K. Scott Allen

I’m looking at 5-7 files of data in CSV format. The columns and format can change every 3 to 6 months. What I wanted was a strongly type wrapper / data transfer object for all the CSV file formats I had to work with, and I thought this would be a good opportunity to try a T4 template.

If you haven’t heard of T4 templates before, then it is a technology that is worth your time to investigate. Hanselman has a pile of links in his post T4 (Text Template Transformation Toolkit) Code Generation – Best Kept Visual Studio Secret.

Each CSV file I’m working with has a header row:

PID,CaseID,BirthDate,DischargeDate,Status …
1001,2001,1/1/1970,6/1/2009,Foo …

… and so on for ~ 100 columns.

I wanted a template that could look at every CSV file it finds in the same directory as itself and create a class with a string property for each column. The template would derive the name of the class from the filename. Here’s a starting point:

<#@ template language="c#v3.5" hostspecific="true" #>
<#@ import namespace="System.IO" #>
<#@ import namespace="System.Collections.Generic" #>

using System;
namespace Acme.Foo.CsvImport {
                
<#
    foreach(var fileName in GetCsvFileNames())
    {
        var fields = GetFields(fileName);
        var className = GenerateClassName(fileName);
#>
    public class <#= className #> : ImportRecord
    { 
        public <#= className #>(string data)
        {
            var fields = data.Split(',');
            <# for(int i = 0; i < fields.Length; i++) { #>
              <#= fields[i] #> = (fields[<#= i #>]);
            <# } #>
        }
    
       <#
       foreach(var field in fields) { #>        
        public string <#= field #> { get; set; } <#
      } #>                        
    }
    
<# } #>
    
} // end namespace

<#+
    private string GenerateClassName(string fileName)
    {
        return Path.GetFileName(fileName).Substring(0,2) 
                          + "_ImportRecord";
    }
    
    private string[] GetFields(string fileName)
    {
        using (var stream = new StreamReader(
                                  File.OpenRead(fileName)))
        {
            return stream.ReadLine().Split(',');
        }
    }
    
    private IEnumerable<string> GetCsvFileNames()
    {
        var path = Host.ResolvePath("");
        return Directory.GetFiles(path, "*.csv");        
    }
#>

The template itself needs some polishing, and the generated code needs some error handling, but this was enough to prove how easy it is to spit out type definitions from a CSV header:

public class AA_ImportRecord
{
    public AA_ImportRecord(string data)
    {
        var fields = data.Split(',');
        ID = fields[0];
        CaseID = fields[1];
        BirthDate = fields[2];           
    }

    public string ID { get; set; }
    public string CaseID { get; set; }
    public string BirthDate { get; set; }

    // ...
}

What I Learned

1. If you want to use the C# 3.5 compiler and LINQ features, make sure to use the correct template and assembly directives. Oleg Sych has the details: Understanding T4: <#@ template #> directive.
2. When the template runs it is an object derived from the TextTransformation class. Oleg’s post that I linked to above also has details on this magic. Use the <#+ #> syntax to add members (a.k.a helper functions) to the class definition for this object.
3. The Host property is of type ITextEngineTemplatingHost. You can use the Host to log errors, find referenced assemblies, and more. In the template I had to use Host.ResolvePath to get the absolute path to the directory where the template file lives. 
4. The biggest issue I had was using <# and #> instead of <% and %>, and <#= #> instead of <%= %>. I created at least 100 errors because my fingers are trained for ASP.NET views instead of text templates.