I'm attempting to do a keyword search which requires all conditions to be met for a result to be shown. I've created a method of making a custom table from a string which stores all of the keywords which are currently required for this search.
I've been able to get it to happen for 'or' using the following
dbo.MultipleTextSearchValuesOR - Is used to make the table of keywords
select Title from vwIncidentSearchView inner join dbo.MultipleTextSearchValuesOR('Testing|Check') on Title Like id
This works great but can't seem to work it out for 'and' (e.g. result must have 'Testing' and 'Check').
Any help would be appreciated
I've figured it out. The way it had to be done was using a cross apply with a function splitting the string into a table.
Then the entries must be grouped by all the fields. Finally they must be checked to see if it meets all the required keywords.
I do realise this is quite a messy method along with it not beingthe most optimized method so if anyone else has any suggestions to improving performance it would be appreciated
Example:
declare #where nvarchar(max)
declare #TitleLikeClause nvarchar(max)
declare #Title nvarchar(max)
set #Title = 'Down&Email'
SELECT
Referenceid,
Title
FROM vwIncidentSearchView
cross apply
(select
Data
from Split(#Title, '&') candidate
where
Title Like candidate.Data) t2o group by Referenceid, Title having count(*) = (select max(ID) from Split(#Title, '&'))
Related
I have a Table which has 3 different Select queries.
e.g.
Staff
------------------------------------------------------------
ID Code Name Phone DOB Email Addr1 Addr2 Addr3
Query1
Select ID, Code, Phone From Staff Where Code = 'ABC'
Query2
Select ID, Code, Phone From Staff Where Name = 'ABCXYZ' And Code = 'B'
Query3
Select ID, Code, Phone From Staff Where Phone= '1234' And Email = 'a#b'
These 3 queries are there in 3 different stored procedure.
I want to find the names of Attributes which i have used in all my Where clauses. But want to Automate this as I have more than 100 tables to look for
Something like this
exec fxGetWhereColList ('Staff');
Result:
Code
Name
Phone
Email
I agree with all posts before, it's a complex case.
I post you a small idea to help you in your work.
If you can retrieve each execution plan, you can parse XML result and get the impacted columns.
For example: On the node <Predicate>
Like Lamak says, doing this right is nearly impossible, at least in SQL.
To do it properly you should use a SQL Parser for the SQL variant you are using. A SQL Parser will identify the objects referenced in the statement, and the elements of the where clauses.
But if you know that the queries are simple and look like the one you are listing, you can use some simple queries to pick apart the statements. I have made an example for a single statement:
declare
#foo nvarchar(max)=N'Select ID, Code, Phone From Staff Where Phone= ''1234'' And Email = ''a#b'' '
, #From nvarchar(max)=' From '
, #Table nvarchar(max)
declare #a int
-- find FROM
set #a= PATINDEX('%'+#From+'%',#foo)
-- Find statement to the right of from
set #foo=ltrim(RIGHT(#foo,len(#foo)-(#a)))
-- Find first space
set #a=CHARINDEX(' ',#foo)
-- find first word, we assume it is the table name
set #Table=ltrim(rtrim(LEFT(#foo,#a)))
-- Find WHERE statement
set #foo=ltrim(rtrim(replace(right(#foo,len(#foo)-#a),'Where','')))
-- Now find matching columns in table, I am using SQL Server so I look up column names in information_schema.columns
select #Table,column_name
from INFORMATION_SCHEMA.COLUMNS
--from (values ('ID'),('Code'),('Name'),('Phone'),('DOB'),('Email'),('Addr1'),('Addr2'),('Addr3'))cols(column_name)
where TABLE_NAME=#table
and #foo like '%'+column_name+'%'
This solution will only work for some simple statements, it assumes a lot of stuff.
For simplicity, assume I have two tables joined by account#. The second table has two columns, id and comment. Each account could have one or more comments and each unique comment has a unique id.
I need to write a t-sql query to generate one row for each account - which I assume means I need to combine as many comments as might exit for each account. This assumes the result set will only show the account# once. Simple?
Sql Server is a RDBMS best tuned for storing data and retrieving data, you can retrieve the desired data with one very simple query but the desired format should be handled with any of the reporting tools available like ssrs or crystal reports
Your query will be a simple inner join something like this
SELECT A.Account , B.Comment
FROM TableA AS A INNER JOIN TableB AS B
ON A.Account = B.Account
Now you can use your reporting tool to Group all the Comments by Account when Displaying data.
I do agree with M. Ali, but if you don't have that option, the following will work.
SELECT [accountID]
, [name]
, (SELECT CAST(Comment + ', ' AS VARCHAR(MAX))
FROM [comments]
WHERE (accountID = accounts.accountID)
FOR XML PATH ('')
) AS Comments
FROM accounts
SQL Fiddle
In my actual project I have this exact situation.
What you need is a solution to aggregate the comments in order to show only one line per account#.
