Linq-to-Sql var null question or observation

[RayLopez99]

This is more an observation than a question. I notice in the below
the 'catch' is never triggered, even if I enter a value that I know is
not present in MyTable. Thus I know for example the Table column
AField, holding strings, does not have any rows that start with "Z",
but if I input aString = "ZZZZ", I don't get var MyMatchingTables =
null, but rather I get zero hits (or .Count = 0). Must be built into
Linq-to-Sql. I've been using it for years but this is the first time
I thought about it, since I just read somewhere that keyword 'var'
does not support any nulls, or rather the right hand side source
cannot return a null.

RL

public List<MyTable> GetMyTablesByName(string aString)
{
List<MyTable> myMyTableList = new List<MyTable>();

try
{
DataContext database = new
DataContext(MyConnectionString);

var MyMatchingTables = from X in db.MyTables
where
X.AField.StartsWith(aString)
select X;

//why is var MyMatchingTables never null? Must be built into Linq to
Sql to never return null?

myMyTableList =
MyMatchingTables.ToList();
}
catch (Exception ex)
{
Debug.WriteLine(ex.Message); //never triggered, even
if no match

}

return myMyTableList;
}

 

[RayLopez99]

On Sat, 25 Dec 2010 00:55:28 -0800 (PST), RayLopez99


How is a List<T> null?

 It's a
 collection. The collection can be insantiaed new and nothing put in
it. Collection.count = 0.

That's what is coming out of the query even if you didn't use
ToList(), an empty  List<T> being created.


Now if you were using First() or FirstorDefault(), the var could be
null, since the object may not be made, because it didn't get a hit
and a single instance of the object was not returned.

var MyMatchingTables = (from X in db.MyTables where
X.AField.StartsWith(aString) select X).FirstorDefault(); //syntax may
be wrong, but you get the point

So this could return a var MyMatchingTables = null? But I thought in
Linq-to-Sql that is not allowed? So it will throw an exception (in
the try/catch block)?

Interesting, please confirm.

One way around this null business of course is to create a custom
object and 'new' it within the query, but that's another question--I'm
curious about this code.

RL

 

[RayLopez99]

The key is to understand the Linq statements and know what they return
or not return when encapsulating data in objects.

Thanks. As a practical matter, it doesn't matter for me since in my
code (web method) it will either return via the try or the catch
block, so if I get some error server side on the client side it won't
be noticeable. Still, at some point I should look more carefully into
this because for performance reasons I think it's better not to rely
on try/catch.

Now that I have your attention, please tell me what you think of
"Reactive Programming" in .NET (Rx)

Rx programming (a push model paradigm, like ObservableCollection):

http://themechanicalbride.blogspot.com/2009/07/introducing-rx-linq-to-events.html

http://blogs.msdn.com/b/rxteam/arch...active-extensions-rx-for-net-silverlight.aspx

http://channel9.msdn.com/shows/Goin...-and-IObservableIObserver-in-the-BCL-VS-2010/

From what I can tell in the videos, and I've not gone though them
carefully, the Rx model might be useful in SQL (ADO.NET) asynchronous
loading where you have several web methods and one depends on the
other. Lazy loading. For example two lists boxes, where the second
list box depends on what you select in the first listbox, and both
inside a grid. To date I've not had an easy way of doing this
(typically the second list box will not load in time and it will stay
null--I've tried all the event handlers I could think of). What I end
up doing, and this works as a workaround, is to load the second list
box "outside" the grid, or "downstream" (later in time) from the first
list box. That works fine as a workaround, but with the Rx model you
can have stuff "magically" pop up on your screen as soon as it's done,
"in it's own sweet time". I think.

Very cool paradigm in any event. Also check out this easy to do mouse
drag and drop using Rx programming:
http://channel9.msdn.com/blogs/j.van.gogh/writing-your-first-rx-application

I'm curious what a Big League .NET architect like you thinks of this
stuff (apparently Rx is still Beta--maybe in Visual Studio 20xx it
will come out).

RL

 

[Registered User]

Just for absolute clarity ... a List is created when ToList() is

var MyMatchingTables = (from X in db.MyTables where
X.AField.StartsWith(aString) select X).FirstorDefault(); //syntax may
be wrong, but you get the point

So this could return a var MyMatchingTables = null? But I thought in
Linq-to-Sql that is not allowed? So it will throw an exception (in
the try/catch block)?

The FirstOrDefault method may return null because null is a perfectly
valid result for a SQL query to return. Attempting to use this null
value may result in a null reference exception being thrown.

Interesting, please confirm.

