Dapper.FSharp

Lightweight F# extension for StackOverflow Dapper with support for MSSQL, MySQL, and PostgreSQL

Features

• No auto-attribute-based-only-author-maybe-knows-magic behavior
• Support for F# records / anonymous records
• Support for F# options
• LINQ Query Provider
• Support for SQL Server 2012 (11.x) and later / Azure SQL Database, MySQL 8.0, PostgreSQL 12.0
• Support for SELECT (including JOINs), INSERT, UPDATE (full / partial), DELETE
• Support for OUTPUT clause (MSSQL only)
• Easy usage thanks to F# computation expressions
• Keeps things simple

Installation

If you want to install this package manually, use usual NuGet package command

Install-Package Dapper.FSharp

or using Paket

paket add Dapper.FSharp

What's new in v3?

3rd version is rather an evolution, than revolution. The major concern was to keep the library maintainable with low possible effort while providing the best type safety. Hence version 3 contains two major (breaking) changes:

• Old API has been dropped in favor of LINQ (watch out for namespace change - Dapper.FSharp.LinqBuilders are no more)
• join has been renamed to innerJoin
• Minimal supported version is NET 5.0

If you still need/want to use v2.0, follow the Version 2 docs.

FAQ

Why another library around Dapper?

I've created this library to cover most of my own use-cases where in 90% I need just a few simple queries for CRUD operations using Dapper and don't want to write column names manually. All I need is a simple (anonymous) record with properties and want to have them filled from the query or to insert / update data.

How does the library works?

This library does two things:

1. Provides 4 computation expression builders for select, insert, update and delete. Those expressions create definitions (just simple records, no worries) of SQL queries.
2. Extends IDbConnection with few more methods to handle such definitions and creates proper SQL query + parameters for Dapper. Then it calls Dapper QueryAsync or ExecuteAsync. How does the library know the column names? It uses reflection to get record properties. So yes, there is one (the only) simple rule: All property names must match columns in the table.

Do I need to create a record with all columns?

You can, but don't have to. If you need to read a subset of data only, you can create a special view record just for this. Also if you don't want to write nullable data, you can omit them in the record definition.

And what about names mapping using Attributes or foreign keys magic?

Nope. Sorry. Not gonna happen in this library. Simplicity is what matters. Just define your record as it is in a database and you are ok.

Can I map more records from one query?

Yes. If you use LEFT or INNER JOIN, you can map each table to a separate record. If you use LEFT JOIN, you can even map the 2nd and/or 3rd table to Option (F# records and null values don't work well together). The current limitation is 3 tables (two joins).

What if I need to join more than 3 tables, sub-select or something special?

Fallback to plain Dapper then. Really. Dapper is an amazing library and sometimes there's nothing better than manually written optimized SQL query. Remember this library has one and only goal: Simplify 90% of repetitive SQL queries you would have to write manually. Nothing. Else.

Getting started

First of all, you need to init registration of mappers for optional types to have Dapper mappings understand that NULL from database = Option.None

Dapper.FSharp.OptionTypes.register()

It's recommended to do it somewhere close to the program entry point or in Startup class.

Example database

Let's have a database table called Persons:

CREATE TABLE [dbo].[Persons](
    [Id] [uniqueidentifier] NOT NULL,
    [FirstName] [nvarchar](max) NOT NULL,
    [LastName] [nvarchar](max) NOT NULL,
    [Position] [int] NOT NULL,
    [DateOfBirth] [datetime] NULL)

As mentioned in FAQ section, you need F# record to work with such table in Dapper.FSharp:

type Person = {
    Id : Guid
    FirstName : string
    LastName : string
    Position : int
    DateOfBirth : DateTime option
}

Hint: Check tests located under tests/Dapper.FSharp.Tests folder for more examples

API Overview

Table Mappings

You can either specify your tables within the query, or you can specify them above your queries (which is recommended since it makes them sharable between your queries). The following will assume that the table name exactly matches the record name, "Person":

let personTable = table<Person>

If your record maps to a table with a different name:

let personTable = table'<Person> "People"

If you want to include a schema name:

let personTable = table'<Person> "People" |> inSchema "dbo"

