NHProfThe stack is not as simple as you wish it to be

time to read 3 min | 509 words

One of the nicest features that NH Prof has to offer is this, allowing you to go from the query issued to the database directly to the line of code that caused this query to be generated. Icon

image 

A few days ago I posted that you can either build something in one day, or in three months, but nothing much in the middle. The proof of concept that convinced me that I can build NH Prof was written during a single evening, along with two pints of Guinness. The overall concept that I have now is drastically different, but it is that evening that made sure that I would start it.

However, I wanted to use this opportunity to talk about some of the things that take a one day project and stretch it into three months. In this case, the query to line of code is a good feature, because it looks simple. After all, the only thing that I need to do, basically, is just to use the built in facility to get the stack trace, and extrapolate from there. Well, even for the normal case, that is not so easy. There are issues of user experience that play a deciding role in how we are going to deal with this.

But this isn't what was really challenging. The challenging part came from the law of least surprises. That is, if I am providing a stack trace for queries, I should provide them for all queries. As it turned out, there was one scenario in which I am not able to give a complete stack trace. When we are mixing queries generated as a result of a lazy proxy being instantiated. On the whole, this is a minor concern, but it really bugged me. You would get a stack trace that just ended in some weird type that you never seen before.

At that point, I felt like I had to solve this.

Let us go back and outline the problem. Stack traces relies on PDBs to do their work. NHibernate's lazy proxies are generated at runtime, have no PDBs and mask the actual call to the method / property that we are actually are interested in. The mix of the two wasn't encouraging, I must say.

Given that, I decided to pull out the big guns, and started messing with the PDB myself. Obviously, I don't want to do it on my own, that is why Cecil is so useful. But it was still a bit tricky to get it right. And the funny part is that I am pretty certain that this is one feature that absolutely no one will actually notice. If it wasn't there, however, people would notice.

Anyway, it 3:30 AM here, and for some reason I managed to do more work tonight than in the previous week(!). NH Prof is now in private beta, and I hope to be able to take it to public beta in a few weeks.

More posts in "NHProf" series:

  1. (02 Dec 2008) What is the role of the DBA?
  2. (01 Dec 2008) The stack is not as simple as you wish it to be
  3. (21 Oct 2008) Logging interception
  4. (13 Oct 2008) Another milestone
  5. (13 Oct 2008) Alive! It is alive!