You can easily confirm this behavior for yourself by writing some
code. Use a console app and set the data context's Log property to
Console.Out, something like this :

CustomNorthwindDataContext dc = new CustomNorthwindDataContext();
dc.Log = Console.Out;
var customer = (from c in dc.Customers
where c.CustomerId.Equals(String.Empty)
select c).FirstOrDefault();


Setting the Log property will let you examine the underlying TSQL.
When FirstOrDefault is used against the query results, the TSQL
statement starts with 'SELECT TOP 1". When there is no TOP 1 to select
the query returns null which becomes the FirstOrDefault method's
return value.

One way around this null business of course is to create a custom
object and 'new' it within the query, but that's another question--I'm
curious about this code.

Doing this won't accomplish much of anything.

var customer = from c in dc.Customers
where c.CustomerId.Equals(String.Empty)
select new { c.CustomerId, c.CompanyName};

The query above returns no results so no custom objects will be
created. The returned enumeration will be empty.

regards
A.G.

 

[Registered User]

I don't see your point here. If more than one object is being returned
out of the query, you don't need ToList() it's going to return a List<T>
by itself of objects in a tubular form.

A ToList() is used to force eager loading as opposed to lazy loading.

<http://weblogs.asp.net/morteza/arch...-to-force-linq-query-execute-immediately.aspx>

I stand corrected, thank you.
regards
A.G.

 

[Registered User]

It also applies to Linq used with ADO.NET Entity Framework.

Apparently I wasn't associating the objects' c'tors being used in the
context of Linq. I dunno why, mayhaps just brain-fade. Thanks again.

regards
A.G.

 

[RayLopez99]

You can easily confirm this behavior for yourself by writing some
code. Use a console app and set the data context's Log property to
Console.Out, something like this :

CustomNorthwindDataContext dc = new CustomNorthwindDataContext();
dc.Log = Console.Out;
var customer = (from c in dc.Customers
                where c.CustomerId.Equals(String.Empty)
                select c).FirstOrDefault();

Setting the Log property will let you examine the underlying TSQL.
When FirstOrDefault is used against the query results, the TSQL
statement starts with 'SELECT TOP 1". When there is no TOP 1 to select
the query returns null which becomes the FirstOrDefault method's
return value.


Doing this won't accomplish much of anything.

var customer = from c in dc.Customers
                where c.CustomerId.Equals(String.Empty)
                select new { c.CustomerId, c.CompanyName};

The query above returns no results so no custom objects will be
created. The returned enumeration will be empty.

regards
A.G.

Right, thanks A.G. I did this.

RL

 

[RayLopez99]

Throwing exceptions in Web applications are expensive.

<http://blogs.msdn.com/b/ricom/archive/2003/12/19/44697.aspx>

In the example I saw of your code, throwing that exception because
something was null that you could have checked for being null without
throwing an exception is bad programming IMHO.

Good point, thanks. Actually in my particular web app I will always
be guaranteed of a hit (almost never an exception, unless the server
is down, which has nothing to do with the database) because the stuff
I am passing in are known, preexisting primary keys. But next time
I'll remember to check for 'null' in the body of the "try" block prior
to .ToList() rather than waiting for "catch" to catch any null
reference exceptions.

You might want to look at this. It's using controls that are directly
connected to ASP.NET controls using Linq-2-SQL or ADO.NET Entity
Framework. The middle tier is built for you. I built the DAL tier with
it with ease but I was not connecting up with controls, because I was
doing MVP/service layer/BLL/DAL/Model using EF just to see how it
worked, which was nice.

I could not follow it, sorry, no fault of you or the blogger (if it's
not you), no doubt it's me. Just could not make sense out of it. I'll
build my own middle tier.

<http://geekswithblogs.net/ranganh/archive/2009/12/14/using-wcf-ria-se...>

There is a MS forum for WCF RIA.




I'll take a look, maybe I can use it in my upcoming contract.

Yes, please do check it out. I also plan to. I like the idea behind
the "push" model of Rx reactive programming--akin to "lazy loading"
but with the difference you don't have to have a human user driving
the loading. That is, the way I do things now, the user clicks on a
button, and if the server has not yet returned values, you get an
error message saying "Please wait, server still working on answer".
With Rx programming, you can still have a message, but the user does
not have to click on any button, but rather the server web method
reply will just will show up. Of course you can still do this today
with a Timer that periodically checks your client side collection
housing the server data to see if .Count > 0, but it's just more
messy. One thing though: Rx apparently is still "beta" which means
(as we all know) a few (hopefully minor) bugs are probably present in
it...but I plan on checking it out later next year.