INSERT

Inserting a single record:

open Dapper.FSharp
open Dapper.FSharp.MSSQL

let conn : IDbConnection = ... // get it somewhere

let newPerson = { Id = Guid.NewGuid(); FirstName = "Roman"; LastName = "Provaznik"; Position = 1; DateOfBirth = None }

let personTable = table<Person>

insert {
    into personTable
    value newPerson
} |> conn.InsertAsync

Inserting Multiple Records:

open Dapper.FSharp
open Dapper.FSharp.MSSQL

let conn : IDbConnection = ... // get it somewhere

let person1 = { Id = Guid.NewGuid(); FirstName = "Roman"; LastName = "Provaznik"; Position = 1; DateOfBirth = None }
let person2 = { Id = Guid.NewGuid(); FirstName = "Ptero"; LastName = "Dactyl"; Position = 2; DateOfBirth = None }

let personTable = table<Person>

insert {
    into personTable
    values [ person1; person2 ]
} |> conn.InsertAsync

Excluding Fields from the Insert:

open Dapper.FSharp
open Dapper.FSharp.MSSQL

let conn : IDbConnection = ... // get it somewhere

let newPerson = { Id = Guid.NewGuid(); FirstName = "Roman"; LastName = "Provaznik"; Position = 1; DateOfBirth = None }

let personTable = table<Person>

insert {
    for p in personTable do
    value newPerson
    excludeColumn p.DateOfBirth
} |> conn.InsertAsync

NOTE: You can exclude multiple fields by using multiple excludeColumn statements.

UPDATE

let updatedPerson = { existingPerson with LastName = "Vorezprut" }

update {
    for p in personTable do
    set updatedPerson
    where (p.Id = updatedPerson.Id)
} |> conn.UpdateAsync

Partial updates are possible by manually specifying one or more includeColumn properties:

update {
    for p in personTable do
    set modifiedPerson
    includeColumn p.FirstName
    includeColumn p.LastName
    where (p.Position = 1)
} |> conn.UpdateAsync

Partial updates are also possible by using an anonymous record:

update {
    for p in personTable do
    set {| FirstName = "UPDATED"; LastName = "UPDATED" |}
    where (p.Position = 1)
} |> conn.UpdateAsync

DELETE

delete {
    for p in personTable do
    where (p.Position = 10)
} |> conn.DeleteAsync

And if you really want to delete the whole table, you must use the deleteAll keyword:

delete {
    for p in personTable do
    deleteAll
} |> conn.DeleteAsync

SELECT

To select all records in a table, you must use the selectAll keyword:

select {
    for p in personTable do
    selectAll
} |> conn.SelectAsync<Person>

NOTE: You also need to use selectAll if you have a no where and no orderBy clauses because a query cannot consist of only for or join statements.

Filtering with where statement:

select {
    for p in personTable do
    where (p.Position > 5 && p.Position < 10)
} |> conn.SelectAsync<Person>

To flip boolean logic in where condition, use not operator (unary NOT):

select {
    for p in personTable do
    where (not (p.Position > 5 && p.Position < 10))
} |> conn.SelectAsync<Person>

NOTE: The forward pipe |> operator in your query expressions because it's not implemented, so don't do it (unless you like exceptions)!

To use LIKE operator in where condition, use like:

select {
    for p in personTable do
    where (like p.FirstName "%partofname%")
} |> conn.SelectAsync<Person>

Sorting:

select {
    for p in personTable do
    where (p.Position > 5 && p.Position < 10)
    orderBy p.Position
    thenByDescending p.LastName
} |> conn.SelectAsync<Person>

If you need to skip some values or to take a subset of results only, use skip, take, and skipTake. Keep in mind that for correct paging, you need to order results as well.

select {
    for p in personTable do
    where (p.Position > 5 && p.Position < 10)
    orderBy p.Position
    skipTake 2 3 // skip first 2 rows, take next 3
} |> conn.SelectAsync<Person>

Option Types and Nulls

Checking for null on an Option type:

select {
    for p in personTable do
    where (p.DateOfBirth = None)
    orderBy p.Position
} |> conn.SelectAsync<Person>

Checking for null on a nullable type:

select {
    for p in personTable do
    where (p.LastName = null)
    orderBy p.Position
} |> conn.SelectAsync<Person>

Checking for null (works for any type):

select {
    for p in personTable do
    where (isNullValue p.LastName && isNotNullValue p.FirstName)
    orderBy p.Position
} |> conn.SelectAsync<Person>

