Quey across a join in nHibernate

My god, I don't think I've ever come across any framework with quite such a poor popularity-to-documentation framework as Fluent nHibernate. Everyone is using it, but no one writes down how.

Stupid nHibernate.

So, after spending ages trying to get something as simple as a rowcount, today I have spent an equally frustrating period trying to get a results set which does a logic OR operation across a joined table. Something as petty as this, in SQL terms:


SELECT * from product p
inner join sku s on s.id = p.id
WHERE p.Name like '%widget%'
OR s.SkuCode like '%wd_%'


The issue arises because even though my mappings are correct, nHibernate throws a wobbler at you if you start with Product and try and get it to recognise that it should be able to filter on a field in the joined sku table. This, for example, compiles but fails horribly even though to my mind it's the intuitive way of presenting the above query in nHibernate:


Dim results As ArrayList = session.CreateCriteria(Of DataTransferObjects.Product) _
.Add(Expression.Or( _
Expression.Like("Name", "widget"), _
Expression.Like("SkuCode", "wd_"))) _
.List()


No, what you have to do is set up a CreateAlias to the table and use that:

Dim results As ArrayList = session.CreateCriteria(Of DataTransferObjects.Product) _
.CreateAlias("Skus", "sku") _
.Add(Expression.Or( _
Expression.Like("Name", "widget"), _
Expression.Like("sku.SkuCode", "wd_"))) _
.List()


Certainly not the way I'd expect a "highly intuitive", productivity-enhancing framework to function. Not by a long shot.

I could have done the required work by creating a couple of custom data handling objects and accompanying T-SQL stored procedures in about ten minutes. To do the same thing in nHibernate has taken two days. What's that about enhancing productivity?

nHibernate Projections in VB.net

Well, this has lead me a merry dance, I can assure you. You'd think it was pretty straightforward to do something as simple as a row count in nHibernate, wouldn't you? Well, in point of fact it actually is pretty simple but it's just not well documented anywhere. Particularly perplexing is the fact that most of the examples seem to have been ripped straight from the Java hibernate implementation and I'm not entirely sure this will even port to C# and work, let alone translate to VB (I haven't tried in C#).

Anyway, just in case some poor soul has the same problem and gets stuck, this is how you do it:


Dim pResults As IList = session.CreateCriteria(GetType(dto.Product)) _
.SetProjection(Projections.RowCount()).List()


Of course "dto.Product" is a custom class I'm using that's specific to what I'm trying to do, so stick your own in there and you should be good to go.

Date Conversions in SQL

Does what it says on the tin. I find this a particularly handy trick to know when trying to copy & paste SQL data from Query Analyser into Excel, because the default date/time format causes Excel to scramble the dates into nonsense.

Stupid Excel.

Anyway, on with the show.


CONVERT(CHAR(19),GETDATE()) -- gives 31 March 2003 11:19
CONVERT(CHAR(11),GETDATE(),106) -- gives 31 March 2003


A SQL Split function

This is so useful it's beyond me why it isn't adopted as standard functionality. Maybe it is in SQL Server 2008. In the meantime, here's how I do it.


CREATE FUNCTION fnSplit(@Data nvarchar(4000),
@Delimiter varchar(10) = ',')
RETURNS @tblSplit TABLE

(ItemId int IDENTITY(1, 1) NOT NULL PRIMARY KEY,
Item nvarchar(4000) NULL)

AS

BEGIN

DECLARE @Delimiter2 varchar(12),
@item nvarchar(4000),
@iPos int,
@DelimWidth int

--had to do this cuz if they send in a > 9 char delimiter I could not pre and post append the % wildcards
SET @Delimiter2 = @Delimiter
SET @Delimiter2 = ISNULL(@Delimiter2, ',')
SET @DelimWidth = LEN(@Delimiter2)

IF RIGHT(RTRIM(@Data), 1) <> @Delimiter2
SELECT @Data = RTRIM(@Data) + @Delimiter2

IF LEFT(@Delimiter2, 1) <> '%'
SET @Delimiter2 = '%' + @Delimiter2

IF RIGHT(@Delimiter2, 1) <> '%'
SET @Delimiter2 = @Delimiter2 + '%'

SELECT @iPos = PATINDEX(@Delimiter2, @Data)

WHILE @iPos > 0
BEGIN
SELECT @item = LTRIM(RTRIM(LEFT(@Data, @iPos - 1)))
IF @@ERROR <> 0 BREAK
SELECT @Data = RIGHT(@Data, LEN(@Data) - (LEN(@item) + @DelimWidth))
IF @@ERROR <> 0 BREAK
INSERT INTO @tblSplit VALUES(@item)
IF @@ERROR <> 0 BREAK
SELECT @iPos = PATINDEX(@Delimiter2, @Data)
IF @@ERROR <> 0 BREAK
END

RETURN
END



You call it like this:


select item from [reference]..fnSplit( @variable, ',' )


Where @variable is a varchar containing your comma-delineated string. Useful for sticking directly into IN clauses.

Reading XML with XPath

There's more than one way to read an XML file in dot net, but the XPath model doesn't seem to be terribly widely documented. Which is a shame because although it's a little unintuitive, it offers some functionality that doesn't come with other tools.


using System.Xml.XPath;

XPathDocument xDoc = new XPathDocument(“filepath”);
XPathNavigator xNav = xDoc.CreateNavigator();

Xnav.MoveToRoot();
Xnav.MoveToFirstChild();

do
{
//do stuff .. like xNav.GetAttribute(""AttributeName", "");
}
while (Xnav.MoveToNext());


So, load your xml file by replacing "filepath" with the path of the - ahem - file you want to load. MoveToRoot and MoveToFirstChild do exactly what you'd expect them to, and you can obviously vary these commands depending on where you want to start reading. Then the do-while loop should pass through each node in turn, allowing you to access its attributes and such.

SQL Cursors

SQL cursors, for those who don't know, are effectively a way of introducing a loop into a chunk of SQL code. Quite why the people who invented SQL couldn't call them "loops" I don't know - DBAs always have to be different, I suppose. Anyway it's a handy thing to know.

Beware though that SQL cursors are resource-intensive. If efficiency at the database end is important to you, it might be better to return the data and loop through it in the handling code. However, there are times when you need to get data, loop through it and then do something else at the DB end with the results of the loop. In those cases using a cursor is likely to be less intensive than the round trips involved in two separate back-and-forth trips to the database server.


Declare c1 cursor read_only
For
Select from
-- select statement picks up
--rows to sort through

Declare @trackingID int
-- declare a variable to hold the primary key
-- of the rows you’re returning

Open c1

Fetch next from c1 into @trackingID

While @@fetch_status = 0
Begin
--do something with the tracking variable,
--such as using it to select some
--data or do an update or somesuch

Fetch next from c1 into @trackingID
End

Close c1
Deallocate c1


Writing files to disc

This is a simple enough procedure but, in honesty, I always, always forget how to do it, and when you go and look it up on the interweb it seems that the examples provided are far more complicated than they really need to be. I guess it must be that the System.IO namespace simply isn't as intuitive as perhaps it should be. So here's the stripped down, basic code you need to write a file to disc - don't forget that the namespace goes at the top of your file, while the actual operational code goes inside it.


using System.IO;

FileStream myStream = new FileStream("insert File path here", FileMode.Create);
StreamWriter myWriter = new StreamWriter(myStream);

myWriter.Write("Some Text");
myWriter.Write(Environment.NewLine);
//this creates a new line. Duh!

myWriter.Close();
myStream.Close();


So put the target path into "insert File path here" and put your text into where it says "some text". Obviously you can keep adding as much text or new lines as you need.