rxx Discussions Rss Feed

New Post: Subject doesn't output events when it's reading an ObservableTcpListener

rolpereira — Wed, 17 Apr 2013 16:06:02 GMT

davedev wrote:

Hello davedev,

Try this instead:

receiveContentObservable.Subscribe(subject.OnNext, subject.OnError);

Yup, that worked. Thank you!

Note that sometimes LINQPad doesn't correctly close the socket [snip]

You can avoid orphaned sockets by disposing of the subscription yourself.
using (tcpListenerPool.Subscribe(
    (TcpClient tcpClient) =>
    {
        // ...
    }))
{
    Console.WriteLine("waiting");
    Console.Read();
}

Weird I thought I had tried that and it had failed. Thanks again.

New Post: Subject doesn't output events when it's reading an ObservableTcpListener

davedev — Wed, 17 Apr 2013 15:29:17 GMT

Hi,

Note that sometimes LINQPad doesn't correctly close the socket [snip]

You can avoid orphaned sockets by disposing of the subscription yourself.

using (tcpListenerPool.Subscribe(
    (TcpClient tcpClient) =>
    {
        // ...
    }))
{
    Console.WriteLine("waiting");
    Console.Read();
}

- Dave

New Post: Subject doesn't output events when it's reading an ObservableTcpListener

davedev — Wed, 17 Apr 2013 15:25:36 GMT

Hi,

It's probably because the subject is receiving OnCompleted and ignoring subsequent notifications.

In other words, this:

receiveContentObservable.Subscribe(subject);

is essentially the same as this:

receiveContentObservable.Subscribe(subject.OnNext, subject.OnError, subject.OnCompleted);

Try this instead:

receiveContentObservable.Subscribe(subject.OnNext, subject.OnError);

It's a fairly common mistake that I make too :)

- Dave

New Post: Subject doesn't output events when it's reading an ObservableTcpListener

rolpereira — Wed, 17 Apr 2013 14:04:34 GMT

Hello all,

I'm trying to pass the output of ObservableTcpListener to a Subject<byte[]>. That subject will then be used in a query.

The problem is that while the ObservableTcpListener receives every socket contents that I send, the query doesn't output anything besides the first two or three events.

This code shows the problem that I'm mentioning. It should be ran in LINQPad as a "C# Program".

void Main()
{
    // Create the subject and the query
    var subject = new Subject<byte[]>();
    var query = subject.Select(reading => reading);
    
    query.Subscribe(Console.WriteLine);
    
    var tcpListenerPool = ObservableTcpListener.Start(IPAddress.Loopback, 20001);
    
    tcpListenerPool.Subscribe(
        (TcpClient tcpClient) =>
        {
            // Debug information just to check if the socket actually has anything in it
            Console.WriteLine("tcpSocket.Avaliable: " + tcpClient.Available);
            
            var receiveContentObservable = tcpClient.Client.ReceiveUntilCompleted(SocketFlags.None);
            
            // Send socket content to the subject
            receiveContentObservable.Subscribe(subject);
        });
        
    Console.WriteLine("waiting");
    Console.Read();
}

// Define other methods and classes here

Note that sometimes LINQPad doesn't correctly close the socket, and gives the error: "Only one usage of each socket address (protocol/network address/port) is normally permitted". If that happens go to the "Query" menu and choose "Cancel All Threads and Reset".

The program that I'm using to send the sockets to the ObservableTcpListener is this python program:

import socket
import random

payload = ["foo", "bar"];

def sendPayload(payload):
    s = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
    s.connect(("localhost", 20001))
    
    s.send(payload)
    
    s.close()

# Send 10 sockets with either "foo" or "bar"    
map(lambda x: sendPayload(random.choice(payload)), range(10))

Is this a bug, or am I doing something wrong?

Edit: I'm using the "Rxx Class Library for .NET 4.0" available in the Downloads section.

New Post: IObservable> conversion

grantchik — Wed, 17 Apr 2013 12:27:08 GMT

Good stuff, worked, thanks

New Post: IObservable> conversion

davedev — Wed, 17 Apr 2013 12:10:12 GMT

Hi,

You can use this overload of Select. It allows you to project both sides of Either.

