PetaPoco – A simple web app
I’ve put together a basic web app using PetaPoco to manage a one-to-many relationship between authors and their articles.
The author poco contains just the author’s name, id and a list of articles the author has written :
public class Author
{
public int Id {get; set;}
public string Name {get; set;}
[ResultColumn]
public List<Article> Articles {get; set;}
public Author()
{
Articles = new List<Article>();
}
}
The Author.Articles list is a [ResultColumn] meaning that PetaPoco won’t try and persist it automatically, it’s ignored during inserts and updates. The Article poco has more properties :
public class Article
{
public int? Id {get; set;}
public string Title {get; set;}
public string Body {get; set;}
[DisplayFormat(DataFormatString = "{0:dd/MM/yyyy}", ApplyFormatInEditMode=true)]
public DateTime Date {get; set;}
[ResultColumn]
public Author Author {get; set;}
[Column( "author_id")]
[DisplayName( "Author")]
public int? AuthorId {get; set;}
public Article()
{
Date = DateTime.Now;
}
}
The Date property has a date format attribute so that in the \Article\Edit.cshtml view @Html.EditorFor(model => model.Article.Date) will display the date how I want it. The Author property is another ResultColumn only used when displaying the Article, not when updating it. In the ArticleRepository I use the MulltiPoco functionality in PetaPoco to populate the Author when the Article record is retrieved.
Setup
By default the app uses the tempdb at .\SQLEXPRESS. This is defined in the web.config :
<add name="PetaPocoWebTest" connectionString="Data Source=.\SQLEXPRESS;Initial Catalog=tempdb;Integrated Security=True" providerName="System.Data.SqlClient" />
The tempdb is always present in SQL Server and is recreated when the database is stopped and started. You can change this to suit your environment. The app will auto generate the tables for you if they don’t exist. In theory if you have SQL Express running you don’t need to do anything to get started
A few details
There are two repository classes; AuthorRepository and ArticleRepository. These classes are where the PetaPoco work lives. As well as loading the Author the AuthorRepository class also loads in the list of articles by that author. I use a relationship relator callback class as demonstrated in the recent PetaPoco object relationship mapping blog post. I made a slight tweak to handle objects that have no child objects at all otherwise there was always a new() version of Article in the list. The fix is simple :
if( article.Id != int.MinValue)
{
// Only add this article if the details aren't blank
_currentAuthor.Articles.Add( article);
}
I tried to get this working with an int? Id, but had type conversion problems in PetaPoco. I’ll investigate that later as it would be useful to have int? Id to indicate a blank record
I’ve used the Glimpse PetaPoco plugin by Schotime (Adam Schroder). Because the Glimpse plugin uses HttpContext.Current.Items all the SQL statements are lost when RedirectToAction() is called from a controller. I got around this by copying the PetaPoco Glimpse data into TempData before a redirect, and from TempData back into the Items collection before the controller executes.
This means you still see update/insert/delete SQL. It is definitely a requirement to perform a RedirectToAction() after such database calls so that it isn’t possible to refresh the screen and re-execute the commands (ie; cause havoc). If the user did refresh the screen (or otherwise repeat the last request) they’d just get the redirect to action and not the request that fired the database action.
I expect this to be a common pattern with Glimpse plugins that use response redirects.
UPDATE
Many thanks to Schotime who showed me that the Remote tab in Glimpse shows you the last 5 requests. All the SQL is there in Glimpse if you look for it. I didn’t realise this when I added the TempData solution
:) but at least using TempData the SQL commands prior to a redirect are visible right on the PetaPoco tab without having to drill down to the previous request, which I kinda like as it’s more intuitive and effortless.
Hope this is useful to someone. You can download the source at GitHub. This is my first project using git so if there are any problems please let me know.
Finally, I’ve cut my finger and I’m angry with my cat.
Help getting started with PetaPoco
I’ve recently spent some time working with PetaPoco. As always with something new I found problems early and often but not real problems with PetaPoco itself, just typical “What does this error mean?” and “How do I do this then?”