RL

 

[RayLopez99]

On 12/25/2010 4:48 PM, RayLopez99 wrote:

<snipped>

If you want to cut to the chase and figure it all out, then this is what
you need.

http://www.dofactory.com/Patterns/Patterns.aspxhttp://www.dofactory.com/Framework/Framework.aspx

Thanks. For only $79, the price of a couple of books, which I buy all
the time (I even have the Gang of 4's original book and have
implemented a few of their templates), you can get source code on
these GOF design templates.

Have you tried this software? Is it well commented? If so I might
buy it.

Also why do you use nHibernate? better than just using Visual
Studio's built in Linq-to-Sql? Looks redundant to what you already
get when you write a Linq-to-SQL command. or is it "better, faster,
less mistake-prone" Linq-to-Sql? I notice a lot of people use it and
recommend it, so it must have some advantages.

RL

 

[RayLopez99]

Yes I have the 3.5 version of the code, and I will be getting the 4.0
version for my upcoming project using VS2010. It's one of the best
investments in myself I have ever made. In its VS web development form
for the Web project, I had trouble making it work to see what the code
was doing, but I remedied that issue by making the Web projects use IIS
and virtual directory for the projects and went from there.

OK I'll keep this in mind if they have not remedied in in version 4.0.

Have you heard of this book, which is in Java, but its in C# in the
DOfactory code examples.

http://headfirstlabs.com/books/hfdp/

You may be able to find a free ebook.

I've seen this book and will add it to my collection. The GO4 book is
very dry; this book looks more good for teaching.

nHibernate has been in the Java world for a long time, ported over to
.NET, and it is more mature than Linq-2-SQL or ADO.NET Entity Framework.
It's Open-source, but it's tedious work in setting up those entities
manually

I see. I'll stick to what I have now, as I hate setting up entities
manually--it's drag-and-drop or nothing for me.

RL

 

[Arne Vajhøj]

This is more an observation than a question. I notice in the below
the 'catch' is never triggered, even if I enter a value that I know is
not present in MyTable. Thus I know for example the Table column
AField, holding strings, does not have any rows that start with "Z",
but if I input aString = "ZZZZ", I don't get var MyMatchingTables =
null, but rather I get zero hits (or .Count = 0). Must be built into
Linq-to-Sql. I've been using it for years but this is the first time
I thought about it, since I just read somewhere that keyword 'var'
does not support any nulls, or rather the right hand side source
cannot return a null.

var MyMatchingTables = from X in db.MyTables
where
X.AField.StartsWith(aString)
select X;

//why is var MyMatchingTables never null? Must be built into Linq to
Sql to never return null?

Why should it be null?

A collection with no elements is a valid collection.

And a collection with no elements seems to be a much better
description than no collection.

If you use XxxCommand ExecuteReader and does not find
any values then you also get a XxxDataReader that will
not return any rows - you do not get null.

Consistent and logical.

Arne

 

[Arne Vajhøj]

Also why do you use nHibernate? better than just using Visual
Studio's built in Linq-to-Sql? Looks redundant to what you already
get when you write a Linq-to-SQL command. or is it "better, faster,
less mistake-prone" Linq-to-Sql? I notice a lot of people use it and
recommend it, so it must have some advantages.

LINQ to SQL is a dead end.

NHibernate and MS EF are both relevant ORM's to consider.

You can use LINQ with both of them.

NHibernate is rather sophisticated.

Arne

 

[Arne Vajhøj]

nHibernate has been in the Java world for a long time, ported over to
.NET, and it is more mature than Linq-2-SQL or ADO.NET Entity Framework.
It's Open-source, but it's tedious work in setting up those entities
manually. They do have database centric nHibernate .Net providers, but
they are not free solutions that will build the model for you.

There are several free and/or open source tools that can
generate mappings and classes from database.

I don't think they are widely used.

For a couple of reasons:
1) you will most likely want to change some of the generated
stuff anyway to make it better fitting or better performing
and then you can almost just as well write it from scratch
2) the true ORM way is to generate the tables from the classes
not the classes from the tables

Arne

 

[Arne Vajhøj]

In doing a linq query with nHibernate, one must use IList and you cannot
use a List<T>, which I found strange that a List<T> could not be used.

I don't see the relevance for Peters point.

ArrayList implements IList
List<> implements IList<>

But ToList still creates the second one.

Arne

PS: Oh - and NHibernate supports both IList and IList<> - you
just pick which of them you want.

 

[RayLopez99]

Why should it be null?

A collection with no elements is a valid collection.

And a collection with no elements seems to be a much better
description than no collection.

If you use XxxCommand ExecuteReader and does not find
any values then you also get a XxxDataReader that will
not return any rows - you do not get null.

Consistent and logical.

