C# struct memory optimization?

Question

I had a job interview for probation(? I'm not sure if that's the word) and interviewer asked me to tell him what are the differences between structure and class.

So I told him everything I knew and everything that I've read on msdn.

And the guy said "not enough", I didn't have a clue. So he said:

The struct are optimized, so if there is and integer and float, which have some bites identical, then it will save this space, so struct with int=0 and float=0 is half the size of int=int.MAX, float=float.MIN.

Okay. So I was like - didn't hear about that.

But then, after the interview I was thinking about it and it doesn't really make sense to me. It would mean, that the struct size differs while we're changing the value of some variable in it. And it can't really be in the same place in memory, what if there is a collision while expanding. And we would have to write somewhere what bits we're skiping, not sure if it would give any optimization.

Also, he asked me at the begging what is the difference in struct and class in Java. To which I have answered, that there is no struct in Java and he said "Not for programmers, but Numeric types are structures" I was kind of like WTF.

Basically the question is:

Did this guy know something, which is very hard to get to know (I mean, I was looking for it on the web, can't find a thing)

or maybe he doesn't know anything about his job and tries to look cool at the job interviews.

Answer 1

The guy seems to be confused about the StructLayoutAttribute that can be applied to C# structs. It allows you to specify how the struct's memory is laid out, and you can, in fact, create a struct that has fields of varying types that all start at the same memory address. The part he seems to have missed is that you're going to only use one of those fields at a time. MSDN has more info here. Look at the example struct TestUnion toward the bottom of the page. It contains four fields, all with FieldOffset(0). If you run it, you can set an integer value to the i field and then check the d field and see that it has chnaged.

Answer 2

To me it looks like (one of you) was not talking about C# structs/classes but rather about more low-level or more general structs.

There is this special sort of memory optimization used e.g. in

1. C (unions)

and in

2. Pascal (variant records)

see e.g. article How do I translate a C union into Delphi? for an example.

Special form of this "structure" with dynamic polymorphic memory allocation is known as

3. http://en.wikipedia.org/wiki/Variant_type

and it was used heavily for inter-process data exchange in OLE automation APIs, in the pre-C# era (for decades in multitude of languages).

4. (s)he might be also talking about structure serialization format vs class in-memory format (see e.g. https://developers.google.com/protocol-buffers/docs/encoding for example of C# structure serialization)

5. you might be also talking about differences of internal JVM memory allocation (see e.g. http://blog.jamesdbloom.com/JVMInternals.html) which reminds me that you might be talking about the class file format and encoding of structs and special numeric literals vs encoding of classes (http://blog.jamesdbloom.com/JVMInternals.html#constant_pool)

So after 5 guesses I believe there is something lost in your translation of your talk with the interviewer and (s)he probably browsed into an area you claimed you know and it turned out you don't. It also might be that (s)he started talking bullshit and checked your reactions. Lying about your skills on the resume is not recommended in any job (e.g. http://www.softwaretestinghelp.com/5-common-interview-mistakes/). I'd vote for the interviewer knew the inteviewing job well enough