October 10, 2014

Reactive Reactive! Read all about it!

Like the late 1800's American newspaper boy used to shout, in this blog post we'll be taking a look at something "extra" in Rx.

Reactive Reactive

Nope, it's not a typo.

It’s like acceleration to speed, or C# to IL, or higher-order functions to functions.

Sometimes fast isn't fast enough.  Sometimes abstract isn't abstract enough.  Sometimes functions aren't... functiony enough?

Sometimes Rx operators aren’t reactive enough.

Many operators in Rx force us to supply static arguments, resulting in an operator's static behavior.  By static I mean upfront, immutable, unvarying values; e.g., typically scalar primitives like Int32, TimeSpan and DateTimeOffset.  Furthermore, stateful operators with static parameters prevent us from varying behavior without a potential loss of data.

Here's a simple example of a stateful operator with a scalar parameter.  Let's say that we have an observable and we want to skip the first 5 notifications:

var xs = observable.Skip(5);

That's great, but what if we don't know how many notifications we want to skip yet?  Or what if we want to change that value later while the observable is active?

Defer doesn't work if we want to subscribe now.

We could attempt to change the value, without causing side effects, by writing some complicated dance within the Publish operator involving subscriptions, disposing, recreating the query and resubscribing; however, we'll lose the state within the operator that tells us how many values have already been skipped.  We could keep track of the number that have already been skipped ourselves, but then what's the point of the Skip operator?

Ok, easy solution, we define a new overload:

public static IObservable<T> Skip<T>(this IObservable<T> source, IObservable<int> count);

Now we can change the count at any time without resubscribing and without losing any state.  We'll simply push in a new value through count whenever we want to change it.

Subscribe now, decide later.

Maybe we'll never push in a value and so everything will be skipped.  Maybe we'll push in a value of 10 before the first 10 notifications have been skipped, thus only the first 10 notifications will be skipped.  Maybe we change our minds and push in a value of 5, but it's too late – we've already skipped the first 7 notifications, thus only the first 7 notifications are skipped.  At this point, the operator unsubscribes from count.1

But wait, are the values in count even necessary anymore?  It seems overly complicated.  Can't we just generalize our solution in terms of time?

public static IObservable<T> SkipUntil<T, U>(this IObservable<T> source, IObservable<U> other);

In other words, instead of an observable count, we'll simply use an observableT is ignored.  When the observable signals or terminates, the operator stops skipping.

This "reactive reactive" pattern is already in use by several operators in Rx, such as SkipUntil and TakeUntil as of Rx 1.0, and some overloads of Window, Buffer, Throttle, Sample, Delay and Timeout as of Rx 2.0, but upon further inspection it seems that Rx is potentially missing a lot of similar opportunities.

Reactive Overloads for Static Parameters
https://github.com/Reactive-Extensions/Rx.NET/issues/18

Merge – A Case Study

In one such example, we can see how data loss is a critical problem.  The Merge(maxConcurrency) operator is very powerful, but its behavior is static.  While inner observables are pushed in, Merge subscribes to them until maximum concurrency is reached, at which point any observables received are enqueued for later.  Often these observables capture non-recoverable application state.  To change the maximum level of concurrency in response to the changing capabilities of the environment requires disposing of the subscription to Merge's observable, reforming the query with a new static argument and resubscribing; however, in the process any observables enqueued by Merge are lost, and their captured application state is lost along with them.

Merge with Dynamic Maximum Concurrency
https://github.com/Reactive-Extensions/Rx.NET/issues/48
https://social.msdn.microsoft.com/Forums/en-US/62743ffd-befd-474a-8f0a-19dbaec7a926/design-question-throttling-hits-on-external-systems

Conclusion

By providing a full suite of reactive/reactive overloads for Rx operators, developers would be enabled to write queries that adjust the behavior of operators dynamically based upon dynamic application state, user input and/or heuristics about the runtime environment to improve the performance of their applications, without losing data and without the costs of having to repeatedly solve similar problems themselves.

 

  1. One could also imagine the next logical scenario in which we raise the value back to 10, thus perhaps the first 7 notifications and the 9th and 10th notifications are skipped, but the 8th notification makes it through.  Although these semantics are interesting, they're not the semantics of the Skip operator.  Skip is all about skipping consecutive notifications from the beginning of the sequence.  I'm not saying that these semantics aren't useful, it just seems more like the behavior of the Sample operator to me and surely a similar analysis applies to that operator.

Tags:

Rx

September 28, 2014

I'm On Twitter

Just thought I'd mention that I'm attempting to be a bit less antisocial.

Follow me on Twitter @IDaveSexton

Recently, I noticed that I've been getting traffic to my website from Twitter and so I've finally decided to see what it's all about. I have to say that I'm actually enjoying the experience of subscribing to interesting people and reading my customized stream each day.

Still, though, I'd rather be coding.

September 13, 2014

Reactive Extensions: The Power of T

The T in IObservable<T> is more powerful than you may realize.

Structural T

T carries data; its most obvious purpose.

T can be any type, such as Int32, String or MyAmazingObject.

T can be void, as in System.Unit.

T can be anonymous, as in:  new { Value = "T", Awesome = true }

T can be a choice, as in Either<TLeft, TRight>, which also seems to fix so-called "glitches".

T can be a group of T, as in IObservable<IGroupedObservable<K,T>>; e.g., GroupBy.

T might be there or it might not, as in Maybe<T>.

Meta T

T carries notification.  Observables notify observers by pushing T.

can be notification reified, as in Notification<T>.

T can be collection change notifications, as in CollectionNotification<T>.

T can be directed mutations, as in CollectionModification<T>.

Temporal T

T carries time.  Observables push T asynchronously (though sequentially-§4.2).  Even pushing synchronously takes time though.

T can be a point in time; implicitly, the time at which it's observed, as in Amb or CombineLatest.

can be time reified, as in Timestamped<T>, TimeInterval<T>, Recorded<T> or simply DateTimeOffset.

Temporal Window T

T carries duration.  A window is a duration of time that has a start and maybe an end.

T can start a window, as in Join, GroupJoin, Window or Buffer.

T can end a window, as in Join, GroupJoin, Window(closingSelector), Buffer(closingSelector), etc.

T can be a window, as in IObservable<IObservable<T>>.

T can be the end of time, as in that which precedes it, interpreted relative to the time of subscription (e.g., Timer), relative to a previous T (e.g., Interval or Throttle), or relative to some out-of-band time (e.g., Sample).

Channel T

T can be a channel or callbackObservers invoke T to affect an observable.

T can be a delegate, as in Action or Func<TResult>.

T can be a subject, as in IObservable<ISubject<T>>.  This establishes a full duplex, asynchronous communication channel.  The observable is a singleton and both ends subscribe to the subject; e.g., a chat application. (See also this blog post.)