That's we concluded, was that 'behind the scenes' the query:

will never return null. It will return, as you say, an empty
collection. This surprised me since I was not using 'new' on the
right hand side. So I would imagine X.AField.StartsWith("a string you
know does not exist in your database") will return zero, and perhaps
(have not tried this) even "X.AField.StartsWith(String.empty)" will
also return zero, and perhaps even "X.AField.StartsWith(null)". It's
built into the system.

RL

 

[RayLopez99]

LINQ to SQL is a dead end.

OK perhaps I was not using the right words. I am using the "Drag and
drop" wizard of Visual Studio 2008, when working with ADO.NET or
Silverlight, to generate database schema, that I can set up
associations with, such as foreign key, etc. Then I am querying the
database tables using Linq, rather than traditional SQL as I have in
the past. I don't know if this is "Linq to Sql" or not but it works
fine.

NHibernate and MS EF are both relevant ORM's to consider.

You can use LINQ with both of them.

NHibernate is rather sophisticated.

Whatever works for me is what I call sophisticated. As another poster
said there's no free "database schema" generator for NHiberate, and
right now the 'built in' wizard of Visual Studio 2008 seems to be
working well enough for me.

RL

 

[RayLopez99]

On 26-12-2010 05:43, Big Steel wrote: u.

There are several free and/or open source tools that can
generate mappings and classes from database.

I don't think they are widely used.

For a couple of reasons:
1) you will most likely want to change some of the generated
    stuff anyway to make it better fitting or better performing
    and then you can almost just as well write it from scratch

2) the true ORM way is to generate the tables from the classes
    not the classes from the tables

This last point #2 is interesting but does not make sense to me--since
when does the class take precedence over the table? I thought the
class lives in the middle tier--it's job is to translate from the user
interface to and from the table. Why should the table, which is a
rigid, 1970s era construction (for example they still use the 'short'
int in database tables) be generated from modern OOP classes? Seems
like you are putting a donkey cart in front of a Ferrari sports car
and then driving the Ferrari.

But maybe with the latest generation of EF this is how things have
gone--I've not looked much at EF for at least a year.

RL

 

[Arne Vajhøj]

That's we concluded, was that 'behind the scenes' the query:


will never return null. It will return, as you say, an empty
collection. This surprised me since I was not using 'new' on the
right hand side. So I would imagine X.AField.StartsWith("a string you
know does not exist in your database") will return zero, and perhaps
(have not tried this) even "X.AField.StartsWith(String.empty)" will
also return zero, and perhaps even "X.AField.StartsWith(null)". It's
built into the system.

A collection has to be created without new if there are data - it
is not so surprisingly to me that it does the same with no data.

Arne

 

[Arne Vajhøj]

OK perhaps I was not using the right words. I am using the "Drag and
drop" wizard of Visual Studio 2008, when working with ADO.NET or
Silverlight, to generate database schema, that I can set up
associations with, such as foreign key, etc. Then I am querying the
database tables using Linq, rather than traditional SQL as I have in
the past. I don't know if this is "Linq to Sql" or not but it works
fine.

"entity data model" is EF.

"LINQ to SQL classes" is guess what.

There are lots of different solutions that works. The question
is what woirk best in a given context and what is most
future safe.

Whatever works for me is what I call sophisticated.

Unusual definition.

As another poster
said there's no free "database schema" generator for NHiberate,

But there is.

and
right now the 'built in' wizard of Visual Studio 2008 seems to be
working well enough for me.

Using the built in stuff always has some advantages, because
you do not need to integrate things yourself.

And if you are spending 50 hours creating the solution, then
spending 10 hours getting the environment working is huge
overhead.

But if you are working on a 50000 hours projects, then those
10 hours of integration does not matter.

Arne

 

[Arne Vajhøj]

This last point #2 is interesting but does not make sense to me--since
when does the class take precedence over the table? I thought the
class lives in the middle tier--it's job is to translate from the user
interface to and from the table. Why should the table, which is a
rigid, 1970s era construction (for example they still use the 'short'
int in database tables) be generated from modern OOP classes? Seems
like you are putting a donkey cart in front of a Ferrari sports car
and then driving the Ferrari.

If you have an existing database that you need to access
by your app, then it is necessary to map tables->classes.
And if that database structure is not ORM friendly then it
can become quite painful.

If you don't have an existing database, then the purpose
of the database is to persist your classes. So you define
you classes as they fit the app best. And then the tables
are generated that fits best with the object and the given
ORM framework. That is a lot easier, but not always possible.

And to you your analogy: it makes more sense to define the
donkey cart after the Ferrari than the Ferrari after the donkey cart.

Arne