Faithlife.Data provides helpers for querying ADO.NET-compatible databases.


Quick Start

The Faithlife.Data class library provides an enhanced API for interacting with ADO.NET-compatible databases. It is similar to Dapper and other “micro” ORMs for .NET.

To use this library, add a NuGet package reference to your project and call DbConnector.Create() to create a DbConnector with a valid IDbConnection from your favorite ADO.NET database provider, e.g. Microsoft.Data.SqlClient for SQL Server or MySqlConnector for MySQL.

Here’s a simple code sample that opens an in-memory SQLite database, creates a table, inserts a few rows within a transaction, and runs a couple of queries. There’s no risk of SQL injection attacks with the interpolated strings, which use formatted SQL, documented below. Try it!

// create connection; open automatically and just in time
await using var connector = DbConnector.Create(
  new SqliteConnection("Data Source=:memory:"),
  new DbConnectorSettings { AutoOpen = true, LazyOpen = true });

// create widgets table
await connector.Command(@"
    create table widgets (
    id integer primary key autoincrement,
    name text not null,
    height real not null)")

// insert widgets in a transaction
var widgets = new[]
  new Widget("First", 6.875),
  new Widget("Second", 1.414),
  new Widget("Third", 3.1415),
await using (await connector.BeginTransactionAsync())
  foreach (var widget in widgets)
    await connector.Command(Sql.Format($@"
        insert into widgets (name, height)
        values ({widget.Name}, {widget.Height})"))
  await connector.CommitTransactionAsync();

// get short widgets
var maxHeight = 5.0;
foreach (var widget in await connector
    $"select name, height from widgets where height <= {maxHeight}"))

// get minimum and maximum heights
var (min, max) = await connector
  .Command("select min(height), max(height) from widgets")
  .QuerySingleAsync<(double, double)>();
Console.WriteLine($"minimum height {min}, maximum height {max}");


With a DbConnector, you can:

Consult the reference documentation for additional details.

What about Dapper?

If you are familiar with Dapper, you will note many similarities between it and this library. So why use Faithlife.Data? Here are a few key differences:

  • DbConnector wraps the connection, whereas Dapper primarly provides extension methods on IDbConnection.
  • With Dapper, you must remember to set the transaction parameter when there is an active transaction. Since Faithlife.Data tracks the current transaction, it attaches it to database commands automatically.
  • Faithlife.Data has direct support for modern C# and .NET, including tuples, IAsyncEnumerable, and the new index/range syntax.
  • The multi-mapping support of Faithlife.Data is simpler and more flexible than the map and splitOn parameters of Dapper.
  • Faithlife.Data avoids type conversion, requiring that the requested type exactly match the provided type, whereas Dapper will try to convert the value with Convert.ChangeType(). This is sometimes aggravating, but we feel it is better to know what the database is returning and avoid the surprises that type conversion can bring.
  • The async methods of Faithlife.Data call the async methods of the database provider more consistently than Dapper.
  • Faithlife.Data makes the choice between buffered and unbuffered queries more explicit by providing separate methods. This makes it more likely that clients will keep the difference in mind, and allows Query() to return an IReadOnlyList<T> instead of an IEnumerable<T>.
  • Faithlife.Data has an easy alternative to using anonymous objects for specifying parameters, which may have better performance for some clients and uses stronger types than Dapper’s param parameter of type object.
  • Faithlife.Data does less caching than Dapper. This may or may not be an advantage, depending on usage.
  • Both Faithlife.Data and Dapper will edit the SQL when substituting a collection parameter for a list of dynamically named parameters. The syntax used by Faithlife.Data is more explicit, so scenarios where the SQL is edited are more predictable.

Creating a connector

Like IDbConnection, DbConnector is not thread-safe, so you will need one instance per connection. Consider defining a method to easily create a connection to your database. This example uses Microsoft.Data.Sqlite.

DbConnector CreateConnector() =>
    DbConnector.Create(new SqliteConnection("Data Source=:memory:"));

If you use formatted SQL, some features require you to specify your SQL syntax with the SqlSyntax setting. (This is not required for simple parameter injection.)

