Skip to content

Custom table-valued functions

In QueryStorm, you can write C# table-valued functions in order to expose external data to SQLite e.g. REST APIs, local files, active directory data, external databases...

You can download the demo workbook from here.

Defining the function

In this example, I'll load data from an exchange rates API and join it with data in an Excel table.

For a start, let's connect via C# and paste in the following function into the editor:

using Newtonsoft.Json.Linq;

///<summary>Gets the exchange rates for the specified currency on the specified date.</summary>
///<param name="date">The date for which to get the currency rates for</param>
public static IEnumerable<Rate> GetExchangeData(DateTime date, string baseCurrency = "EUR", string symbols = null)
    var requestUri = $"{date.ToString("yyyy-MM-dd")}?base={baseCurrency}&symbols={symbols.Replace(" ","")}";
    var response = new WebClient().DownloadString(requestUri);

    string ratesStr = JObject.Parse(response).Property("rates").Value.ToString();

    return JsonConvert
        .DeserializeObject<IDictionary<string, decimal>>(ratesStr)
        .Select(kvp => new Rate() { Currency = kvp.Key, Exchange = kvp.Value });

public class Rate
    public string Currency { get; set; }
    public decimal Exchange { get; set; }


The Newtosoft.Json library is included in QueryStorm, and we can use it right out of the box.

Our GetExchangeData() function takes in the desired date, a base currency, and the desired currencies to return. It hits a web API to load the data and parses the JSON response into a collection of Rate objects.

We can, of course, call this function from C#, but we can also expose it to SQLite. To do so, we need to to do the following:

  1. make sure it successfully compiles (has no compile-time errors)
  2. decorate it with the [SQLFuncTabular] attribute
  3. embed it into the workbook

Once that's done, it will be available to SQLite.

Calling the function

When we connect via SQLite, it will compile the C# code in the workbook and register the function we defined.

Calling C# table-valued function

If we like, we can also use this function in joins. For example, let's assume we have a table with transactions, and each transaction has a Date column and a Target currencies column. The date specifies when the transaction happened, and the target currencies specifies the currencies we need to convert into.


We can join each transaction with the appropriate exchange rates like so:

    transactions tr, 
    GetExchangeData(, tr.baseCurrency, tr.exchangeCurrencies)

Here we're joining each row in the transactions table with the results of calling the GetExchangeData function. We're achieving the join by passing values from transactions rows into the GetExchangeData function. The resulting data looks something like this:


It might seem strange to do a join this way, but it's just syntax sugar. Under the hood, SQLite sees table-valued functions as normal tables and their parameters as hidden columns of those tables. Imagine that GetExchangeData is a normal table that has all exchange rates ever (a better name for it might be simply ExchangeRates, though). With this in mind, we can rewrite the query like so:

    transactions tran
    inner join GetExchangeData exchage on =
        and exchage.baseCurrency = tran.baseCurrency
        and exchage.symbols = tran.exchangeCurrencies


The query works exactly the same as the previous one, only the syntax is different.


Caching is important when table joins are at play. We don't want to repeatedly hit a web service in the background if the function is called multiple times with the same parameters. That's why table-valued function results are cached so that multiple calls with the same arguments only execute the table-valued function once and the result is cached and reused.

The caching scope is per-execution, meaning that if a join results in multiple executions of the function with the same parameters, the function is only executed once, but if you run the query again it will be re-executed.