INNER JOIN dbo.Split(#v2) AS b
ON a.Item = b.Item
)
THEN 1 ELSE 0 END;
You can even reduce this to only call the function once:
SELECT CASE WHEN EXISTS
(
SELECT 1 FROM dbo.Split(#v1)
WHERE ', ' + LTRIM(#v2) + ','
LIKE '%, ' + LTRIM(Item) + ',%'
) THEN 1 ELSE 0 END;
On 2016+:
SELECT CASE WHEN EXISTS
(
SELECT 1 FROM STRING_SPLIT(#v1, ',')
WHERE ', ' + LTRIM(#v2) + ','
LIKE '%, ' + LTRIM([Value]) + ',%'
) THEN 1 ELSE 0 END;
You can use CTEs to split your string into xml nodes, then insert the words into table variables. Joining the table variables will reveal any matches
DECLARE #v1 VARCHAR(200) = 'hello, world, one, two'
DECLARE #v2 VARCHAR(200) = 'jump, down, yes, one'
DECLARE #v1Words TABLE (word VARCHAR(100))
DECLARE #v2Words TABLE (word VARCHAR(100))
;WITH cteSplitV1 AS(
SELECT CAST('<word>' + REPLACE(#v1,', ','</word><word>') + '</word>' AS XML) AS words)
INSERT INTO #v1Words(word)
SELECT word.x.value('.','VARCHAR(100)') AS [word]
FROM cteSplitV1
CROSS APPLY words.nodes('/word') AS word(x)
;WITH cteSplitV2 AS(
SELECT CAST('<word>' + REPLACE(#v2,', ','</word><word>') + '</word>' AS XML) AS words)
INSERT INTO #v2Words(word)
SELECT word.x.value('.','VARCHAR(100)') AS [word]
FROM cteSplitV2
CROSS APPLY words.nodes('/word') AS word(x)
SELECT *
FROM #v1Words v1
JOIN #v2Words v2
ON v1.word = v2.word

SQL Server While Loop Not Returning expected Value

Here the code from my function:
-- Get Country Names
DECLARE getCountriesName CURSOR FOR
SELECT c.Name
FROM VocabCountry c
RIGHT OUTER JOIN ProjectCountryRelations cr
ON cr.CountryId = c.Id
WHERE
c.Id = #project_id;
-- Finally Create list to return
OPEN getCountriesName;
FETCH NEXT FROM getCountriesName INTO #Name;
WHILE ##FETCH_STATUS = 0
BEGIN
SET #listOfItems = #listOfItems + #Name + ', '
FETCH NEXT FROM getCountriesName INTO #Name;
END;
CLOSE getCountriesName;
DEALLOCATE getCountriesName;
I am expecting a list of comma separated values, for example, like so:
Canada, United States of America,
I verified that the SELECT returns the countries expected.
Thanks for any help!
Eric
If you're on SQL Server 2008 or newer, you could easily do this with a single SELECT and some FOR XML PATH magic :-)
SELECT
STUFF(
(SELECT ',' + c.Name
FROM VocabCountry c
RIGHT OUTER JOIN ProjectCountryRelations cr
ON cr.CountryId = c.Id
FOR XML PATH('')
), 1, 1, '')
That should return a comma-separated list of those country names for you. No ugly and awful cursor needed!
Using the FOR XML PATH will concat the values with commas.
Using the STUFF Function will remove the first comma you don't want.
So use this:
SELECT
STUFF(
(SELECT ',' + c.Name
FROM VocabCountry c
RIGHT OUTER JOIN ProjectCountryRelations cr
ON cr.CountryId = c.Id
WHERE cr.Id = #project_id
FOR XML PATH('')
), 1, 1, '');
See this fiddle: http://sqlfiddle.com/#!3/fd648/21