For example:

IObservable<Either<T, Exception>> xs = GetObservableFromExternalSource();

IObservable<Either<T, MyAppException>> ys = xs.Select(t => t, ex => new MyAppException(ex));

See also:
https://rxx.codeplex.com/discussions/401272

- Dave

New Post: IObservable> conversion

grantchik — Wed, 17 Apr 2013 11:15:04 GMT

so i have a subscription that returns IObservable<Either<T, Exception>> from an external source, and i would like to convert that into IObservable<Either<T, MyAppException>>, where MyAppException takes Exception as a parameter in ctor.

can you please advise a way of doing this? thanks

New Post: Rxx release to match Rx v2 due to old dependency

davedev — Mon, 11 Mar 2013 10:17:11 GMT

Hi,

Yep, FromPropertyChangedPattern is on the list.

I agree that portability is the way to go. Rx now supports Windows Phone 8 in their portable library, so Rxx should too.

- Dave

New Post: Rxx release to match Rx v2 due to old dependency

haspemulator — Mon, 11 Mar 2013 09:32:35 GMT

davedev wrote:

Furthermore, I've been working on integrating Rxx into Rx, though admittedly it hasn't been one of my priorities recently. I'll dive back into it soon and hopefully we'll see some of the APIs rolled into Rx vNext. After that, I may continue to deploy Rxx with features that weren't absorbed into Rx.

Integration of Rxx into Rx sounds absolutely great! My personal favorite would be the FromPropertyChangedPattern. Another really cool thing could be implementation of certain pieces in a portable way, so that it's possible to reuse the binaries produced with those extensions between Windows Phone 8 and Windows 8. But I guess it's too forward-looking request. :)

Thanks,
Roman.

New Post: Rxx release to match Rx v2 due to old dependency

davedev — Fri, 08 Mar 2013 17:14:06 GMT

Hi Roman,

I'm having a bit of a problem with a Code Contracts bug at the moment that is preventing me from building a release package; however, that's not really the reason why I haven't deployed a new version to NuGet in a while. At some point I decided that the cost of building, testing, documenting and deploying new release versions of Rxx to match new versions of Rx, for each of the supported platforms, wasn't worth the effort. I figured that building locally was the best solution for everyone, partially because it allows people to shrink the size of the library by eliminating the features that they don't want. But thanks for your feedback, because I realize that it's probably not the best solution for everyone, it's just me being lazy :)

Furthermore, I've been working on integrating Rxx into Rx, though admittedly it hasn't been one of my priorities recently. I'll dive back into it soon and hopefully we'll see some of the APIs rolled into Rx vNext. After that, I may continue to deploy Rxx with features that weren't absorbed into Rx.

- Dave

New Post: Rxx release to match Rx v2 due to old dependency

haspemulator — Fri, 08 Mar 2013 14:31:20 GMT

Hi,

Any updates for the status of Rx2-matching Rxx release recently? Really looking forward for it.

P.S. Building locally is not an attractive option, because of our build system and how it works (without going much into details). I would really prefer having Nuget package for this.

Thanks in advance!

-Roman.

New Post: BufferIntrospective delayed subscription to underlying source

bman654 — Sat, 02 Feb 2013 18:07:56 GMT

Thanks Dave! I had been meaning to create the codeplex discussion ever since you posted but kept forgetting also :)

New Post: BufferIntrospective delayed subscription to underlying source

davedev — Sat, 02 Feb 2013 17:50:53 GMT

This discussion has been copied to a work item. Click here to go to the work item and continue the discussion.

New Post: BufferIntrospective delayed subscription to underlying source

davedev — Sat, 02 Feb 2013 17:48:38 GMT

Hi Brandon,

I had bookmarked the discussion but then I forgot to get back to it, sorry.

I'm currently working on merging various Rxx operators into Rx. The introspective family may end up being absorbed too. If so, I'm probably going to end up rethinking and rewriting them, including __BufferIntrospective__. I'll probably drop WindowIntrospective entirely.

Thanks for reporting this issue. When I get to these operators, I should make a unit test out of it.

-- Dave

New Post: BufferIntrospective delayed subscription to underlying source

bman654 — Sat, 02 Feb 2013 16:39:31 GMT

Hi Dave,

Did you ever have time to repro this issue: http://social.msdn.microsoft.com/Forums/vi-VN/rx/thread/89f3d9c7-654f-41da-96e5-9bf9f5f97f2e

Thanks.
Brandon

New Post: TFS??

davedev — Thu, 31 Jan 2013 06:32:46 GMT

Looks like GIT may become a viable option for me through TFS:

http://blogs.msdn.com/b/codeplex/archive/2013/01/30/using-visual-studio-git-with-codeplex.aspx

New Post: TFS??

davedev — Thu, 24 Jan 2013 18:31:04 GMT

Hi,

Fair question.