Future T

T can be the promise of a future T, as in IObservable<Task<T>>; although, it's somewhat redundant since IObservable<T> is also a promise of a future T and it has a wider cardinality (0,∞).  Task's cardinality is strictly (0,1).

Conclusion

T carries data, notificationtime and duration.  It can also have communication channel or promise semantics.  Best of all, T carries any additional semantics that you choose.

So the next time that you find yourself stuck trying to figure out which Rx operator is the right one to solve your particular problem, pause and give T the attention that it deserves.  You may just find that the answer to your problem lies within how you make use of T.

So, have I missed any categorically interesting uses for T?  How do you use T?

Tags:

Patterns | Rx | Rxx

July 16, 2014

NuGet Package for Sandcastle

Did you ever want to build Visual Studio-quality reference documentation from your XML code comments by simply referencing a NuGet package?
Well now you can!

The latest release of White Tie includes full support for generating documentation using Sandcastle. No configuration necessary. (You must install Sandcastle yourself though - see the details below).

Add a reference to White Tie's NuGet Package and then simply build your project in Release mode to generate reference documentation.

Details

If you haven't installed Sandcastle, then White Tie gracefully skips generating documentation. You must install Sandcastle yourself if you want White Tie to generate documentation for your project. Note that you may have to reboot your computer after installing Sandcastle for its changes to take effect. At the very least, you'll have to restart Visual Studio.

White Tie's only requirements are that you must install Microsoft's HTML Help v1 Compiler and Sandcastle's tools. The required steps are highlighted in the following image of the Sandcastle Guided Installer. The other options provided by the installer, such as SHFB Visual Studio Package, are not used by White Tie at all. Of course, you're free to install them anyway if you'd like.

White Tie's Required Sandcastle Installer Steps

After you've installed Sandcastle, documentation is generated automatically when you build your project in Release mode. This is configurable via MSBuild properties, as is almost everything in White Tie (see the following section for details). The result is a compiled help file (.chm) in your project's output folder (typically, bin\Release\). In addition, White Tie packs the compiled help file into the NuGet package that it generates for your project as well.

Documentation Configuration

The purpose of White Tie is to make packaging and deployment of OSS projects, or any other kind of project, really simple. White Tie uses sensible default configurations across all of its features. However, it's also highly configurable to allow for customizations.

White Tie offers many configuration options in the form of MSBuild properties and items that control if and how documentation is generated, a NuGet package is generated, static analysis is executed and output is copied to a local deployment directory.

The configuration options for generating documentation (as of White Tie v1.3.5) are as follows.
See the official documentation for details.

MSBuild Properties

  • $(BuildDocumentationEnabled)
  • $(DocumentationProjectName)
  • $(DocumentationProductName)
  • $(DocumentationCompanyName)
  • $(DocumentationCopyright)
  • $(DocumentationVersion)
  • $(DocumentationProjectSummary)
  • $(DocumentationRootTopicEnabled)
  • $(DocumentationRootTopicTitle)
  • $(DocumentationFeedbackEmail)
  • $(DocumentationFeedbackLinkText)
  • $(DocumentationHeader)
  • $(DocumentationFooter)
  • $(DocumentationStyle)
  • $(DocumentationLanguage)

MSBuild Items

  • @(DocumentationContent)
  • @(DocumentationImage)
  • @(DocumentationNamespace)
  • @(DocumentationSource)
  • @(DocumentationReference)
  • @(DocumentationFrameworkReference)

Undocumented Sandcastle Configuration Properties and Items

Advanced users of Sandcastle may be wondering how to configure Sandcastle's many other options that are missing from the list above.

Well, in addition to the properties and items listed above, White Tie supports several properties and items that are specific to Sandcastle; however, I've decided to separate these options from the primary group because they aren't general enough for White Tie. In the future, White Tie may support other documentation generators, and perhaps even replace its dependency on Sandcastle entirely. As a result, if you restrict your use to only those configuration options listed above, then in the future you shouldn't have to change anything in order to take advantage of whatever new ways White Tie generates documentation for your project. For example, I can imagine Microsoft offering a very fast and light-weight Roslyn documentation generator and in the future I'd like any project depending on White Tie to switch automatically, with little to no changes necessary, by simply upgrading to a newer version of White Tie.

If you really want to use Sandcastle-specific features anyway, then here they are:

  • $(_SandcastleCopyrightUrl)
  • $(_SandcastleVisibleItems) - Default value is "Attributes, ExplicitInterfaceImplementations, InheritedMembers, InheritedFrameworkMembers, Protected, ProtectedInternalAsProtected"
  • $(_SandcastleMissingTags) - Default value is "IncludeTargets"
  • $(_SandcastleIntelliSenseOutputEnabled) - Default value is "true"
  • $(_SandcastleCodeRemoveRegionMarkers) - Default value is "false"
  • $(_SandcastleHelpFileFormat) - Default value is "HtmlHelp1"
  • $(_SandcastleSdkLinkTarget) - Default value is "Blank"
  • $(_SandcastleSyntaxFilters) - Default value is "Standard"
  • $(_SandcastleHelp2SdkLinkType) - Default value is "Msdn"
  • $(_SandcastleContentPlacement) - Default value is "AboveNamespaces"
  • $(_SandcastleNamingMethod) - Default value is "Guid"
  • $(_SandcastlePreliminary) - Default value is "False"
  • $(_SandcastleBuildAssemblerVerbosity) - Default value is "OnlyWarningsAndErrors"
  • $(_SandcastleNamespaceGrouping) - Default value is "True"
  • $(_SandcastleMaximumGroupParts) - Default value is "2"
  • $(_SandcastleCatalogName) - Default value is "VisualStudio12"
  • $(_SHFBApiFilter) - Default value is "<ApiFilter />"
  • @(SandcastleColorizer) - Default list is "cs;vb"
  • @(SandcastleComponent) - Default list is empty; however, the Code Block Component, API Token Resolution and IntelliSense Component are included immediately before building, so you cannot remove them.
  • @(_SandcastleHelpAttribute) - Default list is "DocSet;DocSet;TargetOS". The Value metadata of these items are {@HtmlHelpName}, NetFramework and Windows, respectively.
  • @(_SHFBTransformArguments) - Default list is empty. You may include: logoFile, logoHeight, logoWidth, logoAltText, logoPlacement, logoAlignment, maxVersionParts and defaultLanguage. Set the Value metadata to the desired value for each item.
  • @(_SandcastleContentLayoutFile) - Default list is empty.
  • @(_SandcastleCodeSnippet) - Default list is empty.
  • @(_SandcastleTokensFile) - Default list is empty.
  • @(_SandcastleResourceItems) - Default list is empty.
June 29, 2014

OSS For OSS

White Tie
Visual Studio Settings Switcher

In this post, I'll introduce you to my new OSS project called White Tie and I'll describe how it relates to Rxx a bit.

