Click here to Skip to main content

Article

Breaking Changes in Argument List Evaluation in C# 5.0

Paulo Morgado

Rate me:

4.86/5 (13 votes)

30 Jul 2012CPOL

35.4K

11

25

In C# 4.0 there was an error in the C# compiler in the order of the evaluation of the arguments in argument lists.

The C# Language Specification states on §7.5.1.2 that “(…) the expressions or variable references of an argument list are evaluated in order, from left to right (…)”. free hit counters

So, when this code is compiled with the C# 4.0 compiler:

C#

static void M(
    int x = 10,
    int y = 20,
    int z = 30)
{
    Console.WriteLine(
        "x={0}, y={1}, z={2}", x, y, z);
}

static void Main(string[] args)
{
    int a = 0;

    M(++a, z: ++a);
}

and run, this unexpected output is obtained:

C#

x=2, y=20, z=1

In fact, fixing this compiler flaw was the cause of one of the few breaking changes introduced in C# 5.0.

Using the 5.0 compiler, the expected result is obtained:

C#

x=1, y=20, z=2

To avoid this type of surprises, expression evaluation should be avoided in argument lists.

With this code:

C#

int a = 0;

int i = ++a;
int j = ++a;

M(i, z: j);

the same result is obtained for both C# 4.0 and C# 5.0:

C#

x=1, y=20, z=2

License

This article, along with any associated source code and files, is licensed under The Code Project Open License (CPOL)

Written By

Paulo Morgado

Software Developer (Senior) Paulo Morgado

Portugal

About Paulo Morgado

Comments and Discussions

eferreyra30-Jul-12 3:05

30-Jul-12 3:05

A MSIL detail would have been nice.

Excellent ARTICLE

Tikoko26-Jul-12 8:04

26-Jul-12 8:04

Thanks!

Re: Excellent ARTICLE

Paulo Morgado26-Jul-12 8:24

26-Jul-12 8:24

THANK YOU!!!

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

johannesnestler26-Jul-12 6:05

johannesnestler

26-Jul-12 6:05

Good to know, thank you - maybe not an "article" though... but you got my 5

Re: My vote of 5

Paulo Morgado26-Jul-12 8:23

26-Jul-12 8:23

THANK YOU!!!

It is a blog post, not an article.

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Not an article...

Dave Kreskowiak26-Jul-12 3:38

Dave Kreskowiak

26-Jul-12 3:38

This is a tip/trick at best.

Frankly, I've never seen anyone use expressions as parameters, and if I did, I would discourage the practice entirely because it makes debugging a bit of a pain in the ass and makes the code more difficult to maintain.

A guide to posting questions on CodeProject[^]

Dave Kreskowiak

Re: Not an article...

Paulo Morgado26-Jul-12 4:04

26-Jul-12 4:04

Dave,

I'm sure you're aware that you don't know every programmer out there and haven't seen every C# code ever written, right?

My recomendation is also to not use it, as said in my blog post.

But the language specification allows it and the C# 4.0 compiler generates the wrong code. The C# 5.0 compiler will not have this flaw.

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Richard MacCutchan25-Jul-12 23:09

Richard MacCutchan

25-Jul-12 23:09

This is a well known feature of the increment operators and is discussed at length in many articles. ~~You may consider posting this as a tip but it is far too short to be an article.~~

[edit]
It appears this got redesignated while I was posting my comment.
[/edit]

One of these days I'm going to think of a really clever signature.

Re: Not an article

Paulo Morgado26-Jul-12 0:34

26-Jul-12 0:34

This has nothing to do with increment operators.

You can replace the increment operator with:

C#

static void M(
    int x = 10,
    int y = 20,
    int z = 30)
{
    Console.WriteLine(
        "x={0}, y={1}, z={2}", x, y, z);
}

static int i;
static int next()
{
    i++;
    return i;
}

static void Main(string[] args)
{
    int a = 0;

    M(next(a), z: next(a));
}

The output will still be

x=2, y=20, z=1

with the C# 4.0 compiler and

x=1, y=20, z=2

with the C# 5.0 compiler.

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Not an article

Richard MacCutchan26-Jul-12 1:17

Richard MacCutchan

26-Jul-12 1:17

Paulo Morgado wrote:
This has nothing to do with increment operators.

It has everything to do with increment operators, and is a well known issue. Try some research on it and you will find some good descriptions of why this occurs.

One of these days I'm going to think of a really clever signature.