I choose TFS for all of my open source projects because:

It meets all of my basic needs when developing open source software on CodePlex:
1. Get Latest (Only really useful where team > 1)
2. Check In
3. Check Out
4. View History
5. Compare
6. Resolve Conflicts (Only really useful where team > 1)
It's very simple to use.
It integrates seamlessly into Visual Studio.
It requires no command-line knowledge; i.e., it's entirely UI-driven. However, it also offers a command-line for advanced scenarios (which I never need).
It's the original source control system used by CodePlex and many of my previous employers, so I have more experience with it than any other.

I had used SVN at a previous employer and found it to be difficult and unpleasant to use, especially with Visual Studio. The developers in our company that were using Visual Studio 100% of the time, including myself, agreed that SVN was not the best choice. We wanted as much transparency in our source control system as possible, which meant avoiding the command-line and shell extensions, instead opting for simple context menu commands in Visual Studio. We also wanted to avoid having annoying artifacts in the file system, such as .svn or .vss folders. We had tried installing SVN plug-ins for Visual Studio, but IIRC they had bugs and none of them were nearly as seamless or as easy to use in general as TFS. The company eventually transitioned to using TFS source control and task management for all their projects, not just .NET projects. I was primarily responsible for maintaining source control along with one other person. We decided to install SVN Bridge so that our Java developers and Mac users (designers) could use the same source control. They found it to work well with their Java/Mac dev tools, so TFS was a big unifying win for the company.

More recently, I've taken a look at Hg and GIT so that I can contribute to other projects here on CodePlex. It seems Hg and GIT are even more complicated than SVN, though they also have some similarities; e.g., they're both command-line driven and they seem to push shell integration as their primary GUI. I'll never understand the appeal of shell integration - it's not like my computer is only useful for programming. I.e., I don't see my entire file system as a bloated source control client :)

Furthermore, neither appear to offer simple commands to accomplish common tasks like those listed above. For example, when researching GIT, I asked How to Get Latest From Source Control? and What Commands are Supported in the GUI? The answers were surprising to me. In response to the latter question:

For your example "Get Latest" I would suggest using rebasing. Which is quite simple to use.

[snip - 8 step procedure]

So you have the latest and your changes combined. This is just a typical workflow I use and suggest. There might be different point of views out there :)

Quite simple to use? 8 steps? Typical workflow? Different points of view? All to simply Get Latest, which TFS offers as a discrete command? Sorry, but GIT is certainly not going to be my first choice. I still haven't quite figured it out yet :)

I don't know much about Hg, but I suspect that it's similar to GIT. At the very least, given that it seems to rely on Python scripts and a command-line interface, I highly doubt that it gives the seamless experience in Visual Studio that I get with TFS, even if I used VisualHg. (Note that I've already tried, briefly.)

So to answer your question:

TFS?? Why?

Even if I could get the same seamless experience using a source control system other than TFS, perhaps a better question is:

Why not?

Full Disclosure: I'm not entirely a TFS enthusiast. I've found that it has problems too, some of which were very frustrating. (Though maybe they've been resolved in the latest version - I don't know yet.)

For me, the perfect source control system would be one that never prompted me for anything and offered no context menu arguments at all, perhaps with the exception that sometimes I like to view history and comparisons. Mostly it would be as if it wasn't even there. 99% transparency. Perhaps that's an entirely naïve and idealistic perspective, especially when considering the need for conflict resolution (or is there really a need for it?), but that's how I feel. Remember that I mostly work as an individual, so source control to me probably has a different meaning than to those people that mostly work on teams.

- Dave

New Post: TFS??

damianh — Thu, 24 Jan 2013 09:57:06 GMT

Why? (seriously)

New Post: Usage of Either

DanHarman — Wed, 28 Nov 2012 09:09:35 GMT

Dave

Ahah I see your point, I hadn't considered the rammifications of closures intersecting with subscriptions. Thanks a lot for that catch, it may explain some of the problems I'm seeing at the moment. Also thanks for that link to the rx programming best practise you wrote and the MS one as well. Hadn't seen them before and they are very useful.

Dan

New Post: Usage of Either

davedev — Tue, 27 Nov 2012 20:35:19 GMT

Hi Dan,

Ah, I see. But my point was that the local is being shared across multiple subscriptions, which it seems wasn't intended. Consider compromising by wrapping the entire query, including the local variable, in a call to Observable.Defer. That way, every call to Map and every call to Subscribe get fresh locals, though multiple notifications for a given subscription share the same local. Defer can also serve as documentation that the closure isn't a mistake.

- Dave