I'll also introduce you to a new feature of another OSS project of mine, Visual Studio Settings Switcher.  This new feature allows you to easily share VS settings files with your solution, and OSS contributors to easily take advantage of it.

White Tie and VS Settings Switcher provide a set of features that I've specifically intended to be used by OSS developers, though anyone can benefit from them really.

I know that some of the ideas in these projects are probably not new, but perhaps the way in which I've implemented those ideas as features is new enough, and simple enough, to make them better than competing projects.

Your feedback is welcome.

White Tie

I've recently created the White Tie project on CodePlex.  I like to think of it as "formal buildware for use after 6PM".

White Tie is basically just a couple of unintrusive MSBuild files that are imported by your VS projects.  It's primarily intended for OSS developers who use Visual Studio (or MSBuild) to build their projects into NuGet packages, and optionally use Code Contracts, run static analysis and like to copy all build output into a local aggregate directory.  There are also work items currently under consideration to build reference documentation via Sandcastle and to enable remote deployment scenarios, for example.

White Tie makes building, packaging and deployment easier by offering sensible, out-of-the-box configuration, though it's also highly customizable by adding particular MSBuild properties and items to your project file.  (Read the Configuration documentation for details.)

White Tie is delivered as a NuGet package, so it's really easy to integrate into your projects.  Simply build your project in Release mode and a NuGet package will be generated automatically.  No configuration is necessary for existing projects, in general, though read the note at the bottom here for details about build errors caused by missing assembly-level attributes in new projects.

One of White Tie's key features is the NuGetFlavor item group.  By adding project references as NuGetFlavor items, rather than as ProjectReference items, White Tie will automatically include the referenced projects' output into the primary project's NuGet package.

For example, let's say that you have a primary class library project called Project A, which targets the full .NET Framework 4.5.1.  If you add a NuGet reference to White Tie, then whenever Project A is built in Release mode (by default), White Tie will generate a NuGet package for Project A.  Now let's say that you also have a portable class library project called Project B, which targets a portable profile that includes Windows 8 and Windows Phone 8.  Project B is intended to be the same as Project A, only targeting different platforms.  You might refer to Project B as a "flavor" of Project A.  Now you can simply edit the project file of Project A to add a NuGetFlavor item that points to the project file of Project B.  Whenever you build Project A in Release mode, White Tie ensures that Project B is built as well, and it also includes the portable library built by Project B into the NuGet package that it builds for Project A, and in the correct target framework subfolder.

White Tie offers several more features and configuration options.  Take a look at the documentation to get started, and let me know what you think.

White Tie's Origins and Future

White Tie is something that I've been working on for a few years, in one form or another.  Much of the work began as part of the Rxx project, such as its support for NuGet, though some features even predate Rxx.

Since Rxx it has evolved into something a bit more useful, though I've only been using it internally for my own OSS projects and business-related products (under the name DaveSexton.Build).  I finally decided recently that if I was ever going to "officially" release Rxx again, I'd have to bring its build system and project configurations up-to-date first.  White Tie is basically the best of DaveSexton.Build, formalized and revised (or "reimagined", if you prefer advertising jargon) to take advantage of NuGet 2.8 features, including complete support for portable libraries.  My transition to White Tie is not yet complete, and there are currently some road blocks, but I'm committed to getting it working fully.

Note that this isn't a guarantee that I'm going to build another NuGet package for Rxx, although it's certainly a large step in the right direction.  Not only have I checked in updates to Rxx's source code to include references to White Tie already, but I've also begun to think about and make changes to the portable library configurations that Rxx will target in the future.  But even after that, I'm still going to have to comb over all of its APIs before I'll release an "official" package again.  I have OCD and I haven't reviewed its source code or work items thoroughly since Rx 2.0 was released, so my confidence level in Rxx is (unreasonably, perhaps) lower than it should be.  :)

Visual Studio Settings Switcher

I've recently released version 1.2 of the Visual Studio Settings Switcher project on CodePlex, also available in the Visual Studio Gallery.

The latest release adds a new feature that I think OSS developers really want: Solution Settings Files

If you have an OSS project, VS Settings Switcher makes it really easy to generate a Visual Studio settings file that you can check into source control along with your solution, for your OSS contributors to use.  This settings file is in the typical VS settings file format.  It's also XML, which makes it easy to edit in Visual Studio, which you should definitely do before checking it in.  I highly recommend removing all of the extraneous settings that nobody wants to be shared in OSS projects, like the positions of your VS tool windows and other environment settings.  Typically, you'll want to reduce this file to only a small subset, such as your code formatting settings for your project's programming language.

This feature is also great for those of us that contribute to OSS projects.  It makes applying an OSS project's Visual Studio settings a breeze, without any manual intervention required.  If a solution has a .vssettings file with the same name and in the same directory, then VS Settings Switcher will automatically apply its settings whenever your open that solution.  And when you close that solution or open a different solution later, your original settings are automatically restored.

Spread the word: It's a good idea now to check in a partial .vssettings file with your solution!

April 30, 2014

Checklist for Creating Rx Operators

This blog post was originally posted here on the Rx forum back in 2012. I've noticed through reference tracking to my blog that it still gets some hits once in a while, so perhaps people are finding it to be useful. I figure it may get more visibility if I copy it to my blog. Note that I've made a few small elisions, corrections, clarifications and additions here, though I've left it essentially the same as the original post.

Prerequisites

  • Know the existing Rx Design Guidelines document.  You will find references to it throughout this checklist.
  • Know how to properly use existing Rx operators.  This checklist is not about how to properly use existing Rx operators, it's about how to define new operators for use in your business applications, or for inclusion into custom reactive frameworks that are based on the same design principles as Rx.

For the purposes of this checklist, an observable operator is any instance, static or extension method that returns IObservable<T>, including those that are typically considered to be static factory methods or combinators. (§6.3)

Also for the purposes of this checklist, the term observable refers to any local variable, parameter or member that is of the type IObservable<T> or a delegate to a function that returns an instance of IObservable<T>.

Checklist

Below is a categorized list of factors that I think are important to consider when creating new observable operators, ordered in general from most important to least important. This is not intended to be an exhaustive list, yet.  I'll be happy to update it based on your feedback.