DbConnector CreateConnector() =>
    DbConnector.Create(new SqliteConnection("Data Source=:memory:"),
    new DbConnectorSettings { SqlSyntax = SqlSyntax.Sqlite });

Executing a command

Once you have a connector, you can open the connection with OpenConnectionAsync(), create a command with Command(), and execute the command with ExecuteAsync().

await using (var connector = CreateConnector())
await using (await connector.OpenConnectionAsync())
    await connector.Command(@"
create table widgets (
  id integer primary key autoincrement,
  name text not null);").ExecuteAsync();

Note that asynchronous methods in this library return ValueTask, not Task, so be sure to follow the relevant guidelines, e.g. don’t await a ValueTask more than once.

Automatically opening connections

Calling OpenConnectionAsync every time you use a connector can get tiresome. You could move the OpenConnectionAsync call to your connector creation method, or you can use the AutoOpen setting.

Consider using the LazyOpen setting as well, which waits to actually open the database connection until just before it is first used. This ensures that the connection is opened asynchronously (as long as you use asynchronous methods to execute commands). Also, if you have a code path that doesn’t actually execute any commands, the connection will never be opened in that scenario.

DbConnector OpenConnector() => DbConnector.Create(
    new SqliteConnection("Data Source=:memory:"),
    new DbConnectorSettings
      AutoOpen = true,
      LazyOpen = true,
      SqlSyntax = SqlSyntax.Sqlite,

Accessing the database synchronously

Every method that has communicates with the database has an synchronous equivalent without the Async suffix, e.g. Execute(). Consider using the synchronous methods if your ADO.NET provider doesn’t actually support asynchronous I/O.

using (var connector = CreateConnector())
using (connector.OpenConnection())
create table widgets (
  id integer primary key autoincrement,
  name text not null);").Execute();

Using transactions

To leverage a database transaction, call BeginTransactionAsync() before executing any commands, and then call CommitTransactionAsync() before disposing the return value of BeginTransactionAsync().

await using (var connector = OpenConnector())
await using (await connector.BeginTransactionAsync())
    await connector.Command(@"
create table widgets (
  id integer primary key autoincrement,
  name text not null,
  height real not null);").ExecuteAsync();

    await connector.Command(@"
insert into widgets (name, height)
  values ('First', 6.875);
insert into widgets (name, height)
  values ('Second', 3.1415);").ExecuteAsync();

    await connector.CommitTransactionAsync();

If CommitTransactionAsync() is not called, the transaction will be rolled back when the return value of BeginTransactionAsync() is disposed.

ADO.NET requres that the Transaction property of IDbCommand be set to the current transaction; DbConnector takes care of that automatically when executing commands.

Mapping database records

The querying methods of this class library (e.g. QueryAsync()) are generic, allowing the caller to specify the type to which each database record should be mapped.

Simple types

When selecting a single column, use a simple type as the generic parameter.

async Task<IReadOnlyList<string>> GetWidgetNamesAsync(
    DbConnector connector,
    CancellationToken cancellationToken = default)
    return await connector.Command("select name from widgets;")

For compact documentation, many examples will use the synchronous API.

IReadOnlyList<string> GetWidgetNames(DbConnector connector) =>
    connector.Command("select name from widgets;").Query<string>();

Simple types must match the type returned from IDataRecord.GetValue() exactly. No conversions are performed, except between nullable and non-nullable types. Be sure to use a nullable type if the value could be null, e.g. int?; an exception will be thrown if the field is null but the type is not nullable.

The supported simple types are string, long, int, short, byte, ulong, uint, ushort, sbyte, double, float, decimal, bool, Guid, DateTime, DateTimeOffset, and TimeSpan.

Use byte[] or Stream to return the bytes from a “blob” column. If a non-null Stream is returned, be sure to dispose it.

Enumerated types (defined with enum in C#) are supported like simple types, except that conversions are performed if needed. Specifically, strings are parsed to the enumerated type (ignoring case) and integral types are cast to the enumerated type.