The kind of objects I want to use with PetaPoco are domain entities so there’s a bit more meat to them than a flat poco. Here are the test classes I started with :
public class Author
{
public int id {get; internal set;}
public string name {get; internal set;}
}
public class Article
{
public int id {get; internal set;}
public string content {get; internal set;}
public string title {get; internal set;}
public Author author {get; internal set;}
}
The property setters are internal so that client (application) code using the entities cannot change their values. The only way these properties would be changed is either :
- in the repository when the entities is being loaded (hydrated)
- during construction or through a builder class
- through public methods on the entities
In my real code I’ll use the InternalsVisibleTo attribute on the classes to ensure that outside of the domain classes only the repository assembly can access the setters. This test is stripped right back so the only remnant of these requirements are the internal setters. Also notice that there is a relationship between Article and Author; when I load an Article I want the Author to be brought back too. So now here are the learning curve issues I had :
Querying data
Let’s load a list of articles from the database and join the author table too :
db.Query(
"SELECT * FROM article left outer join " +
"author on author.id = article.author_id")
This throws an exception :
Unhandled Exception: System.ArgumentNullException: Value cannot be null. Parameter name: meth at System.Reflection.Emit.DynamicILGenerator.Emit(OpCode opcode, MethodInfo meth)
This happens because PetaPoco currently only looks for public setters when mapping to pocos and all my setters are internal. This is easily fixed in PetaPoco.cs by changing all instances of GetSetMethod() to GetSetMethod( true) (update: this change was added to PetaPoco core shortly after this post, so you won’t need to apply it)
Now I want to load a single article. Let’s try using the new multi-poco support :
var article = db.Single(
"SELECT * FROM article left outer join author" +
" on author.id = article.author_id");
This doesn’t compile because the multi-poco feature only applies to the Query() and Fetch() functions at the moment so I can only pass in the type of the result which is Article. The solution is simple :
var article = db.Query(
"SELECT * FROM article left outer join author " +
" on author.id = article.author_id where article.id=@0",
1).SingleOrDefault();
Inserting data
Ok let’s try and insert a new article. We’ll leave out the author for now and see if the foreign key constraint in the database kicks in.
int newRecordId = (int)db.Insert( new Article() {title="Article 1",
content="Hello from article 1"});
PetaPoco didn’t like it :
Unhandled Exception: Npgsql.NpgsqlException: ERROR: 42P01: relation “Article” does not exist at Npgsql.NpgsqlState.d__a.MoveNext()
I’m using PostgreSQL and it seems that it is case sensitive, if I change my class names to all lower case it works. But I don’t want to do that so all I need to do is use the TableName attribute which overrides the convention of class name equals table name :
[TableName( "author")]
public class Author
{
public int id {get; internal set;}
public string name {get; internal set;}
}
[TableName( "article")]
public class Article
{
public int id {get; internal set;}
public string content {get; internal set;}
public string title {get; internal set;}
public Author author {get; internal set;}
}
Now I can get a bit further but I still get an error :
Unhandled Exception: Npgsql.NpgsqlException: ERROR: 42703: column “author” of relation “article” does not exist at Npgsql.NpgsqlState.d__a.MoveNext()
This happens because my poco has an author property and this doesn’t map directly to the author_id column on the database. The author_id column is an integer, so how can I pass my poco into Update() and have it use an author id? You can’t. Instead you need to use a different overload of the Insert() method :
int newRecordId = (int)db.Insert( "article", "id", true,
new {title="A new article",
content="Hello from the new article",
author_id=(int)1});
We are now no longer using the poco, we are using an anonymous type. This is doubly useful because we now have the option to just update specific columns if we want to. Code calling my repository will still pass in a poco but the repository will strip out the values it wants to update. I like that.
In this example I already knew the author id because the user will have selected the author from a drop down while entering the article details.