SQL Server 2005 - column names based on variable passed in pivot query

I have following query which takes 2 parameters.
YearNumber
MonthNumber
In my pivot query, I am trying to select columns based on #Year_Rtl variable. I need to select data for the year passed, last year and last last year. Since the data being displayed on UI is table format divided by #Year_Rtl, I decided to write a pivot query for that as below.
In the query, it works fine if I hard code [#Year_Rtl], [#Year_Rtl - 1], [#Year_Rtl - 2] to [2012], [2011], [2010]. But since the year passed can be anything, I want columns to be named dynamically.
DECLARE #Month_Rtl int
DECLARE #Year_Rtl int
SET #Year_Rtl = 2012
SET #Month_Rtl = 1
SELECT
'Data 1', [#Year_Rtl], [#Year_Rtl - 1], [#Year_Rtl - 2]
FROM
(SELECT [Yr_No], Qty
FROM dbo.Table1 t
WHERE (t.Col1 = 10) AND
(t.Col2 = '673') AND
((t.Mth_No = #Month_Rtl AND t.Yr_No = #Year_Rtl) OR
(t.Mth_No = 12 AND t.Yr_No IN (#Year_Rtl - 1, #Year_Rtl - 2)))
) p PIVOT (SUM(Qty)
FOR [Yr_No] IN ([#Year_Rtl], [#Year_Rtl-1], [#Year_Rtl-2])
) AS pvt
Above query throws following errors:
Error converting data type nvarchar to smallint.
The incorrect value "#Year_Rtl" is supplied in the PIVOT operator.
Invalid column name '#Year_Rtl - 1'.
Invalid column name '#Year_Rtl - 2'.
Since you can use dynamic SQL, I'd go with a macro-replacement approach. You're identifying areas of the query that must be dynamically replaced with placeholders (e.g. $$Year_Rtl) and then calculating their replacement values below. I find that it keeps the SQL statement easy to follow.
DECLARE #SQL NVarChar(2000);
SELECT #SQL = N'
SELECT
''Data 1'', [$$Year_Rtl], [$$Year_RtlM1], [$$Year_RtlM2]
FROM
(SELECT [Yr_No], Qty
FROM dbo.Table1 t
WHERE (t.Col1 = 10) AND
(t.Col2 = ''673'') AND
((t.Mth_No = $$Month_Rtl AND t.Yr_No = $$Year_Rtl) OR
(t.Mth_No = 12 AND t.Yr_No IN ($$Year_RtlM1, $$Year_RtlM2)))
) p PIVOT (SUM(Qty)
FOR [Yr_No] IN ([$$Year_Rtl], [$$Year_RtlM1], [$$Year_RtlM2])
) AS pvt';
SELECT #SQL = REPLACE(#SQL, '$$Year_RtlM2', #Year_Rtl - 2);
SELECT #SQL = REPLACE(#SQL, '$$Year_RtlM1', #Year_Rtl - 1);
SELECT #SQL = REPLACE(#SQL, '$$Year_Rtl', #Year_Rtl);
SELECT #SQL = REPLACE(#SQL, '$$Month_Rtl', #Month_Rtl);
PRINT #SQL;
-- Uncomment the next line to allow the built query to execute...
--EXECUTE sp_ExecuteSQL #SQL;
Since consuming code will also have to be flaky under this scheme (e.g. selecting columns based on "position" rather than name) - why not normalize the columns by performing a DATEDIFF(year,Yr_No,#Year_Rtl), and work from there? Those columns will always be 0, -1 and -2...
You need to look into Dynamic SQL Pivoting.
I recommend reading Itzik Ben-Gan's T-SQL Fundamentals where he goes over how to do this.
Alternatively try this article if you don't want to buy the book.
Maybe this will help:
First getting the columns with a tally function like this:
DECLARE #Month_Rtl int,
#Year_Rtl int,
#Year_Rtl_Start INT,
#cols VARCHAR(MAX),
#values VARCHAR(MAX)
SET #Year_Rtl = 2012
SET #Month_Rtl = 1
SET #Year_Rtl_Start=2009
;WITH Years ( n ) AS (
SELECT #Year_Rtl_Start UNION ALL
SELECT 1 + n FROM Years WHERE n < #Year_Rtl )
SELECT
#cols = COALESCE(#cols + ','+QUOTENAME(n),
QUOTENAME(n)),
#values = COALESCE(#values + ','+CAST(n AS VARCHAR(100)),
CAST(n AS VARCHAR(100)))
FROM
Years
ORDER BY n DESC
The variable #cols contains the columns that is in the pivot and the variable #values contains the years for the IN. The #Year_Rtl is the end year and the #Year_Rtl_Start is the start for you range.
Then declaring and executing the dynamic pivot like this:
DECLARE #query NVARCHAR(4000)=
N'SELECT
''Data 1'', '+#cols+'
FROM
(
SELECT
[Yr_No], Qty
FROM
dbo.Table1 t
WHERE
t.Col1 = 10
AND t.Col2 = ''673''
AND
(
(
t.Mth_No = '+CAST(#Month_Rtl AS VARCHAR(10))+'
AND t.Yr_No = '+CAST(#Year_Rtl AS VARCHAR(10))+'
)
OR
(
t.Mth_No = 12
AND t.Yr_No IN ('+#values+'))
)
) p
PIVOT
(
SUM(Qty)
FOR [Yr_No] IN ('+#cols+')
) AS pvt'
EXECUTE(#query)

Resources