Use tuples to map multiple record fields at once. Each tuple item is read from the record in order. The record field names are ignored, as are the tuple item names, if any.

IReadOnlyList<(string Name, double Height)> GetWidgetInfo(DbConnector connector) =>
    connector.Command("select name, height from widgets;").Query<(string, double)>();


If the type isn’t a simple type or a tuple, it is assumed to be a DTO (data transfer object) type, i.e. a type with properties that correspond to record fields. Both read/write and read-only properties are supported; DtoInfo<T>.CreateNew from Faithlife.Reflection is used to create the instance.

class WidgetDto
    public long Id { get; set; }
    public string? Name { get; set; }
    public double Height { get; set; }

When a DTO type is used, a new instance of the DTO is created, and each record field is mapped to a DTO property whose name matches the field name, ignoring case and any underscores (so full_name would map successfully to FullName, for example). If the property has a Column attribute with a non-null Name property (e.g. from System.ComponentModel.DataAnnotations), that name is used instead of the field name. Not every property of the DTO must be used, but every mapped field must have a corresponding property.

IReadOnlyList<WidgetDto> GetWidgets(DbConnector connector) =>
    connector.Command("select id, name, height from widgets;").Query<WidgetDto>();

For more ways to map records, see advanced record mapping below.

Parameterized queries

When executing parameterized queries, the parameter values are specified with the DbConnector.Command() method. One way to specify command parameters is via one or more string/object tuples after the command SQL.

void InsertWidget(DbConnector connector, string name, double height) =>
        "insert into widgets (name, height) values (@name, @height);",
        ("name", name), ("height", height)).Execute();

Formatted SQL

Typing parameter names in the SQL command text and parameters objects seems redundant. String interpolation can be used to put the parameters in the SQL safely by using Sql.Format() to format an interpolated string into command text and parameters.

void InsertWidget(DbConnector connector, string name, double height) =>
        $"insert into widgets (name, height) values ({name}, {height});"

This is equivalent to the following:

void InsertWidget(DbConnector connector, string name, double height) =>
        "insert into widgets (name, height) values (@fdp0, @fdp1);",
        ("fdp0", name), ("fdp1", height)).Execute();

(fdp is just an arbitrary prefix for the automatically named parameters; it stands for “Faithlife.Data parameter”.)

Note that using an interpolated string without Sql.Format() is still a SQL injection vulnerability:

// Don't do this!
    $"insert into widgets (name, height) values ({name}, {height});"

SQL text and parameters can be composed using instances of the Sql class with Sql.Format. Sql instances in interpolated strings are used to build the SQL statement. (They are not converted into parameters like any other value would be.) Sql.Format returns a Sql instance, so Sql.Format can be composed with Sql.Format as needed.

IReadOnlyList<WidgetDto> GetWidgets(DbConnector connector,
    double? minHeight = null, string[]? fields = null)
    var fieldsSql = fields is null ? Sql.Raw("*") : Sql.Join(", ", fields.Select(Sql.Name));
    var whereSql = minHeight is null ? Sql.Empty : Sql.Format($"where height >= {minHeight}");
    return connector.Command(Sql.Format($"select {fieldsSql} from widgets {whereSql};"))

To create Sql instances, use static members on the Sql class:

  • Sql.ColumnNames generates a list of column names from a DTO for SELECT and INSERT statements.
  • Sql.ColumnParams generates a list of parameters from a DTO for an INSERT statement.
  • Sql.Concat (or operator +) concatenates SQL fragments.
  • Sql.Empty is an empty SQL fragment.
  • Sql.Join joins SQL fragments with a separator.
  • Sql.LikePrefixParam generates a parameter with a LIKE pattern for prefix matching.
  • Sql.Name creates SQL that quotes the specified identifier.
  • Sql.Param generates a parameter for the specified value.
  • Sql.Raw creates raw SQL from the specified string.

Since commands are commonly created with a single call to Sql.Format, the CommandFormat method can be used as shorthand.

void InsertWidget(DbConnector connector, string name, double height) =>
        $"insert into widgets (name, height) values ({name}, {height});"

Collection parameters

Database providers do not typically support collections as parameter values, which makes it difficult to run queries that use the IN operator. To expand a collection into a set of numbered parameters, use ... after the parameter name in the SQL and Faithlife.Data will make the necessary substitutions.

