As I said in previous post I had to drop AppDomains from my solution to allow replacing assembly while the FbNetExternalEngine is running. Calls across AppDomains are simply too slow (read: I can’t make it fast enough. 😉) for this kind of project. Why I even needed AppDomains? The specs are simple: Allow assembly replacing (rewrite) while it’s loaded in Firebird via my FbNetExternalEngine. Can it be solved without AppDomains?
Over the last few days I was working on performance of FbNetExternalEngine. Basically doing all I can to make it execute the code quickly as possible. OK, maybe not all I can, but something that gives biggest gain while being able to program it in a reasonable timeframe, because there’s always room for more.
I completely missed some handy methods on Entity Framework Core’s
DbContext. Maybe you did too. And that would be a pity.
When I was writing the original post I knew the code was right on the edge. It worked on 2.5 and that mattered at that time. My hope was that it will eventually work forever, because why would you change such thing… Nope.
While writing previous post I realized the deconstruction to tuple can be added to any type. As far as compiler is concerned the
Deconstruct method needs to exist, doesn’t matter whether it’s an extension method or instance one. That means I can write a
Tail-like methods in C# with nice syntax. I also dusted off my Haskell knowledge and wrote the same logic there for comparison. And then I did it in Python and F# as well. Brace yourself for a true polyglot post.
As I’m playing more and more with tuple in C# 7 to find out where this is useful and where it falls short and actually bites back later on. And during this I kind of by blind luck discovered more succinct syntax for deconstruction. I reminds me Python, which I like.
Did you know you can store
NaN value in Firebird? You didn’t? Me neither, until last week when Slavomír Skopalík showed me whole new world. It was in Delphi, so I was curious how it will turn out from .NET/FirebirdClient.
O/R mappery jsou výborný sluha, ale špatný pán. Abych tak trochu parafrázoval známé přísloví. David Gešvindr dobře zachytil, že v rámci různých Entity Framework přednášek apod. trousím, často i jako odpovědi na otázky z davu, různá doporcučení co nedělat (nebo co naopak dělat). Drobečky jsou fajn, ale dát to trochu do kupy a celkově sjednotit není od věci. A o tom je tato přednáška.
I realized I have some pieces of code to show some specific behavior, mostly around threading and synchronization, all over my notes. Some of these pieces may be 10+ years old. I use these pieces during my “threading/parallel/async” course, but why not to share it publicly. Maybe I’ll stumble on it after some years, maybe .NET will be history, and it will be interesting to re-read and re-think the code. The code isn’t unique or something where I’m the first to realize it. It’s really just an example code.
Today I have a small example that shows how optimizations (compiler, JIT, ...) can make multithreaded code behave differently than expected. I don’t remember where I’ve seen the code. Some book or blog post from - I guess - Jeffrey Richter. Or maybe Joe Duffy or Stephen Toub or ... These guys are very smart.
Because I can load any .NET code in FbNetExternalEngine I had and idea, probably since November last year, to try to load Firebird Embedded inside it all that in server (I know completely useless). So I did it.
I’m playing a little with Windows Subsystem for Linux (also known as WSL or LXSS), mostly to see what’s what and how it might be useful for my everyday usage. As I was reading some articles on how it’s implemented and how it works inside I’ve got an idea. What if I create an application, in .NET Core, that uses some resource from kernel and I execute it both regular Windows session and in LXSS, will this resource be shared across?
Today, as I’m teaching my parallel/threading/async course, I was showing how the tearing on reads happens. I used the same code I published couple of months ago. And then idea came (yes, it happens sometimes 😉). I wanted to show that the
decimal, being 128-bit structure, splits on underlying values. Here’s the code.
It’s been a while since the last time I blogged about FbNetExternalEngine. That doesn’t mean there was nothing happening. I was slowly working on it (really slowly, because of ... time), plus some companies took interest in it, confirming it’s worth working on it and providing some feedback or nice-to-have features wishes (feel free to share your ideas in comments). So what’s really new?
New version 126.96.36.199 of ADO.NET provider for Firebird is ready for download. Only one small new feature was added - support for
[CREATE|ALTER|DROP] FUNCTION in
FirebirdSql.Data.Isql for Firebird 3.
I don’t like comments. Most of the time information in comments is wrong and obsolete. I believe the code should be clear as a fresh snow. Most of the time. Some design decisions or measurements can be commented, but not the code itself. If it needs commenting, it should be rewritten.
One cardinal sin I can’t live with is commenting out code, just because you might need it later or because it’s being “refactored”. Nonsense. Sadly, this is done so often in a team where I’m responsible for code quality. Lucky me.
Thus, to prevent it I decided to test for big chunks of comments in the codebase using a code. More specifically NDepend or Roslyn. Whatever will do the job.
I don’t know why, but last week I decided to check whether the 2 seconds timeout for single finalizer and 40 seconds for all finalizers at the end of .NET application are still in place. If you haven’t heard about these two numbers ever, don’t worry. Because if you’ve ever hit one or the other you’ve been doing something seriously wrong. That’s how I’ve got to learn about it as well. 😉
The implementation of events via the
FbRemoteEvent class was nor wrong nor correct. It was correct in a way that it did what it should do, but it was not clean and it was not guiding the developer into the pit of success, you had to know how it works under the covers to work with it properly. I decided to rewrite the implementation, fix some issues and more importantly give it more high-level feel, where you’re actually using some infrastructure, instead of just close the raw implementation.
New version 188.8.131.52 of ADO.NET provider for Firebird is ready for download. Except for few bug fixes, the main focus was on re-implementing events.
I write my posts in Word, mostly. Because Word, with its superb proofing tools, it single handedly the best tool for that. But I also use Markdown to store the posts and I publish from it, because Markdown is just a plain text. The finishing touches like code pasting and YAML editing I do in PSPad. Again, because it’s the editor I like and used since forever. However, I use typographer’s quotes in Word. Why wouldn’t it? And although I can leave these in the final text, I prefer to use straight quotes and let these be processed in Markdown to HTML transformation.
I have a small lifehack that I’d like to share with you. The fact of today is that you can’t life without an email. OK, maybe you can, but in my world, it would be pretty damn difficult and inconvenient. And so I have email always with me, on my phone. But what if you don’t want to be email-able? Like on vacation?