Re: Not an article

Paulo Morgado26-Jul-12 3:46

26-Jul-12 3:46

No it doesn't. And I have no recolection of any problem between the language specification and the code generated by the compiler regarding the increment operator.

There is, however, a known problem with the order of evaluation of arguments that will be solved in C# 5.0 like this[^] MSDN article referenced in the blog post. The MSDN article clearly states that "Side effects from named and positional arguments in a method call now occur in the correct order".

But I'm more than willing to explain to you why this is not an increment operator problem if your care to explain what your doubts are.

Or you could try to prove that the C# 4.0 compiler is working just fine regarding argument list evaluation.

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Not an article

Richard MacCutchan26-Jul-12 5:18

Richard MacCutchan

26-Jul-12 5:18

Paulo Morgado wrote:
And I have no recolection of any problem

It still exists; using multiple increment/decrements on the same variable within a single expression, as in a function/method call, does not guarantee the correct order of evalution. This is because compilers, and their writers, are allowed freedom to choose the order they do it. The fact that C# 5 is changing this is a recognition that the problem exists.

One of these days I'm going to think of a really clever signature.

Re: Not an article

Paulo Morgado26-Jul-12 5:27

26-Jul-12 5:27

Once again, this has nothing to do with increment/decrement operators and the C# 5.0 compiler is changing nothing whatsoever about increment/decrement operators.

Can you give, at least one, example where the correct order of evaluation of increment/decrement operators in C# is not according to the language specification?

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Not an article

Richard MacCutchan26-Jul-12 5:56

Richard MacCutchan

26-Jul-12 5:56

Offhand, no, but I have seen enough discussions of this issue to know that it exists.

One of these days I'm going to think of a really clever signature.

Re: Not an article

Paulo Morgado26-Jul-12 6:14

26-Jul-12 6:14

I'de like to see just one.

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Not an article

Matt T Heffron26-Jul-12 8:29

26-Jul-12 8:29

Richard, I'm going to have to agree with Paulo here. Apparently, this problem is due to the way that the named parameter is being handled. It seems that it is being evaluated before the positional parameters. To test this, I changed the invocation of M to be:
M(++a, 20, ++a);
Making the y parameter explicit, and providing the z parameter as a positional parameter.
The result is the correct:
x=1, y=20, z=2

Richard MacCutchan wrote:
This is because compilers, and their writers, are allowed freedom to choose the
order they do it.

The C# Language Specification that Paulo cited, specifically denies compiler writers the freedom to choose the order that the individual expressions of the argument list are evaluated. The error in the C# 4.0 compiler appears to be what I stated above.

Re: Not an article

Paulo Morgado26-Jul-12 8:50

26-Jul-12 8:50

THANK YOU!!!

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Requires more explanation and clarity

PIEBALDconsult25-Jul-12 7:10

25-Jul-12 7:10

Please clearly explain what the problem is.

Re: Requires more explanation and clarity

Paulo Morgado25-Jul-12 13:41

25-Jul-12 13:41

You're kidding, right?

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Requires more explanation and clarity

PIEBALDconsult26-Jul-12 3:21

26-Jul-12 3:21

No. You seem to assume that everyone knows about the problem. I've never heard of it. I don't use the new features of C# 4.

Re: Requires more explanation and clarity

Paulo Morgado26-Jul-12 3:34

26-Jul-12 3:34

I didn't new that myself becaus I don't usually use this specific feature in this specific way.

This blog post shows what happens in C# 4.0 and won't happen if C# 5.0.

It assumes that the reader knows (of) the feature (named arguments), but not the problem. If you want to know about named arguments, see it here: C# 4.0: Named And Optional Arguments[^]

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Requires more explanation and clarity

PIEBALDconsult26-Jul-12 3:57

26-Jul-12 3:57

Explain it, don't just include a link. Mad | :mad:

Mad | :mad:

Re: Requires more explanation and clarity

Paulo Morgado26-Jul-12 4:06

26-Jul-12 4:06

I already did. The link is for one of my blog posts. Do you want me to repost it here?

Cheers,

Paulo Morgado

Portugal - Europe's West Coast

Re: Requires more explanation and clarity

PIEBALDconsult26-Jul-12 5:23

26-Jul-12 5:23

Yes.

General News Suggestion Question Bug Answer Joke Praise Rant Admin

Use Ctrl+Left/Right to switch messages, Ctrl+Up/Down to switch threads, Ctrl+Shift+Left/Right to switch pages.

Go to top