SQL Select Top with Left Join and Where clause

Question

I have two database tables:

Cities with columns:

Country_Code | City_Code | City_Name  

Countries with columns

Country_Code | Country_Name

Based on a few chars entered by User, it checks the City_Name column to return results to populate a City autocomplete box. The result needs to have the city code, city name, country code, and country name, hence the need for a join.

The query I am using is

 SELECT TOP 10 
     ci.Country_Code, ci.City_Code, ci.City_Name, co.Country_Name  
 FROM 
     Cities ci 
 LEFT OUTER JOIN 
     Countries co ON ci.Country_Code = co.Country_Code  
 WHERE 
     ci.City_Name LIKE '@CityName'  
 ORDER BY 
     ci.City_Name  

The results I get are correct, but the query takes a long time to complete. From what I understand, first, the results contain join of both the tables, then the where clause kicks in to get the specific rows only, which are ordered by City Name and top 10 results returned.

My question is, is there a way to speed up the query. Have the where clause checked, and then only perform the join, better still perform it only on the top 10 results? I tried putting my WHERE clause in the ON clause, but that gave wrong results.

---

EDIT : @CityName contains 2-3 chars entered by the user and then a '%'.

Answer 1

I'd suggest start with adding clustered index on Countries.Country_Code (also making it the primary key of the Countries table if it is not already so). An index would sort the table such that the search speed in join is increased.

Answer 2

This appears to be your query:

SELECT TOP 10 ci.Country_Code, ci.City_Code, ci.City_Name, co.Country_Name  
FROM Cities ci LEFT OUTER JOIN 
     Countries co
     ON ci.Country_Code = co.Country_Code  
WHERE ci.City_Name LIKE @CityName  
ORDER BY ci.City_Name  ;

Quotes should not be needed around @CityName .

I don't understand the LEFT JOIN . It suggests that there are cities without a valid Country_Code -- and that seems unlikely.

Assuming @CityName does not start with a wildcard (as suggested by your question), then this can make use of an index. I would suggest the following indexes:

• cities(city_name, country_code)
• countries(country_code, country_name)

The second is not needed if country_code is a primary key.