Behavior

  1. Implement your new operator by composing existing Rx operators together, whenever possible; otherwise, just use any overload of Observable.Create.  Do not implement IObservable<T> yourself.  (§§6.1, 6.2)
  2. Ensure that your operator satisfies the Rx grammar and assume that IObservable<T> parameters are well-behaved. (§§4.1, 6.6)
    OnNext* (OnCompleted|OnError)?
  3. Ensure that your operator serializes notifications and assume that IObservable<T> parameters are well-behaved. (§§6.7, 6.8, 4.2)
  4. IObservable<T> is a model for concurrency; therefore, assume that observable parameters execute concurrently when Subscribe is called. Ensure that your operator is thread-safe and, when merging or combining the notifications of multiple observable parameters, ensure that your output observable serializes notifications.  (§§4.4, 5.9)
  5. Assume that IObservable<T> parameters are cold. Convert their temperature before subscribing multiple times to avoid duplicating subscription side effects, if your operator does not have retry or repeat semantics.  (§5.10)
  6. Protect calls to user code.  (§6.4)
  7. Do not catch exceptions thrown by observers; i.e., calls to OnNext, OnError and OnCompleted.
  8. Implement lazy (deferred) execution when generating a cold observable.  Check arguments up front, but do not cause any side effects until Subscribe is called.  This includes scheduling work, iterating enumerable parameters, and mutating external state or parameters.  (§6.16)
  9. Implement unsubscription, and do it properly.  (§§6.17, 6.18)
  10. Parameterize scheduling when appropriate.  (§§6.9, 6.10, 6.11, 5.4)
  11. Avoid deep call stacks caused by recursion.  (§6.15)
  12. Watch for reentry when executing user code and assigning the result to a SerialDisposable.  Use the double indirection pattern to avoid the effects of race conditions.

Performance and Memory

  1. Do not block.  Execute asynchronously instead.  (§6.14)
  2. Do not introduce extra asynchrony. Avoid the ObserveOn operator, whenever possible. Rely on your notification source, if any, to handle asynchrony itself; alternatively, consider offering parameterized scheduling, especially if you're generating your own notifications.  (§§5.5, 6.9, 6.12)

Semantics

  1. Choose a name that is semantically appropriate for the operator based on its business requirement and intended usage rather than behavioral details; e.g., TakeUntil is a better name than SecondStopsFirstGetCustomerOrders is a better name than SendOrdersRequest or GetServerResponses.
  2. Do not include implementation details in names except to distinguish between otherwise ambiguous operators and parameters.
  3. Use pluralization to indicate that an observable may generate more than one notification; e.g., LoadImages.
  4. Consider naming an extension method returning a hot observable as if it's a property; e.g., MouseMoves.
  5. Add an "Observable" suffix to distinguish an operator from existing synchronous and asynchronous methods that have similar names; e.g., Stream.ReadObservable.

Documentation

  1. Specify whether the observable returned by your operator is synchronous, asynchronous or concurrent when Subscribe is called.
  2. Specify whether the observable returned by your operator is hot or cold.  Be specific about what, if any, side effects may occur when the operator is called and/or when Subscribe is called.

Style

  1. Consider whether creating an asynchronous method(C# 5; VB 11) is a better fit.  This may be true when the generated observable is a singleton (cardinality = 1) and callers aren't necessarily dependent on Rx.  Any Task-returning operator is easily converted by callers into an observable via the ToObservable method.  Note that when complex control flow is required or the operator depends on Task-returning methods (e.g., when await is useful) you don't necessarily have to define an async method.  Instead, Rx 2.0 Release and Rx 1.1 Experimental define overloads of Observable.Create for defining async iterators.
  2. Avoid using subjects.  It's alright to use them implicitly if necessary; e.g., Publish.  (To Use Subject or Not To Use Subject)
  3. Avoid closing over local variables defined in the outer method body.  Sometimes this pattern is useful, but often it's a mistake that causes an otherwise cold observable to behave unpredictably because state is shared among multiple calls to Subscribe.

Tags:

Rx

August 05, 2013

Hot and Cold Observables

I've touched on the topic of observable temperature while answering questions on the Rx forum and in my previous post on subjects.  I tried explaining it more deeply in this particular discussion, but now I'm finally ready to provide a deep and comprehensive1 analysis in this blog post.

I'll answer the following questions: 

What really makes an observable hot or cold?
Why should I care?
How can I tell if an observable is hot or cold?
How can I change the temperature of an observable?

The analysis that follows is not intended for beginners, though you may find the following section and the conclusion to be informative.  You may also want to watch this first and look here for additional information.

Common Sense

We like to think of hot observables from two different perspectives as follows.

From the perspective of observers, there's the potential to miss notifications.  Hot observables are "live" sequences, meaning that they are happening whether we observe them or not.  The canonical example is a mouse, which of course is free to move regardless of whether you subscribe to an IObservable<MouseMoveEventArgs> or not.  In general, if I hand you an observable and tell you that it's hot, then as an observer you'd infer that you might have missed notifications that happened before you subscribed.  Pretty simple.

From the perspective of observables, hot observables broadcast notifications to all observers.  The canonical example is converting an event into an observable; e.g., FromEventPattern.  If three observers subscribe to IObservable<MouseMoveEventArgs>, then each observer will serially2 observe the same notifications, as opposed to observing different notifications or concurrent notifications.

We like to think of cold observables from the same perspectives as follows.

From the perspective of observers, there's the potential for each observer to get different notifications.  Cold observables are "generated" sequences, meaning that they can generate different notifications for every observer.  Furthermore, observers may receive notifications asynchronously, with respect to each other.  The canonical example is Create, which can asynchronously generate notifications whenever an observer subscribes.

From the perspective of observables, cold observables won't generate notifications until an observer subscribes, and they generate notifications each time that an observer subscribes.  The canonical example is Range, which generates a range of numbers whenever an observer subscribes.

In summary, common sense tells us that:

Hot observables are always running and they broadcast notifications to all observers.

Cold observables generate notifications for each observer.

I intend to show that these ideas are more like symptoms rather than definitions.  I'll also identify a pattern to reduce them into a primitive concept: side effects.  Ultimately, I'd like to ensure that hot and cold are well-defined terms based primarily on the concept of side effects.3

Enough skipping rocks over the topic.  Let's dive into the watery refraction and expose the peculiarities of our naïve understanding of these concepts.

Uncommon Sense

We know that an observable can be either hot or cold, but can they mix?  Can an observable be generated while broadcasting?  Can it be generated  and yet always running?  Can it be always running without broadcasting?  Can the temperature change dynamically?

That last question is particularly interesting.  Can an observable change its temperature after an observer subscribes?  E.g., Observer E subscribes first, then F subscribes; gets a hot observable while F gets a cold observable.

Can an observable change its temperature for the same observer after it subscribes?  E.g., Observer G subscribes to a cold observable that eventually transitions into a hot observable, for G, while it remains subscribed.

What about specialized subscription behaviors, such as one that causes an observable to behave like it's hot until all observers unsubscribe and then it becomes cold again for a subsequent subscription, which flips it back to being hot again -- does that behavior exist?

