Code review #1 – dapper and varchar parameters

This is a first post about great code review feedback, that I either gave or received. It will always consist of 3 parts: context, review feedback and explanation. You can go ahead and read previous ones here: So lets not wait anymore and get to it.

The context

This is a simple ASP.Net application, that is requesting database to get count of elements filtered by one parameter. In this case we need a number of users providing a country code, that is always two character string.

This is how DB schema looks like:

Code in .net app is written with Dapper nuget package, that extends functionalities of IDbCommand and offers entities mapping with considerably good performance. It looks like this:

Looks pretty standard, right? What is wrong here then?

Review feedback

Please convert countryCode parameter to ANSI string in GetCountByCountryCode method, cause if you use it like that, it’s not optimal.


Notice, that CountryCode in database schema is a varchar(2) and this means that it stores two 1-byte characters. On the contrary nvarchar type is 2-byte per character type, that can store multilingual data. When using .net String type we are using unicode strings by default and therefore if we pass countryCode string to SQL it will have to be converted to ANSI string first.

The correct code should look like this:

If we run SQL Server Profiler and check what requests are we doing, this is what we will get:

As you can see first query needs to convert CountryCode parameter from nvarchar(4000) to varchar(2) in order to compare it.

In order to check how that would impact the performance, I created a SQL table with 1000000(one milion) records and compared results.

Before review it took 242 miliseconds and after review it took only 55 miliseconds. So as you see it is more that 4 times performance improvement in this specific case.

  All code posted here you can find on my GitHub:

7 thoughts on “Code review #1 – dapper and varchar parameters

  1. Nice, thanks for sharing that 🙂
    Small code pieces not a lot of people would think to refactor. Especially that if you sum many small things like this one, performance can be significantly harmed.

  2. If my parameter declaration is this

    p.Add(“@SessionToken”, SessionToken, dbType: DbType.String, direction: ParameterDirection.Input);

    would i then change it to the following

    .Add(“@SessionToken”, SessionToken, dbType: DbType.AnsiString, direction: ParameterDirection.Input);

  3. There is an easier solution, and possibly slightly quicker, and no need to specify length as it’s absolute:

    BEFORE: “SELECT count(*) FROM [Users] WHERE CountryCode = @CountryCode”, new blah blah …..
    AFTER: $”SELECT count(*) FROM [Users] WHERE CountryCode = ‘{CountryCode}'”).ConfigureAwait(false)…

    Here we are using the string interpolation operator, and simply adding single quotes around the {CountryCode}. I went from 2000ms queries to 1ms queries with this simple change.

  4. > As you can see first query needs to convert CountryCode parameter from nvarchar(4000) to varchar(2) in order to compare it.

    I think at least in some cases the situation is even worse: Since it is not possible to convert nvarchar to char without possible information loss it is actually the row data that gets converted. After all a single convert from nvarchar to char should not even register as a performance loss. But converting the char(2) to nvarchar for each row to do the comparison is a significant hit, which is probably what you are seeing here.

    Also @Nicholas…
    > Here we are using the string interpolation operator…

    Sure, it’s faster. But this is how SQL injection security nightmares starts. I have seen it first hand. Suppose country code comes from a form post from the web. What happens when someone posts

    “‘; DELETE FROM [Users]; Select ‘PAWND”

    as CountryCode?

Leave a Reply

Your email address will not be published. Required fields are marked *