Is this a correct way to allow wildcard search for a user?

Question

Given a textbox name for example, the user requirement wants to be able to do a wildcard search (such as contains, starts with, ends with).

Is it ok to accept the sql wildcard characters ('%' and '_') as input as long as I am still using parameterized query in the backend (Java)? Effectively, allowing the user to build his own regular expression which is what the user's requirement is all about.

Example:

1. User types in the

   textbox = '%are%'
   

2. This parameter is feed to the backend as such:

   public class PersonDaoImpl {
   
           public List<Person> search(String name){//name gets the value from textbox w/ sql wildcards
           Query q = mgr.createNativeQuery('select * from Person where name like :name'); //default to always use like since expecting searchkey with sql wildcards    
           q.setParameter('name', name);//gives the input from the screen
           return q.getResultList();
           } 
   }  
   

3. The result set would include people with names 'Waren', 'Jared', 'Clare', 'Blare' as expected since user provided a regular expression.

With the SQL Parameterize Query, I can ensure that I won't be allowing SQL Injection. This implements the user requirement for wildcard search, but perhaps does it violate anything that I may have missed?

UPDATES: Just found out that Google allows wildcard too, from their help page.

Answer 1

Well, it violates the fact that the user needs to know (or be told) how to construct SQL "LIKE" syntax, but that's all. You could end up with a slow query this way, in that it won't usually be able to use an index, but I wouldn't be concerned in terms of security or correctness.

Answer 2

It's "safe", but probably not a good idea, for two reasons:

1. It's probably not the best ui design to require your users to know sql syntax for this.
2. It's horrible for performance: these queries often can't use your indexes, so they are slow to execute. And they require a lot of cpu time to compare all that text, so they add a lot of load (disproportionate to the already high execution time) to your server. You want a solution that relies on a full-text index instead.

Answer 3

I am curious, how does the name parameter end up getting set in the request? What platform is this? (OP missed setParameter earlier)

As you noted the user need to know the wild-card syntax i.e. the use of %, _, etc. A more popular approach is to just get the string from the username, along with an option for 'exact match'/'starts-with'/'anywhere-in-name'. If you go that route you will also be able to execute a more efficient query in the first two cases.

EDIT:

If the customer insists on contains query then I think your current approach of requiring the end-user to input a pattern better then converting the input string to pattern by putting % around it.

This is because the users will still have the option of not adding (or selectively adding) the % to the search string, resulting in faster query execution. For example:

• If the user enter search string Don the query is select ... from ... where name like 'Don'. The RDBMS will most likely use the index on name.

• If the user enter search string Don% the query is select ... from ... where name like 'Don%'. The RDBMS will still quite likely use the index on name.

• If the user enter search string %Don or %Don% then the index cannot be used.