Okay, so I run into a couple of problems with NUnit versioning, but it's here. I added support for nullable parameters in Brail, so now you can do:
And if the user has not been set, it'll return null. You can get the latest bits here. I won't publish a Castle Demo App today, but I'll certainly attempt to use the new feature tomorrow in the Ajax part.
Here
are the graphs that I promised; we'll start off with the number of posts in
each month:
I do
wonder what happened in June, 05, I posted like mad.
Here
are the posts per day of the week:
Every once in a while you see a couple of messages from the guys that use a Blog with a DB backend post some wonderful statistics about their Blog. I couldn't be happier with dasBlog (can you say zero headache in over two years of using it?), but I really like the statistics abilities. Well, I like to think of myself as a developer, and as such, I can solve such problems. Put this together with the SQL stuff I've been doing lately, and that dasBlog has Web Services functionality, and the solution is obvious, isn't it?
I'm going to explain how I did it now, and then post the stats in the next post (along with some nice graphs, I hope.) The first thing that I did was to take my previous post about dasBlog's Web Services, and took it for a spin, here are the results:
import System
import System.Data
import System.Data.SqlClient
import CookComputing.XmlRpc
import newtelligence.DasBlog.Runtime.Proxies
user = "Ayende Rahien"
password = "Ayende@ayende.com"
start = DateTime.Now
log = IO.File.CreateText("log.txt")
dasBlog = BloggerAPIClientProxy()
dasBlog.Url = "http://www.ayende.com/Blog/blogger.aspx"
posts = dasBlog.metaweblog_getRecentPosts("",user, password, 1500)
print "Got ${posts.Length} posts from server"
using con = SqlConnection("""Data Source=.;AttachDbFilename=F:.mdf;Integrated Security=True;User Instance=True"""):
con.Open()
using command = con.CreateCommand():
command.CommandText = "INSERT INTO BlogStats(Time, Title) VALUES(@time, @title)"
time = command.Parameters.Add("time", SqlDbType.DateTime)
title = command.Parameters.Add("title", SqlDbType.NChar)
for post in posts:
try:
time.Value = post.dateCreated
title.Value = post.title
command.ExecuteNonQuery()
except e:
print "Error updated post: ${post.title}"
print e.Message
log.WriteLine(e.ToString())
log.WriteLine("--- --- ---")
print "Took ${DateTime.Now - start}"
Before we start with our regular content, a couple of bug fixes, the ValidateLength on the Name and Email properties of the User class should looks like this:
[
ValidateLength(0,50)]Specifying just one number means the exact length of the string.
Okay, so far we have seen how to edit users and then save them to the database, but we haven't seen yet how to actually create the users. For myself, I think the users will be less than inclined to add themselves to the database by hand, using SQL, and I can just imagine the expression on the face of the DBA when he hears this suggestion.
Okay, I hear you, I won't bring that up again, looks like we have to do some work to make it happen. The question is, how much work is needed? We did a fair amount of work yesterday, to support editing users, how much of this work do you think we can use to support creating users?