I solve it by creating a function to concatenate the comments, like this:
create function dbo.aggregateComments( #accountId integer, #separator varchar( 5 ) )
as
begin;
declare #comments varchar( max ); set #comments = '';
select #comments = #comments + #separator + YouCommentsTableName.CommentColumn
from dbo.YouCommentsTableNAme
where YouCommentsTableName.AccountId = #accountId;
return #comments;
end;
You can use it on you query this way:
select account#, dbo.aggretateComments( account#, ',' )
from dbo.YourAccountTableName
Creating a function will give you a common place to retrieve your comments. It's a good programming practice.
I have a table which looks something like
Event ID Date Instructor
1 1/1/2000 Person 1
1 1/1/2000 Person 2
Now what I want to do is return this data so that each event is on one row and the Instructors are all in one column split with a <br> tag like 'Person 1 <br> Person 2'
Currently the way I have done this is to use a function
CREATE FUNCTION fnReturnInstructorNamesAsHTML
(
#EventID INT
)
RETURNS VARCHAR(max)
BEGIN
DECLARE #Result VARCHAR(MAX)
SELECT
#result = coalesce(#result + '<br>', '') + inst.InstructorName
FROM
[OpsInstructorEventsView] inst
WHERE
inst.EventID = #EventID
RETURN #result
END
Then my main stored procedure calls it like
SELECT
ev.[BGcolour],
ev.[Event] AS name,
ev.[eventid] AS ID,
ev.[eventstart],
ev.[CourseType],
ev.[Type],
ev.[OtherType],
ev.[OtherTypeDesc],
ev.[eventend],
ev.[CourseNo],
ev.[Confirmed],
ev.[Cancelled],
ev.[DeviceID] AS resource_id,
ev.Crew,
ev.CompanyName ,
ev.Notes,
dbo.fnReturnInstructorNamesAsHTML(ev.EventID) as Names
FROM
[OpsSimEventsView] ev
JOIN
[OpsInstructorEventsView] inst
ON
ev.EventID = inst.EventID
This is very slow, im looking at 4seconds per call to the DB. Is there a way for me to improve the performance of the function? Its a fairly small function so im not sure what I can do here, and I couldnt see a way to work the COALESCE into the SELECT of the main procedure.
Any help would be really appreciated, thanks.
You could try something like this.
SELECT
ev.[BGcolour],
ev.[Event] AS name,
ev.[eventid] AS ID,
ev.[eventstart],
ev.[CourseType],
ev.[Type],
ev.[OtherType],
ev.[OtherTypeDesc],
ev.[eventend],
ev.[CourseNo],
ev.[Confirmed],
ev.[Cancelled],
ev.[DeviceID] AS resource_id,
ev.Crew,
ev.CompanyName ,
ev.Notes,
STUFF((SELECT '<br>'+inst.InstructorName
FROM [OpsInstructorEventsView] inst
WHERE ev.EventID = inst.EventID
FOR XML PATH(''), TYPE).value('.', 'nvarchar(max)'), 1, 4, '') as Names
FROM
[OpsSimEventsView] ev
Not sure why you have joined OpsInstructorEventsView in the main query. I removed it here but if you needed you can just add it again.
A few things to look at:
1) The overhead of functions makes them expensive to call, especially in the select statement of a query that could potentially be returning thousands of rows. It will have to execute that function for every one of them. Consider merging the behavior of the function into your main stored procedure, where the SQL Server can make better use of its optimizer.
2) Since you are joining on event id in both tables, make sure you have an index on those two columns. I would expect that you do, given that those both appear to be primary key columns, but make sure. An index can make a huge difference.
3) Convert your coalesce call into its equivalent case statements to remove the overhead of calling that function.
Yes make it an INLINE Table-Valued SQL function:
CREATE FUNCTION fnReturnInstructorNamesAsHTML
( #EventID INT )
RETURNS Table
As
Return
SELECT InstructorName + '<br>' result
FROM OpsInstructorEventsView
WHERE EventID = #EventID
Go
Then, in your SQL Statement, use it like this
SELECT ]Other stuff],
(Select result from dbo.fnReturnInstructorNamesAsHTML(ev.EventID)) as Names
FROM OpsSimEventsView ev
JOIN OpsInstructorEventsView inst
ON ev.EventID = inst.EventID
I'm not exactly clear how the query you show in your question is concatenating data from multiple rows in one row of the result, but the problem is that ordinary UDFs are compiled on use, on EVERY use, so for each row in your output result the Query processopr has to recompile the UDF again. THis is NOT True for an "inline table valued" UDF, as it's sql is folded into the outer sql before it is passed to the SQL optimizer, (the subsystem that generates the statement cache plan) and so the UDF is only compiled once.
I'm querying on my products table for all products with code between a range of codes, and the result brings a row that should't be there.
This is my SQL query:
select prdcod
from products
where prdcod between 'F-DH1' and 'F-FMS'
order by prdcod
and the results of this query are:
F-DH1
F-DH2
F-DH3
FET-RAZ <-- What is this value doing here!?