Updating data
Now let’s update an existing article :
var newArticle = db.Query(
"SELECT * FROM article left outer join author on" +
"author.id = article.author_id where article.id=@0",
1).SingleOrDefault();
newArticle.content += " The end.";
db.Update( newArticle);
Here is this code’s exception :
Unhandled Exception: System.InvalidCastException: Can’t cast PetaPoco_MultiPoco_
Test.Author into any valid DbType.
at Npgsql.NpgsqlParameter.set_Value(Object value)
Again, PetaPoco can’t map the author property to a column on the article table. The way around this is the tell PetaPoco that the author property shouldn’t be used during inserts or updates. We use the ResultColumn attribute for this :
public class Article
{
public int id {get; internal set;}
public string content {get; internal set;}
public string title {get; internal set;}
[ResultColumn]
public Author author {get; internal set;}
}
Now we get another error, remember that I’m using PostgreSQL here :
Unhandled Exception: Npgsql.NpgsqlException: ERROR: 42703: column “ID” does not exist at Npgsql.NpgsqlState.
d__a.MoveNext()
This happens because PetaPoco will use “ID” as the default primary key if you don’t specify it. There are three solutions to this. Firstly I could change my database to use “ID” as the column name. Secondly I could change PetaPoco.cs to use “id” as the default. This would be done in the constructor for the PocoData internal class :
TableInfo.PrimaryKey = a.Length == 0 ? "id" : (a[0] as PrimaryKeyAttribute).Value;
Thirdly, and preferably, I’ll use the PrimaryKey attribute on my poco classes :
[TableName( "author")]
[PrimaryKey("id")]
public class Author
{
public int id {get; internal set;}
public string name {get; internal set;}
}
[TableName( "article")]
[PrimaryKey("id")]
public class Article
{
public int id {get; internal set;}
public string content {get; internal set;}
public string title {get; internal set;}
[ResultColumn]
public Author author {get; internal set;}
}
Also note that if we want to update the author we’d just do a separate call to Update(). If I find anything else worth sharing I’ll update this post
The move towards micro ORMs
Many of us know about the apparently recent appearance of micro ORMs thanks to Rob Conery’s Massive followed by Sam Saffron’s (and Stack Overflow’s) Dapper and more. However, having read around the web it is clear that micro ORMs have existed for as long as ADO.NET, but without the publicity. Some developers don’t like their SQL taken away from them and don’t like the idea of using a heavy ORM if they can help it. So they write their own object mapper that works for them. Of course there are many advantages to using a mature fullsome ORM, it just depends on how you look at it and what’s important in your world
I’ve been resisting using LINQ to SQL / EF / NHibernate partly because I was concerned about efficiency and control but mainly because at work I’ve just not needed to change the data access layers nor have I had the opportunity to do so. I inherited a data access layer that mainly uses stored procedures and ADO.NET and there’s no need to change this. There’s nothing wrong with it and it’s stable and switching to an ORM would be virtually impossible. Does a stored procedure heavy application even suit an ORM? Perhaps for reads
But at home it’s different
At home it’s different. I’m building an application that is still in it’s early stages so I can spend time experimenting with micro ORMs. From what I’ve seen so far I like them all because :
- they are closer to the metal
- the source code is right there for you to step into
- they are much faster than full ORMs
- the learning curve is much shorter
The downside is that you need to know a little SQL to make the most of them, but this excites me! SQL is already a mature ubiquitous DSL for accessing data. If you can talk this language you are at a serious advantage than if you’re relying on an ORM to generate the SQL for you. Dapper and PetaPoco both share a slight optimisation complication in that they both dynamically generate poco mapping methods for your pocos using System.Reflection.Emit.ILGenerator. I tend to skip past that bit in the source for now 
but guess what… these two are the fastest C# based micro ORMs available right now
I’ve experimented a little with Massive, Simple.Data, Dapper and PetaPoco with PetaPoco looking like my personal favourite. I’ve spent most of my experimentation time with PetaPoco so I’ll follow this post up with a list of learning curve problems I had.