    "select id from widgets where name in (@names...);",
    ("names", new[] { "one", "two", "three" })).Execute();

This is equivalent to:

    "select id from widgets where name in (@names_0, @names_1, @names_2);",
    ("names_0", "one"), ("names_1", "two"), ("names_2", "three")).Execute();

This works with formatted SQL as well.

var names = new[] { "one", "two", "three" };
    $"select id from widgets where name in ({names}...);"

Important note: If the collection is empty, an InvalidOperationException will be thrown, since omitting the parameter entirely may not be valid (or intended) SQL.

For more ways to specify query parameters, see advanced parameters below.

Single-record queries

If your query is for a single record, call QuerySingleAsync(), which throws an exception if the query returns multiple records, or QueryFirstAsync(), which does not check for additional records and therefore may be more efficient.

double height = await connector.Command(
    "select height from widgets where name = @name;",
    ("name", "First")).QuerySingleAsync<double>();

If your single-record query might also return no records, call QuerySingleOrDefaultAsync() or QueryFirstOrDefaultAsync(), which return default(T) if no records were found.

double? height = await connector.Command(
    "select height from widgets where name = @name;",
    ("name", "First")).QueryFirstOrDefaultAsync<double?>();

As always, drop the Async suffix for the synchronous API.

Multiple result sets

If your query has multiple result sets, all of the records from all of the result sets will be read and mapped to the same type.

If you want to map each result set to its own type, call QueryMultipleAsync() and then call ReadAsync() for each result set.

await using (var sets = await connector.Command(
    "select name from widgets; select height from widgets;").QueryMultipleAsync())
    IReadOnlyList<string> names = await sets.ReadAsync<string>();
    IReadOnlyList<double> heights = await sets.ReadAsync<double>();
    // ...

Unbuffered queries

The Query and Read methods read all of the records into an IReadOnlyList<T>. Reading all of the data as quickly as possible is often best for performance, but you can read the records one at a time by calling Enumerate() or EnumerateAsync() instead.

var averageHeight = connector.Command("select height from widgets;")

Note that EnumerableAsync returns an IAsyncEnumerable<T>. Use await foreach and/or System.Linq.Async to enumerate the values.

double? minHeight = null;
double? maxHeight = null;
await foreach (var height in connector.Command("select height from widgets;")
    minHeight = minHeight is null ? height : Math.Min(minHeight.Value, height4);
    maxHeight = maxHeight is null ? height : Math.Max(maxHeight.Value, height4);

Cached commands

Use Cache() to potentially improve performance when executing the same query with different parameter values. Cached commands can particularly help with SQLite, which typically doesn’t have as much I/O overhead as other databases. The IDbCommand object is cached indefinitely with the DbConnector object, so avoid caching commands that will only be executed once.

foreach (var (name, size) in widgets)
        $"insert into widgets (name, size) values ({name}, {size});"

Prepared commands

Use Prepare() to automatically call Prepare on the IDbCommand before it is executed. This can be particularly beneficial with the Npgsql provider for PostgreSQL. Consider using Cache() as well, which caches the prepared command and then reuses it without preparing it again. Be sure to measure any performance advantage of caching and/or preparing commands; in particular, preparing the command may hurt performance in some scenarios.

foreach (var (name, size) in widgets)
        $"insert into widgets (name, size) values ({name}, {size});"

Command timeout

To override the default timeout for a particular command, use WithTimeout().

    $"insert into widgets (name, size) values ({name}, {size});"

Stored procedures

DbConnector also works with stored procedures. Simply call StoredProcedure() instead of Command() and pass the name of the stored procedure instead of a SQL query.

connector.StoredProcedure("CreateWidget", ("name", name), ("size", size)).Execute();

Bulk insert

The BulkInsert() and BulkInsertAsync() extension methods allow simple and efficient insertion of many rows into a database table.