F-FMC
F-FML
F-FMS
How can this odd value make it's way into the query results?
PS: I get the same results if I use <= and >= instead of between.
According to OP request promoted next comment to answer:
Seems like your collation excludes '-' sign - this way results make sense, FE is between FD and FM.
:)
between and >= and <= are primarily used for numeric operations (including dates). You're trying to use this for strings, which are difficult at best to determine how those operators will interpret the each string.
Now, while I think I understand your goal here, I'm not entirely sure it's possible using SQL Server queries. This may be some business logic (thanks to the product codes) that needs implemented in code. Something like the Entity Framework or Linq-to-SQL may be better suited to get you the data you're looking for.
How about adding AND LEFT(prdcod, 2) = 'F-'?
Try replacing the "-" with a space so the order is what you would expect:
DECLARE #list table(word varchar(50))
--create list
INSERT INTO #list
SELECT 'F-DH1'
UNION ALL
SELECT 'F-DH2'
UNION ALL
SELECT 'F-DH3'
UNION ALL
SELECT 'FET-RAZ'
UNION ALL
SELECT 'F-FMC'
UNION ALL
SELECT 'F-FML'
UNION ALL
SELECT 'F-FMS'
--original order
SELECT * FROM #list order by word
--show how order changes
SELECT *,replace(word,'-',' ') FROM #list order by replace(word,'-',' ')
--show between condition
SELECT * FROM #list where replace(word,'-',' ') between 'F DH1' and 'F FMS'
In my mssql database I have a table containing articles(id, name, content) a table containing keywords(id, name) and a link table between articles and keywords ArticleKeywords(articleId, keywordID, count). Count is the number of occurrences of that keyword in the article.
How can I write a SP that gets a list of comma separated strings and gives me the articles that have this keywords ordered by the number of occurrences of the keywords in the article?
If an article contains more keywords I want to sum the occurrences of each keyword.
Thanks, Radu
Although it isn't completely clear to me what the source of your comma-separated string is, I think what you want is an SP that takes a string as input and produces the desired result:
CREATE PROC KeywordArticleSearch(#KeywordString NVARCHAR(MAX)) AS BEGIN...
The first step is to verticalize the comma-separated string into a table with the values in rows. This is a problem that has been extensively treated in this question and another question, so just look there and choose one of the options. Whichever way you choose, store the results in a table variable or temp table.
DECLARE #KeywordTable TABLE (Keyword NVARCHAR(128))
-- or alternatively...
CREATE TABLE #KeywordTable (Keyword NVARCHAR(128))
For lookup speed, it is even better to store the KeywordID instead so your query only has to find matching ID's:
DECLARE #KeywordIDTable TABLE (KeywordID INT)
INSERT INTO #KeywordTable
SELECT K.KeywordID FROM SplitFunctionResult S
-- INNER JOIN: keywords that are nonexistent are omitted
INNER JOIN Keywords K ON S.Keyword = K.Keyword
Next, you can go about writing your query. This would be something like:
SELECT articleId, SUM(count)
FROM ArticleKeywords AK
WHERE K.KeywordID IN (SELECT KeywordID FROM #KeywordIDTable)
GROUP BY articleID
Or instead of the WHERE you could use an INNER JOIN. I don't think the query plan would be much different.
For the sake or argument lets say you want to look-up all articles containg the keywords Foo, Bar and Shazam.
ALTER PROCEDURE spArticlesFromKeywordList
#KeyWords varchar(1000) = 'Foo,Bar,Shazam'
AS
SET NOCOUNT ON
DECLARE #KeyWordInClause varchar(1000)
SET #KeyWordInClause = REPLACE (#KeyWords ,',',''',''')
EXEC(
'
SELECT
t1.Name as ArticleName,
t2.Name as KeyWordName,
t3.Count as [COUNT]
FROM ArticleKeywords t3
INNER JOIN Articles t1 on t3.ArticleId = t1.Id
INNER JOIN Keywords t2 on t3.KeywordId = t2.Id
WHERE t2.KeyWord in ( ''' + #KeyWordInClause + ''')
ORDER BY
3 descending, 1
'
)
SET NOCOUNT OFF
I think I understand what you are after so here goes ,(not sure what lang you are using but) in PHP (from your description) I would query ArticleKeywords using a ORDER BY count DESC statement (i.e. the highest comes first) - Obviously you can "select by keywordID or articleid. In very simple terms (cos that's me - simple & there may be much better people than me) you can return the array but create a string from it a bit like this:
$arraytostring .= $row->keywordID.',';
If you left join the tables you could create something like this:
$arraytostring .= $row->keywordID.'-'.$row->name.' '.$row->content.',';
Or you could catch the array as
$array[] = $row->keywordID;
and create your string outside the loop.
Note: you have 2 fields called "name" one in articles and one in keywords it would be easier to rename one of them to avoid any conflicts (that is assuming they are not the same content) i.e. articles name = title and keywords name= keyword