在ASP.NET MVc Web应用程序中如何使用.contains进行搜索和自动完成搜索

Question

i have found a lot of tutorials and books that implements the auto-complete search in MVC web applications as in :-

public ActionResult ArtistSearch(string q)
{
var artists = GetArtists(q);
return PartialView(artists);
}
private List<Artist> GetArtists(string searchString)
{
return storeDB.Artists
.Where(a => a.Name.Contains(searchString))
.ToList();
}

but this raised a question of how much is this approach scalable in real applications that might have thousands of records ???,, so will using Contains() scale well Or there is a much better approach?? BR

Answer 1

If I remember correctly string.Contains() is translated into a LIKE Query with a wildcard on each side of the query string. This makes it very difficult/impossible to use an index, so you can expect your performance to be O(n) on your dataset since SQL Server does a full table scan (see Does SQL Server optimize LIKE ('%%') query? ).

To optmize your query, you might want to take a look at the Full-Text Indexing Capabilities, more info here: SQL Server: how to optimize "like" queries? ).

If you could use .StartsWith instead of .Contains you will have a LIKE query with a wildcard at the end, and you can use an Index on the queried column to get fast lookups (be sure to check the query execution plan!).

I guess you will have much better perceived performance if you try to focus on the UX of your auto-complete feature: Start auto-complete search after a short lock period (when the user stops typing), and make sure it doesn't block (happens in the background).

Answer 2

It depends on the data your serving. In a "real world" application, you need to think about noise/stop words ("and", "the"), abbreviations ("st" for "street"), etc and difficult languages.

In these scenarios, .Contains won't fit the bill, and you'll need to employ a full-text search indexing engine, such as Lucene.NET or SQL Server Full Text Search.

Answer 3

Substring queries cannot use index seeks. But they still can use indexes. Scanning a few thousand records is really nothing if you don't do it too often.

So I recommend that you create an index on Name and you'll be fine for an awful lot of data.

Answer 4

Looks like you are using LINQ with the Entity Framework. LINQ gets converted into SQL and the call to contains gets converted into a LIKE WHERE clause, so you can simply run SELECT * FROM Artists WHERE Name LIKE '%whatever%' to get an idea about performance.

Note there are a couple of things you can do to reduce the impact. One you can limit the number of results .Take(20) . Also you can wait until the user takes at least a couple of characters before triggering auto complete. Finally you can 'throttle' the call to auto complete, so that you don't call auto complete every time they type a character, instead of wait until they go say a half a second without typing an additional character.