Comparing an Option Type

let dob = DateTime.Today

select {
    for p in personTable do
    where (p.DateOfBirth = Some dob)
    orderBy p.Position
} |> conn.SelectAsync<Person>

JOINS

For simple queries with join, you can use innerJoin and leftJoin in combination with SelectAsync overload:

let personTable = table<Person>
let dogsTable = table<Dog>
let dogsWeightsTable = table<DogsWeight>

select {
    for p in personTable do
    innerJoin d in dogsTable on (p.Id = d.OwnerId)
    orderBy p.Position
} |> conn.SelectAsync<Person, Dog>

Dapper.FSharp will map each joined table into a separate record and return it as list of 'a * 'b tuples. Currently, up to 2 joins are supported, so you can also join another table here:

select {
    for p in personTable do
    innerJoin d in dogsTable on (p.Id = d.OwnerId)
    innerJoin dw in dogsWeightsTable on (d.Nickname = dw.DogNickname)
    orderBy p.Position
} |> conn.SelectAsync<Person, Dog, DogsWeight>

The problem with LEFT JOIN is that tables "on the right side" can be full of null values. Luckily we can use SelectAsyncOption to map joined values to Option types:

// this will return seq<(Person * Dog option * DogWeight option)>
select {
    for p in personTable do
    leftJoin d in dogsTable on (p.Id = d.OwnerId)
    leftJoin dw in dogsWeightsTable on (d.Nickname = dw.DogNickname)
    orderBy p.Position
} |> conn.SelectAsyncOption<Person, Dog, DogsWeight>

Aggregate functions

Aggregate functions include count, avg, sum, min, and max. To fully support these functions in builder syntax, the groupBy, groupByMany and distinct keywords are supported as well.

See this example of how to get the amount of persons having a position value greater than 5:

select {
    for p in persons do
    count "*" "Value" // column name and alias (must match the view record property!!!)
    where (gt "Position" 5)
} |> conn.SelectAsync<{| Value : int |}>

Or get the maximum value of the Position column from the table:

select {
    for p in persons do
    max "Position" "Value"
} |> conn.SelectAsync<{| Value : int |}>

Please keep in mind that work with aggregate functions can quickly turn into a nightmare. Use them wisely and if you'll find something hard to achieve using this library, better fallback to plain Dapper and good old handwritten queries™.

OUTPUT clause support (MSSQL & PostgreSQL only)

This library supports OUTPUT clause for MSSQL & PostgreSQL using special methods: InsertOutputAsync, UpdateOutputAsync and DeleteOutputAsync. Please check tests located under tests/Dapper.FSharp.Tests folder for more examples.

Deconstructor

To provide better usage with plain Dapper, this library contains Deconstructor converting Dapper.FSharp queries to a tuple of parameterized SQL query and Map of parameter values.

let r = {
    Id = Guid.NewGuid()
    FirstName = "Works"
    LastName = "Great"
    DateOfBirth = DateTime.Today
    Position = 1
}

let sql, values =
    insert {
        into personTable
        value r
    } |> Deconstructor.insert

printfn "%s" sql 
// INSERT INTO Persons (Id, FirstName, LastName, Position, DateOfBirth) 
// VALUES (@Id0, @FirstName0, @LastName0, @Position0, @DateOfBirth0)"

printfn "%A" values
// map [("DateOfBirth0", 11.05.2020 0:00:00); 
//      ("FirstName0", "Works");
//      ("Id0", 8cc6a7ed-7c17-4bea-a0ca-04a3985d2c7e); 
//      ("LastName0", "Great");
//      ("Position0", 1)]

IncludeColumn vs ExcludeColumn (there can be a 🐲)

New keywords added in v2 - excludeColumn and includeColumn are a great addition to this library, especially when you want to do partial updates / inserts. However, be aware that you should never mix both in the same computation expression!

ExcludeColumn

If used for the first time within computation expression all fields from the record will be used and removed (ignored) those you provided in a keyword. When used more times, already filtered fields will be filtered again.

IncludeColumn

If used, only specified columns will be used and all the others will be ignored.

With great power comes great responsibility.

Contribution Guide

Every new idea of how to make the library even better is more than welcome! However please be aware that there is a process we should all follow:

• Create an issue with a description of proposed changes
• Describe the expected impact on the library (API, performance, ...)
• Define if it's minor or breaking change
• Wait for Approve / Deny
• Send a PR (or wait until taken by some of the contributors)