The answer to all of these questions is YES.  Well, it seems more like YES and SORT OF, based on our common sense definitions.  I'll try to explain.

Those behaviors might seem strange, but you've probably used them together before.  They're implemented by a couple of well-known Rx operators; e.g., Replay and RefCount.  Similarly, PublishLast and some particular overloads of Publish have strange behaviors with regard to temperature.

For example, keeping in mind our common sense definitions of hot and cold, let's assume that you have an observable O and you call O.Replay() to get an IConnectableObservable<T>, which you assign to C and then call C.Subscribe(J) and C.Subscribe(K).  I think we'd all consider C to be hot (or at least warm, whatever that means) because all of its notifications are broadcast to J and K, even though notifications won't arrive until Connect is called.  It's not entirely hot because it doesn't necessarily satisfy the always running factor, hence the introduction of the confusing term warm.

So far, Replay is identical to Publish.

Then you call C.Connect(), which may cause J and K to observe notifications from O.  If O is hot, then the always running factor was satisfied anyway, but if O is cold, then it's not satisfied; regardless, I think we'd all consider C to be hot at this point because of its broadcasting behavior alone.

Finally, you call C.Subscribe(L), and now there's a race condition on L with respect to our common sense understanding of observable temperature.  If O generated any notifications sometime between the calls to C.Connect() and C.Subscribe(L), then L will immediately observe replayed notifications, thus C was technically a cold observable when L subscribed because those particular notifications aren't rebroadcast to J and K; i.e., C generates specific notifications for L.  It means that while C is hot for J and K, it's cold for L; however, it's not permanently cold for L because after C replays all of the notifications that L missed, then L will be observing the same exact broadcast notifications as J and K, which means that C transitions from cold to hot for L.  Although, if O hadn't generated any notifications before you called C.Subscribe(L), then C would start out hot for L just like J and K.

Apparently, sometimes Replay is hot (though only if its source is hot, and only before Connect is called), sometimes it's warm (though only if its source is cold, and only before Connect is called), sometimes it transitions from warm to hot (though only for observers that subscribed before calling Connect) and sometimes it transitions from cold to hot, with a race condition (though only for observers that subscribe after Connect was called).

Part of the reason that this is confusing is because always running, broadcast and generate are actually relative concepts.  They implicitly refer to the time of subscription, though we tend to ignore this fact.  Always running refers to pre-subscription, generate refers to the time of subscription and broadcast is a post-subscription concept.  This explains how operators like Replay can offer strange combinations of these behaviors.

The common sense meanings of hot and cold are diluted by their differences in relation to the time of subscription and the connectability of operators like Replay, with a dependence on the temperature of the source and the presence of a race condition.

Technically it's possible for an observable to be always running without broadcasting; e.g., simply don't subscribe to a hot observable.  Our common sense definition tells us that it's hot even though it's not broadcasting.  Always running and generating are opposite behaviors, which makes broadcasting seem more like a symptom rather than a definition.

It's also possible for an observable to be generated while broadcasting; e.g., publishing a cold observable.  Again, common sense tells us that it's hot even though it's not always running, because we know that an underlying observable is actually responsible for generating notifications, which are broadcast to all observers.  This makes always running seem more like a symptom rather than a definition.

And finally, as shown by the Replay example above, an observable can be always running and yet generated at the same time.  They'll overlap until generation completes, like emptying a buffer, after which the observable seamlessly transitions into the always running sequence.  It's hard to say whether this kind of sequence is purely cold based on the common sense definition of generation alone.  We almost want to call it, "hot with some additional notifications".  This makes generation seem more like a symptom rather than a definition.

In general, our common sense definitions of hot and cold aren't precise enough to be useful to observers so we tend to ignore them and make assumptions based on what really matters: side effects.  All we really care about is answering one simple question:

If I subscribe to your observable, might it cause side effects?

It would be great if we could specify this behavior in a word.  The terms cold and hot should provide the answers yes and no, respectively.

Subscription Side Effects

Wikipedia defines side effect as:

Modifying some state or causing an observable interaction with calling functions of the outside world.

However, sometimes mutation is the primary effect; e.g., assignment.  Clearly the definition is relative.

So let's define side effect for our purposes as:

Any effect that is not the primary effect.

The primary effect of subscribing to an observable (i.e., calling Subscribe) is to register an observer for callbacks, which probably means that the observer will be added to the observable's internal list of observers and a disposable will be created and returned to represent unsubscription.

Likewise, the primary effect of .NET event registration is to add a handler to the list of delegates in the event's MulticastDelegate instance.  We can easily convert any event into an observable.  The primary effect of subscribing to an observable event is to convert an observer into an event handler, register it with the event, and return a new disposable that unregisters the observer from the event.

That pattern remains the same for other native asynchronous conversions as well, such as FromAsyncPattern and ToObservable (Task/Task<T>).

In addition, the compositional nature of Rx (more on that later in this post) allows operators to form subscription chains.  Subscribing to the outer-most operator's observable causes it to subscribe to the previous operator's observable, which subscribes to another observable, and so on.  The entire subscription chain is certainly part of the primary effect of subscribing.  Ultimately, the goal is to subscribe to the inner-most observable through as many intermediary subscriptions as necessary.

Subscription chains inherit side effects.  If any inner subscriptions cause side effects, then the outer-most subscription inadvertently causes them too; therefore, our definition of subscription side effects implicitly includes the sum of all inherited subscription side effects.

Based on the ideas above, a subscription side effect is:

Any side effect other than subscribing to another observable, adding an observer to an observable's "list of observers" and creating a new disposable for unsubscription.

where the meaning of "side effect" is defined by wikipedia and "list of observers" is defined by the necessary conversion to the underlying asynchronous model, if any.

Perhaps for the sake of simplicity I should reduce that definition.  Side-effect inheritance is implied, as is adding an observer to a "list of observers" and returning a disposable for unsubscription.  All of these behaviors are intuitively part of the subscription mechanism.

Therefore, a subscription side effect is:

Any effect beyond an observable's subscription mechanism.

Examples of subscription side effects include:  Calling Schedule, OnNext, OnError, OnCompletedGetEnumerator or MoveNext, mutating a field, creating an object in memory, running a CPU-intensive computation, sending a web request, reading a file, ending a process, formatting your C drive, or really anything else that you can think of that isn't merely the subscription mechanism and may cause an observable effect on the outside world, including notifications.