The simplest way to insert many rows into a database table is to execute INSERT commands in a loop. Unfortunately, this can be extremely slow, even when the commands are all executed in a single transaction. (SQLite is a notable exception here; inserting one row at a time in a loop can be considerably faster than doing a bulk insert.)

Each DBMS has its own preferred approaches for efficiently inserting many rows into a database, but the most portable way is to execute an INSERT command with multiple rows in the VALUES clause, like so:

insert into widgets (name, size) values ('foo', 22), ('bar', 14), ('baz', 42)

Building a SQL statement for a large number of rows is straightforward, but runs the risk of SQL injection problems if the SQL isn’t escaped propertly.

Using command parameters is safer, but building and executing the SQL is more complex. Furthermore, databases often have a limit on the maximum number of command parameters that can be used, so it can be necessary to execute multiple SQL statements, one for each batch of rows to insert.

BulkInsert() builds the SQL commands for each batch and injects the command parameters as needed.

var widgets = new[]
    new { name = "foo", size = 22 },
    new { name = "bar", size = 14 },
    new { name = "baz", size = 42 },
connector.Command("insert into widgets (name, size) values (@name, @size)...")

The ... after the VALUES clause must be included. It is used by BulkInsert to find the end of the VALUES clause that will be transformed. The call above will build a SQL statement like so:

insert into widgets (name, size) values (@name_0, @size_0), (@name_1, @size_1), (@name_2, @size_2)

The actual SQL statement will have as many parameters as needed to insert all of the specified rows. If the total number of command parameters would exceed 999 (a reasonable number for many databases), it will execute multiple SQL commands until all of the rows are inserted.

All of the transformed SQL will be executed for each batch, so including additional statements before or after the INSERT statement is not recommended.

Execute the method within a transaction if it is important to avoid inserting only some of the rows if there is an error.

The BulkInsert() and BulkInsertAsync() methods of the BulkInsertUtility static class are extension methods on DbConnectorCommand. They support an optional BulkInsertSettings parameter that allows you to change the maximum number of command parameters and/or the maximum number of rows per batch.

The method returns the total number of rows affected (or, more specifically, the sum of the row counts returned when executing the SQL commands for each batch).

Advanced record mapping

This section documents additional scenarios for mapping records to values.


Record fields can be mapped to object or dynamic. If a single field is mapped to object or dynamic, the object from IDataRecord.GetValue() is returned directly.

var heights = connector.Command("select height from widgets;")
    .Query<object>(); // returns boxed doubles

If multiple fields are mapped to object or dynamic, an ExpandoObject is returned where each property corresponds to the name and value of a mapped field.

dynamic widget = connector.Command("select name, height from widgets;")
string name =;

Unfortunately, object and dynamic cannot have different algorithms. The implementation cannot distinguish between them, because typeof(T) == typeof(object) when T is dynamic. To avoid confusion, use object when mapping a single field and dynamic when mapping multiple fields.


Record fields can also be mapped to a dictionary of strings to objects, in which case each field gets a key/value pair in the dictionary. The supported dictionary types are Dictionary<string, object>, IDictionary<string, object>, IReadOnlyDictionary<string, object>, and IDictionary.

var dictionary = connector.Command("select name, height from widgets;")
    .Query<Dictionary<string, object>>()[0];
double height = (double) dictionary["height"];

Advanced tuples

Tuples can include multi-field types like DTOs and dynamic.

IReadOnlyList<(WidgetDto Widget, long NameLength)> GetWidgetAndNumber(DbConnector connector) =>
    connector.Command("select id, height, length(name) from widgets;")
        .Query<(WidgetDto, long)>();

If the tuple has two or more multi-field types, all but the last must be terminated by a null record value whose name is null.

IReadOnlyList<(WidgetDto Widget, dynamic Etc)> GetWidgetAndDynamic(DbConnector connector) =>
    connector.Command("select id, height, null, 1 as more, 2 as data from widgets;")
        .Query<(WidgetDto, dynamic)>();

Mapping delegate

For full control over the mapping, the client can specify the map parameter, which is of type Func<IDataRecord, T>. That delegate will be called for each IDataRecord instance returned by the query.

