In SQL I (sadly) often have to use \"LIKE
\" conditions due to databases that violate nearly every rule of normalization. I can\'t change that right now. But tha
I know this is very late, but I had a similar situation. I needed a "Like In" operator for a set of stored procedures I have, which accept many parameters and then uses those parameters to aggregate data from multiple RDBMS systems, thus no RDBMS-specific tricks would work, however the stored procedure and any functions will run on MS SQL Server, so we can use T-SQL for the functionality of generating the full SQL statements for each RDBMS, but the output needs to be fairly RDBMS-independent.
This is what I've come up with for the moment to turn a delimited string (such as a parameter coming into a stored procedure) into a block of SQL. I call it "Lichen" for "LIKE IN". Get it?
Lichen.sql
SET ANSI_NULLS ON
GO
SET QUOTED_IDENTIFIER ON
GO
-- =======================================================================
-- Lichen - Scalar Valued Function
-- Returns nvarchar(512) of "LIKE IN" results. See further documentation.
-- CREATOR: Norman David Cooke
-- CREATED: 2020-02-05
-- UPDATED:
-- =======================================================================
CREATE OR ALTER FUNCTION Lichen
(
-- Add the parameters for the function here
@leadingAnd bit = 1,
@delimiter nchar(1) = ';',
@colIdentifier nvarchar(64),
@argString nvarchar(256)
)
RETURNS nvarchar(512)
AS
BEGIN
-- Declare the return variable here
DECLARE @result nvarchar(512)
-- set delimiter to detect (add more here to detect a delimiter if one isn't provided)
DECLARE @delimit nchar(1) = ';'
IF NOT @delimiter = @delimit
SET @delimit = @delimiter
-- check to see if we have any delimiters in the input pattern
IF CHARINDEX(@delimit, @argString) > 1 -- check for the like in delimiter
BEGIN -- begin 'like in' branch having found a delimiter
-- set up a table variable and string_split the provided pattern into it.
DECLARE @lichenTable TABLE ([id] [int] IDENTITY(1,1) NOT NULL, line NVARCHAR(32))
INSERT INTO @lichenTable SELECT * FROM STRING_SPLIT(@argString, ';')
-- setup loop iterators and determine how many rows were inserted into lichen table
DECLARE @loopCount int = 1
DECLARE @lineCount int
SELECT @lineCount = COUNT(*) from @lichenTable
-- select the temp table (to see whats inside for debug)
--select * from @lichenTable
-- BEGIN AND wrapper block for 'LIKE IN' if bit is set
IF @leadingAnd = 1
SET @result = ' AND ('
ELSE
SET @result = ' ('
-- loop through temp table to build multiple "LIKE 'x' OR" blocks inside the outer AND wrapper block
WHILE ((@loopCount IS NOT NULL) AND (@loopCount <= @lineCount))
BEGIN -- begin loop through @lichenTable
IF (@loopcount = 1) -- the first loop does not get the OR in front
SELECT @result = CONCAT(@result, ' ', @colIdentifier, ' LIKE ''', line, '''') FROM @lichenTable WHERE id = @loopCount
ELSE -- but all subsequent loops do
SELECT @result = CONCAT(@result, ' OR ', @colIdentifier, ' LIKE ''', line, '''') FROM @lichenTable WHERE id = @loopCount
SET @loopcount = @loopCount + 1 -- increment loop
END -- end loop through @lichenTable
-- set final parens after lichenTable loop
SET @result = CONCAT(@result, ' )')
END -- end 'like in' branch having found a delimiter
ELSE -- no delimiter was provided
BEGIN -- begin "no delimiter found" branch
IF @leadingAnd = 1
SET @result = CONCAT(' AND ', @colIdentifier, ' LIKE ''' + @argString + '''')
ELSE
SET @result = CONCAT(' ', @colIdentifier, ' LIKE ''' + @argString + '''')
END -- end "no delimiter found" branch
-- Return the result of the function
RETURN @result
END -- end lichen function
GO
The delimiter detection is possibly planned, but for now it defaults to a semicolon so you can just put default
in there. There are probably bugs in this. The @leadingAnd
parameter is just a bit value to determine if you want a leading "AND" put in front of the block so it fits in nicely with other WHERE clause additions.
Example Usage (with delimiter in argString)
SELECT [dbo].[Lichen] (
default -- @leadingAND, bit, default: 1
,default -- @delimiter, nchar(1), default: ';'
,'foo.bar' -- @colIdentifier, nvarchar(64), this is the column identifier
,'01%;02%;%03%' -- @argString, nvarchar(256), this is the input string to parse "LIKE IN" from
)
GO
Will return a nvarchar(512) containing:
AND ( foo.bar LIKE '01%' OR foo.bar LIKE '02%' OR foo.bar LIKE '%03%' )
It will also skip the block if the input does not contain a delimiter:
Example Usage (without delimiter in argString)
SELECT [dbo].[Lichen] (
default -- @leadingAND, bit, default: 1
,default -- @delimiter, nchar(1), default: ';'
,'foo.bar' -- @colIdentifier, nvarchar(64), this is the column identifier
,'01%' -- @argString, nvarchar(256), this is the input string to parse "LIKE IN" from
)
GO
Will return a nvarchar(512) containing:
AND foo.bar LIKE '01%'
I'm going to continue work on this, so if I've overlooked something (glaringly obvious or otherwise) please feel free to comment or reach out.
I may have a solution for this, although it will only work in SQL Server 2008 as far as I know. I discovered that you can use the row-constructor described in https://stackoverflow.com/a/7285095/894974 to join a 'fictional' table using a like clause. It sounds more complex then it is, look:
SELECT [name]
,[userID]
,[name]
,[town]
,[email]
FROM usr
join (values ('hotmail'),('gmail'),('live')) as myTable(myColumn) on email like '%'+myTable.myColumn+'%'
This will result in all users with an e-mail adres like the ones provided in the list. Hope it's of use to anyone. The problem had been bothering me a while.
One approach would be to store the conditions in a temp table (or table variable in SQL Server) and join to that like this:
SELECT t.SomeField
FROM YourTable t
JOIN #TempTableWithConditions c ON t.something LIKE c.ConditionValue
In Oracle you can use a collection in the following way:
WHERE EXISTS (SELECT 1
FROM TABLE(ku$_vcnt('bla%', '%foo%', 'batz%'))
WHERE something LIKE column_value)
Here I have used a predefined collection type ku$_vcnt
, but you can declare your own one like this:
CREATE TYPE my_collection AS TABLE OF VARCHAR2(4000);
With PostgreSQL there is the ANY or ALL form:
WHERE col LIKE ANY( subselect )
or
WHERE col LIKE ALL( subselect )
where the subselect returns exactly one column of data.
If you are using MySQL the closest you can get is full-text search:
Full-Text Search, MySQL Documentation