Before anything else, I'm getting tired of having to remember all those URLs in my head, we'll add an index page to our administration section. Add an empty method called Index() to the AdminController class and create the following view at Views\Admin\Index.boo
<ul>
<li>
${HtmlHelper.LinkTo('Admin Projects','projects')}
</li>
<li>
${HtmlHelper.LinkTo('Admin Users', 'users')}
</li>
</ul>
Now we create the CreateUser method:
public void CreateUser()
{
RenderView("EditUser");
}
In the EditUser.boo view, we need to make a single change, and that is to initialize the id varaible to "0" instead of empty string, like this (fragement):
<?brail
name = ""
email = ""
id = "0"
Let's beef up our SaveUser method, to make it capable of handling both existing and new users, the new stuff is bolded:
public void SaveUser([ARDataBind("user",AutoLoadUnlessKeyIs=0)]User user)
{
RenderView("GenericMessage");
try
{
user.Save();
Flash["good"] = string.Format(
"User {0} was saved successfully.", user.Name);
}
catch (Exception e)
{
Flash["DataBindErrors"] = user.ValidationErrorMessages;
Flash["bad"] = e.Message;
}
}
I added some better handling for errors, and we now direct the framework to automatically load the user from the database unless the id is equal to 0. I think that this may require a couple of changes to the GenericMessage view, like this:
<p>
<font color="red">
<?brail
if IsDefined('bad'):
output bad
end
?>
</font>
</p>
<?brail if IsDefined('DataBindErrors'): ?>
<table>
<tr>
<thead>Validation Errors</thead>
</tr>
<?brail for error in DataBindErrors: ?>
<tr>
<td>${error}</td>
</tr>
<?brail end ?>
</table>
<?brail end ?>
<p>
<font color="green">
<?brail
if IsDefined('good'):
output good
end
?>
</font>
</p>
Okay, we're done with creating users. Play with it a bit, try to add a non email adress, put it strings that are too long, use names that already eixst, etc. I won't claim that this is the greatest UI in the world, but it's functional.
Next stop, deleting users, should be simple. For now, we'll build it using normal pages & post backs. In a short while, I'll see how to use the Ajax support in MonoRail. We'll start by editing the users.boo page to add a link to the delete page, like this (fragement):
<td>
<?brail
output HtmlHelper.LinkTo("Delete","admin","ConfirmDeleteUser",user.Id)
?>
</td>
As long we're on a roll, we'll add a link to the create user page, like this (fragement):
${HtmlHelper.LinkTo('Create New User','admin','CreateUser')}
Now we get to write the delete functionality itself. We'll start with Views\Admin\ConfirmDeleteUser.boo, which looks like this:
<p>
${user.Name} says he is sorry, are you still angry?
</p>
<p>
${HtmlHelper.LinkTo('Yes, Delete!', 'Admin', 'DeleteUser',User.Id)}
<p>
</p>
${HtmlHelper.LinkTo('Well, no...', 'Admin', 'Users')}
</p>
And now to write the code that will actually handle all of this. Here it is in full:
public void ConfirmDeleteUser([ARFetch("id",Create=false)]User user)
{
if (user == null)
{
RenderView("NoSuchUser");
return;
}
PropertyBag.Add("User", user);
}
public void DeleteUser([ARFetch("id", Create = false)]User user)
{
if (user == null)
{
RenderView("NoSuchUser");
return;
}
user.Delete();
RedirectToAction("Users");
}
This require some explanation. The ARFetch attribute will take an id and will pull it out of the database and into our parameter (passing null if it can't find it). We just make sure that it is handled properly. This is it, we know has full CRUD for a user. I'm going to leave the implementation of similar operation for projects as an exercise for the reader.
Next time, we'll see how we can add some Ajax stuff to it, so we can keep up with the hype.
One of my wishes just came true today, Boo just got a new feature that I wanted for a long time. It's not something that you're likely to find in any way useful unless you mess with the guts of Boo. The new feature is adding @,$,? prefixes to the Boo Parser. Why is this a good thing?
Beacuse it let me, and any other Boo Macro Developer, has more ways to get infromation from the user without it being a true pain in the ass. For instance, I can have this:
with longVariableName:
@Foo = "name"
@Blah = "foo"
You could do this previously, but this require a naming convention (prefix everything with a _, which is a valid variable name). Another possibility is this:
ignoreNulls:
?something.Foo()
This can take anything prefixed with a ? and emit code that would skip the call if the variable is null. I'm going to be putting it in Brail in the next couple of hours (hopefully :-) ).
Okay, I just went through a SQL refactoring tha I consider interesting. We'll start with this code:
Cursor transactions IS
SELECT L.CustomerId ,
Sum(Y.Debt) Sum_Debt ,
Sum(Y.Credit) Sum_Credit
FROM Transactions Y join
Customers L
ON L.CustomerId = Y.CustomerId
AND L.Status NOT IN (1 , 2 , 3)
GROUP BY L.CustomerId
For rec0 IN transactions
UPDATE CustomerStatus
SET Debt = rec0.Sum_Debt,
Credit = rec0.Sum_Credit
WHERE CustomerId = rec0.CustomerId
End
The above is using Oracle syntax, which I find much saner for cursors than what SQL Server requires you to do. This is undoubtly one of the reasons that cursors are so much more popular in Oracle than in SQL Server.