Covering All Bases

Duality

If you're interested in the duality between observables and enumerables, then copy and paste the Covering All Bases section into your favorite text editor and replace the following character sequences (not necessarily on word boundaries):

asynchronous synchronous
observable enumerable
observe enumerate
subscription enumeration
subscribing enumerating
subscribe MoveNext
notification element
always running pre-calculated
broadcast broadcast/share
subject
connection
connect
IBuffer
pushed
listening in to
calling OnNext, OnError or OnCompleted
yielding

It works!  Some of the grammar and spelling is a bit off though, and some of the ideas and operators may be entirely irrelevant due to the lack of asynchrony in enumerables, but the given definitions of hot and cold and the summarization are correct.

The concept of subscription side effects completely covers our common sense understanding of temperature as follows.

Recall our definition of subscription side effects from above:

Any effect beyond an observable's subscription mechanism.

Recall our common sense definitions of temperature from above:

Hot observables are always running and they broadcast notifications to all observers.

Cold observables generate notifications for each observer.

Understanding how cold behavior relates to subscription side effects is easy:

Generate means calling OnNext, OnError or OnCompleted when Subscribe is called.  This can be done synchronously or asynchronously with an IScheduler.  Either way these actions are, by definition, subscription side effects.  That's pretty simple.

Now let's jump to the common sense definition of hot and see how it compares:

Always running means that as soon as an observer subscribes it begins listening in to notifications.  It implies that notifications will not be generated as a side effect of subscribing, but that the mere act of subscribing enables an observer to receive notifications that would have occurred anyway.  In other words, subscribing to an observable that is always running does not necessarily cause subscription side effects.  The primary effect of subscribing is all that is needed to begin receiving notifications asynchronously.

Broadcast means that when a notification is ready to be observed it is pushed to all currently subscribed observers, which does not seem to imply anything about subscription side effects.  Theoretically, an observable could asynchronously generate notifications upon subscription, pushing them to the original observer that caused these subscription side effects while also broadcasting them to any observers that are fast enough to subscribe before the notifications are generated.  That's not the same behavior as Replay, for example, because Replay doesn't broadcast the notifications that it generates for an individual observer.  Apparently this behavior doesn't exist in any Rx primitives.  When an observer subscribes to a cold observable, without publishing, it observes notifications that are generated for it alone; i.e., cold observables don't broadcast.

Broadcast in Rx means that an observable is not responsible for generating notifications itself, but instead broadcasts notifications on behalf of some underlying observable.  That's the difference between the common sense definitions of cold and hot, respectively; therefore, broadcasting does not cause subscription side effects, meaning simply that when you subscribe to a broadcasting observable there will be no subscription side effects.

Broadcasting a cold observable has to cause subscription side effects once, though it's not caused by an observer subscribing to the broadcasting observable, but the broadcasting observable subscribing to the cold observable; a.k.a., "connecting".  Connection is exposed as a public operation, either by a subject's ability to act as an observer or by an operator returning IConnectableObservable<T>.  Theoretically, connection isn't necessary if the underlying observable is always running because observers that haven't subscribed yet are going to miss notifications whether the broadcasting observable is connected or not; however, if the underlying observable isn't always running (i.e., it's cold), then connection is required to ensure that observers don't miss notifications before they've subscribed to the broadcasting observable.  As a result, exposing the connection behavior of broadcasting observables defers subscription side effects of the underlying observable.  This implies again that broadcasting doesn't cause subscription side effects.  Subscribing to a broadcasting observable is all that is needed to begin receiving notifications asynchronously from the underlying observable, although notifications may be deferred until connection if the observable is connectable.

An emerging pattern here is that subscription side effects aren't expected at all in hot observables, including the case where a hot observable broadcasts notifications for a cold observable, because subscription side effects are deferred until connection.  We'll call them connection side effects instead.

Connectability may be hidden from observers.  Subjects are observables, thus the AsObservable operator can be used to hide their connectability.  The same applies to IConnectableObservable<T>.  In either case, hot behavior is retained even when observers know nothing about the underlying connectability of observables; therefore, simply knowing that an observable broadcasts implies that it doesn't cause subscription side effects.  It may have caused or will cause connection side effects, but that's beyond an observer's control when connectability has been hidden.

In addition, some stateful behaviors in Rx use connection side effects to separate notifications from other kinds of subscription side effects.  This is perhaps the most confusing case as far as the common sense definitions of hot and cold are concerned.

Recall from a previous section, when we reviewed the strange behaviors of Replay and RefCount, we determined that our common sense definitions are diluted because of their differences in relation to the time of subscription and the connectability of operators, with a dependence on the temperature of the source and the presence of a race condition.

Replay, PublishLast, and overloads of Publish that have an initialValue parameter, broadcast all subscription side effects, including notifications, to any observer that subscribes before they are connected to the underlying observable.  They also broadcast all subscription side effects, including notifications, to any observer that subscribes after connection but before the underlying observable generates any notifications.  They broadcast all subscription side effects, excluding missed notifications but including future notifications, to any observer that has missed notifications.  The missed notifications aren't broadcast, they are generated specifically for the particular observer that missed them.

These operators show that hot and cold are relative terms with regard to different kinds of subscription side effects, of which notifications are just one example.  They separate missed notifications from other kinds of subscription side effects to return a relatively cold observable or a relatively hot observable, respectively.  This is possible because broadcasting is a post-subscription concept and the ability to miss notifications is a pre-subscription concept that can be remedied at the time of subscription through generation, thus allowing an observable to be hot for some observers and cold for others, simultaneously.

Furthermore, the ability of these kinds of observables to transition from the common sense definitions of cold to hot is of no importance now, given that all we really care about is subscription side effects.  In other words, the newly defined hot and cold terms are relative to the time of subscription and to different kinds of subscription side effects, so the ability to transition is irrelevant.  An observable either causes subscription side effects or it does not.

Thus we no longer need any confusing terms such as warm to describe the behavior of connectables.  A connectable observable is always hot before it's connected and is generally hot afterwards, though some operators apply special behavior to missed notifications to ensure that they aren't actually missed.  These operators return an observable that is cold with regard to missed notifications, for a particular observer, and hot with regard to any other kind of subscription side effect, including future notifications; however, because our new definitions of hot and cold make transitioning irrelevant, we can say absolutely that an observer that has missed notifications will get a cold observable from these operators.  (The race condition's effect on our new definition of cold will be addressed later in this post.)

In summary, our new understanding of temperature with regard to subscription side effects tells us that:

Hot and cold are relative terms with regard to the time of subscription and to different kinds of subscription side effects.

Hot observables either don't cause subscription side effects or defer them until connection, where they become connection side effects.

Cold observables rely on subscription side effects.

Perhaps we can distinguish between these terms further by defining whether subscription side effects may or may not occur:

Hot observables do not cause subscription side effects.

Cold observables do cause subscription side effects.  (Though keep reading because later on I'll relax this definition a bit.)

Finally, we now have a specific and useful definition of observable temperature:

Temperature indicates the propensity of an observable to cause subscription side effects.