IReadOnlyList<string> GetWidgetNames(DbConnector connector) =>
    connector.Command("select name from widgets;")
        .Query(x => x.GetString(0));

The DataRecordExtensions static class provides Get<T>() extension methods on IDataRecord for mapping all or part of a record into the specified type.

IReadOnlyList<double> GetWidgetHeights(DbConnector connector) =>
    connector.Command("select name, height from widgets;")
        .Query(x => x.Get<double>(1));

Fields can also be accessed by name, though that uses IDataRecord.GetOrdinal() and is thus slightly less efficient.

IReadOnlyList<double> GetWidgetHeights(DbConnector connector) =>
    connector.Command("select name, height from widgets;")
        .Query(x => x.Get<double>("height"));

You can also read multiple fields.

IReadOnlyList<(string Name, double Height)> GetWidgetInfo(DbConnector connector) =>
    connector.Command("select id, name, height from widgets;")
        .Query(x => x.Get<(string, double)>(index: 1, count: 2));

C# 8 range syntax can be used:

IReadOnlyList<(string Name, double Height)> GetWidgetInfo(DbConnector connector) =>
    connector.Command("select id, name, height from widgets;")
        .Query(x => x.Get<(string, double)>(1..3));

You can also use the delegate to avoid the null terminator when reading two or more multi-field types. To avoid having to count fields, we can use a Get<T>() overload that takes a start name and an end name to specify the range.

IReadOnlyList<(WidgetDto Widget, dynamic Etc)> GetWidgetAndDynamic2(DbConnector connector) =>
    connector.Command("select id, height, 1 as more, 2 as data from widgets;")
        .Query(x => (x.Get<WidgetDto>("id", "height"), x.Get<dynamic>("more", "data")));

Advanced parameters

The DbParameters structure can be used to build a list of parameters by calling one of the Create() methods.

You can add additional parameters by calling the Add() methods, but note that DbParameters is an immutable collection, so you will need to use the return value of the Add method.

var tallWidgets = connector.Command(
    "select id from widgets where height >= @minHeight and height <= @maxHeight;",
    DbParameters.Create("minHeight", minHeight).Add("maxHeight", maxHeight)).Query<long>();

Parameters from DTOs/collections

Use DbParameters.FromDto() or DbParameters.AddDto() to create parameters from the names and values of public properties and fields, e.g. of anonymous types.

var newWidget = new WidgetDto { Name = "Third", Height = 1.414 };
    "insert into widgets (name, height) values (@Name, @Height);",

var tallWidgets = connector.Command(
    "select id from widgets where height >= @minHeight and height <= @maxHeight;",
    DbParameters.FromDto(new { minHeight = 1.0, maxHeight = 100.0 })).Query<long>();

Use DbParameters.FromDtos() or DbParameters.AddDtos() to create parameters for many DTOs at once.

Use DbParameters.FromMany() or DbParameters.AddMany() to explicitly create parameters from a collection.

Provider-specific parameters

If the parameter value implements IDbDataParameter, that object is used as the parameter after setting its ParameterName property.

double height = await connector.Command(
    "select height from widgets where name = @name;",
    ("name", new SqliteParameter { Value = "Bob", SqliteType = SqliteType.Text }))

Enhancing the API

To avoid defining too many method overloads, starting a command and executing a query use chained methods. Feel free to reduce typing by creating your own extension methods that match your usage of the library.

For example, this extension method can be used to execute a query with parameters from a DTO in one method call:

public static IReadOnlyList<T> Query<T>(this DbConnector connector,
    string sql, object param) =>
        connector.Command(sql, DbParameters.FromDto(param))

Use with Dapper

If you like the Dapper query API but want to use DbConnector to track the current transaction, use extension methods to call Dapper, accessing the Connection and Transaction properties as needed.

public static IEnumerable<T> Query<T>(this DbConnector connector,
    string sql, object param = null, bool buffered = true) =>
            .Query<T>(sql, param, connector.Transaction, buffered: buffered);