From now on I'm using SQL Server syntax, btw. The first thing that I did was to remove the cursor. I did it by simply adding a FROM clasue to the update, like this:
UPDATE CustomerStatus
SET Debt = rec0.Sum_Debt,
Credit = rec0.Sum_Credit
FROM
(
SELECT L.CustomerId ,
Sum(Y.Debt) Sum_Debt ,
Sum(Y.Credit) Sum_Credit
FROM Transactions Y join
Customers L
ON L.CustomerId = Y.CustomerId
AND L.Status NOT IN (1 , 2 , 3)
GROUP BY L.CustomerId
) rec0
WHERE CustomerId = rec0.CustomerId
WITH rec0 (CustomerId, Sum_Debt, Sum_Credit) AS
(
SELECT L.CustomerId ,
Sum(Y.Debt) Sum_Debt ,
Sum(Y.Credit) Sum_Credit
FROM Transactions Y join
Customers L
ON L.CustomerId = Y.CustomerId
AND L.Status NOT IN (1 , 2 , 3)
GROUP BY L.CustomerId
)
UPDATE CustomerStatus
SET Debt = rec0.Sum_Debt,
Credit = rec0.Sum_Credit
FROM rec0
WHERE CustomerId = rec0.CustomerId
I find this much easier to understand than the previous one, and it's performance it just fine. One thing I wish I could do is either nest WITH statements or cascade them.
By nesting I mean something like:
WITH DrunkCustomers (CustomerId) AS
(
WITH Bar(CustomerId)
AS
{
SELECT CustomerId FROM Transactions
WHERE Payee like '%bar%'
}
SELECT L.CustomerId ,
Sum(Y.Debt) Sum_Debt ,
Sum(Y.Credit) Sum_Credit
FROM Bar Y join
Customers L
ON L.CustomerId = Y.CustomerId
AND L.Status NOT IN (1 , 2 , 3)
GROUP BY L.CustomerId
)
By cascading I mean:
WITH Bar(CustomerId)
AS
{
SELECT CustomerId FROM Transactions
WHERE Payee like '%bar%'
}
WITH DrunkCustomers (CustomerId) AS
(
SELECT L.CustomerId ,
Sum(Y.Debt) Sum_Debt ,
Sum(Y.Credit) Sum_Credit
FROM Bar Y join
Customers L
ON L.CustomerId = Y.CustomerId
AND L.Status NOT IN (1 , 2 , 3)
GROUP BY L.CustomerId
)
Sadly, none of those options works.
So Microsoft has posted a lot more snippets, a Jeff has made them more accessible. I downloaded them and took a look. I'm pretty sure that you could hear my cries of horror all the way to Remond.
Why? That is the question I have. Snippets are useful for things like generating a foreach (although the Microsoft implementation sucks, try the way ReSharper does it.) or maybe generating a method skeleton. I see them doing things there like Hashing A Password, Encrypting Files, etc.
Why? This is just a recipe for disaster. If I want to hash a password, I write a method to do it and then I call it. I don't create a template for hashing a password. By all means, provide more utility methods, but don't do it like this. Allow me to hardwire method calls to intelli sense, cool. Encouraging users toward code duplication, Not Cool.
Okay, I got around (finally!) to work a little with Brail, and the end result is that there were a plenty of improvements:
You can get the latest stuff from the Castle Repository.
Have fun.
Patrick Logan has a post web services. I've very little to do with web services so far (notice the lack of posts complaining about stuff in Indigo :-) ).
What caught my eye was this quote:
I agree whole heartedly, and for much more than merely SOAP or Web Services.
No future posts left, oh my!
+972 52-548-6969