Furthermore, we can largely ignore the relativity to particular kinds of subscription side effects since it only matters when discussing the special behaviors of Replay, Publish and PublishLast.  And those operators only distinguish missed notifications from other kinds of subscription side effects, so they can easily be described as generally hot yet cold with respect to missed notifications.

In general, the terms hot and cold cover all kinds of subscription side effects, including notifications. 

Concerning Replay, PublishLast, and Publish overloads with an initialValue parameter, they return observables that are cold for particular observers that have missed notifications, yet they are hot with respect to any other kind of subscription side effect.

Now let's test the reverse: We'll see if our original common sense definitions can be inferred from our new definitions of hot and cold.

Given an observable that is known to not cause any subscription side effects, one must infer that it's either always running, which never causes subscription side effects, or it must broadcast, which defers subscription side effects into connection side effects; therefore, we must refer to the observable as hot.

Our new definition seems to be holding up so far.  Let's see how it compares to the common sense definition of cold.

Given an observable that is known to cause subscription side effects, one can infer that it might generate notifications when an observer subscribes, thus we may refer to it as cold.

It seems like we've got a match; however, notice that I wrote "might generate notifications".  That's because, as described previously, subscription side effects are far more general than just generating notifications.  A cold observable, by our new definition, can cause any subscription side effect.  For example, it can send a web request or asynchronously read from a file, without ever generating any notifications.  Of course, generating a notification would be useful so it's quite common for these examples in particular to generate notifications asynchronously once the operation has completed.

There are other kinds of subscription side effects that do not generate any notifications - not even asynchronously.  For example, imagine tracking the number of observers subscribed to your observable; each time Subscribe is called an integer field is incremented by 1.  That's similar to the behavior of the RefCount operator, described above.  Another example is subscription logging, for diagnostic purposes, which may cause a file to be updated on disc every time an observer subscribes.  The common sense definition of cold doesn't cover these cases, but our new definition does!

It seems to me that our new definition of temperature entirely covers the common sense definition, and does so elegantly as diametrically opposed concepts.  Furthermore, our new definition of hot nicely ties together the two orthogonal concepts that make up the common sense definition of hot, while our new definition of cold identifies the important factor and generalizes it.

But why are subscription side effects the important factor?

Composition

LINQ operators enable the composition of observables into "queries".  Applying any operator to one or more observables means that it will subscribe to those observables when you subscribe to the query.

When defining a query, it's not uncommon to reference an observable multiple times, which means passing the same observable to different operators or to the same operator as multiple arguments.  Since we know that operators subscribe to the observables that you pass to them, referencing the same observable multiple times means that it will potentially have multiple subscriptions.

Some operators have the semantics of multiple subscriptions without multiple references.  For example, Retry only accepts a single observable and potentially subscribes to it multiple times.  In this case it's desirable to duplicate subscription side effects.  For example, if we pass to Retry an observable that makes a web request as its subscription side effect, then we want Retry to execute that subscription side effect every time it re-subscribes.  In other words, we want to retry the web request.

That's why subscription side effects are the important factor.  From the point of view of the Retry operator, we don't necessarily want to duplicate the observable's notifications (as indicated by our common sense definition of cold), we simply want to duplicate the subscription side effects, whatever they might be.  In other words, it's not about generating the notifications, it's about executing the side effects, which are often responsible for generating the notifications.

Sometimes an observable is referenced multiple times without the semantics of allowing duplicate subscription side effects.  For example, you could apply the Where operator twice to the same observable to create two different observables, then apply the Zip operator to merge them back into a single observable in a pairwise fashion.  The query's output semantics aren't important here; what's important is that you don't want to create two subscriptions to the original observable.  Instead, you want broadcast behavior.

Broadcasting is essentially the same thing as sharing subscription side effects among multiple observers.  In the previous example, imagine that the observable to which you are applying the Where operator is cold; e.g., each time that you subscribe it sends a web request to open a reactive socket connection to a server and then streams the responses as notifications.  The Zip operator is going to subscribe twice, but we don't want to send two web requests.  Our desire to broadcast the notifications is the same as our desire to broadcast the web request, because it's the web request (i.e., the subscription side effect) that is responsible for generating notifications in the first place.

And again, that's why subscription side effects are the important factor.  In the former case we wanted to duplicate them, while in the latter case we wanted to broadcast them.

Composition in Custom Operators

The previous cases are special because we had prior knowledge of an observable's temperature.  But what if we didn't?

Composite operators are queries themselves; however, they don't typically define contracts for the temperatures of their observable parameters.  In other words, when we define a custom operator, we must make assumptions as to the temperatures of observable parameters.

An obvious question is: Why can't we define temperature contracts on operators?

Well, technically you could, but then you'd be forcing callers in some cases to explicitly change the temperature of observables before calling your operator, which changes the declarative nature of some operators by bringing temperature into the foreground when perhaps it's unnecessary.

Furthermore, if a caller is unaware of the temperature, then an assumption has to be made anyway.  Unless the contract is propagated to a caller that can control the temperature - and now we've got a viral effect, as is common with contracts.

Alternatively, you'd probably want to have two overloads for every operator where temperature is of concern: one overload for cold observables, one for hot observables.  But how can you overload operators by temperature without defining new types?

Do we really need this complexity?  What are the disadvantages of making assumptions at the operator level?

If you were to assume that all observable parameters are hot, then it's fine if your operator relies on broadcast behavior, but it's wrong if your operator must duplicate subscription side effects; e.g., Retry.  Alternatively, if you were to assume that they are cold, then the reverse is true: It's fine if duplicating subscription side effects is what you want, but it's wrong if you need broadcast behavior.

Therefore, we need to consider the possibilities of conversion before we can determine which assumption is best in general.

Temperature Conversion

First, a quick note about language:

IObservable<T> is an immutable type; therefore, there's no way to mutate its temperature.  When I refer to "changing" an observable to a particular temperature or "making" an observable a particular temperature, I'm using those terms in the general sense of unary operator application.  For example:

var x = 10;
var y = -x;

You might say that the second line "negates x", but technically it doesn't because after the negation operation, x hasn't changed.  It's applying the negation operator to x and assigning the result to y.  The former is just a natural way of describing the essence of the operation.  Although it's ambiguous in this situation, it's not when dealing with IObservable<T> because it's immutable.  Therefore, "changing" an observable's temperature or "making" it have a different temperature simply refers to the application of some operator and the assignment of the result into a new variable, which holds a different observable with the desired temperature.

We can change a cold observable into a hot observable by applying the Publish operator, which has overloads that project the observable into a selector function as a hot observable (though the operator itself returns a cold observable) and overloads that return a connectable observable.  The latter defers subscription side effects until connection, effectively broadcasting them across multiple observers as connection side effects.

Publish makes a cold observable hot.4

Defer makes a hot observable cold.

The Defer operator allows us to add subscription side effects to a hot observable, thus making it cold.

Defer also allows us to convert invocation side effects into subscription side effects; e.g., given a function that returns a hot observable like Start or a C# 5.0 async method that returns a Task<T>, which you then convert into an observable with ToObservable, you can wrap either function in Defer to change its invocation side effects into subscription side effects, thus creating a cold observable from a hot function.

Alternatively, you can change the temperature of a hot observable by combining it in various ways with a cold observable.  For example, if you Concat a hot observable to a cold observable, then the result is a cold observable, by our new definition, because the subscription side effects of the cold observable are prepended onto the hot observable.  If you were to apply the Merge operator instead, then the result would also be a cold observable because, again, the new observable retains the cold observable's subscription side effects.  Remember, it's the subscription side effects that are important, which is why combining a hot observable with a cold observable in such a way that the subscription side effects of the cold observable are preserved yields a cold observable.

Notice, however, that converting from cold to hot simply adds broadcasting behavior, while converting from hot to cold relies on the addition of subscription side effects.  In other words, going from cold to hot is as simple as calling one operator, but going in reverse requires side effects to be added.

In the previous section, we left wondering which assumption is best for an observable with an unknown temperature: cold or hot.  Perhaps we can answer that now by examining the consequences of making wrong assumptions with respect to temperature conversions.

When we need a hot observable, assuming that it's already hot and being wrong means that subscription side effects are duplicated, which breaks the semantics of our query; therefore, let's assume that every observable is cold and apply Publish.  When we're wrong, we'll end up applying broadcast behavior to a hot observable, which would simply have no effect.  That's 1 point for assuming that observables are cold, because when we're wrong, applying Publish does no real harm.5

When we need a cold observable, assuming that it's already cold and being wrong means that subscription side effects won't be duplicated, which breaks the semantics of our query; therefore, we must assume that every observable is hot and... do what?  Remember, the semantics of our query aren't to introduce new subscription side effects, we want to duplicate the subscription side effects of the observable that was handed to us.  We can't make a hot observable cold without introducing subscription side effects; therefore, all we can do is assume that the observable is already cold and hope for the best!  That's 2 points for cold, 0 points for hot.

In summary:

We must assume that any observable with an unknown temperature is cold.

In other words, it's safest to assume that any given IObservable<T> may cause subscription side effects because it's easy to broadcast subscription side effects when necessary, but once broadcast it's impossible to reverse it.  There's no way to force a broadcasting observable's Subscribe method to cause subscription side effects that were already broadcast.

Note that there is a way to reverse the broadcast behavior of an IConnectableObservable<T>: simply dispose of the object that was returned by the Connect method.  In fact, that's exactly how RefCount works.  But this can't be done with any IObservable<T>.  Even trying to cast it to IConnectableObservable<T> won't help because casting doesn't give you a reference to the disposable that was returned by Connect.

Given that it's safest to assume that any unknown IObservable<T> is cold, it follows that sometimes we're going to be wrong.  We should probably relax our previous definition of cold to account for that fact.

Here are our latest definitions: 

Hot observables do not cause subscription side effects.

Cold observables may cause subscription side effects.

That should cover all possibilities.

Classification of Subscription Side Effects

A subscription side effect can be classified as any of the following.6

  • Synchronous
    • Notification
      E.g., directly calling OnNext, OnError or OnCompleted.
      (Technically, notifications are always Asynchronous in Rx due to its implicit use of the CurrentThreadScheduler to schedule subscriptions, though we'll ignore that fact and define Synchronous in relation to the outer-most subscription only, where synchronously calling OnNext, OnError or OnCompleted actually blocks Subscribe from returning.)
    • Out-of-band
      E.g., directly assigning a field, creating a large object in memory, executing a CPU-intensive computation or (synchronously) deleting a file on disc.
  • Asynchronous
    • Notification
      E.g., using an async mechanism, such as Task or IScheduler, to schedule calls to OnNext, OnError or OnCompleted.
    • Out-of-band
      E.g., making a web request or reading a file on disc.  Often these kinds of operations are used for Asynchronous Notification as well.
  • Composition
    Subscribing to another cold observable that causes its own subscription side effects, either synchronously or asynchronously.
  • Computation
    Executing costly or side-effecting code for each notification; e.g., a costly projection via Select or imperative side effects via Do.

Perhaps these classifications could be useful when documenting operators.

For example, the documentation for the Range overload without an IScheduler parameter could state that it returns an observable that causes Synchronous Notification subscription side effects.  The overload with an IScheduler parameter may cause Synchronous or Asynchronous Notification subscription side effects, depending upon the specified IScheduler; this is similar for all operators that are overloaded with an IScheduler parameter.

The documentation for the Interval overload without an IScheduler parameter could state that it returns an observable that causes Asynchronous Notification subscription side effects.

The documentation for the Defer operator could state that it returns an observable that may cause Synchronous Out-of-band subscription side effects, depending upon the specified function.  It could also state that it may cause Composition subscription side effects, which basically means that it inherits the subscription side effects of the observable that is returned by the specified function.

The documentation for the Select operator could state that it returns an observable that may cause Composition subscription side effects, which basically means that it inherits the subscription side effects of the source.  It could also state that it may cause significant Computation side effects, depending upon the specified selector function.

Most of these terms should be self-explanatory by now; however, Computation is particularly interesting since I haven't mentioned it yet.  I consider a computation subscription side effect to be:

Any effect beyond an observable's subscription and notification mechanisms that occurs for each notification.

The notification mechanism is the minimally required effects for notifying; i.e., directly calling OnNext, OnError or OnCompleted are the primary effects of notification.  Anything else is a computational side effect.

Although this kind of side effect is part of an observable's computation, it won't occur until subscription.  Subscribe is essentially the beginning of the computation, thus subscription side effects are technically computation side effects.  But since subscription side effects are the focus of temperature, that's why I refer to computation side effects as a kind of subscription side effect, in the sense that computation side effects don't occur until an observer subscribes.

Computation side effects are part of the definition of cold observables, though it basically means that applying any computational or side-effecting operator to a hot observable converts it into a cold observable; e.g., applying Select to a hot observable means that all observers will receive different notifications, generated specifically for them, which is the common sense definition of cold.  That's not a very useful definition of cold, so I propose to separate the notion of computation side effects from subscription side effects and to exclude them from the new definition of cold, introducing new terms: active and passive.  This separation allows a hot observable to be referred to as hot even if it causes computation side effects.  And you can still apply Publish to broadcast the computation side effects similar to how you'd publish a cold observable to make it hot.

Passive observables do not cause any significant computation side effects.

Active observables do cause significant computation side effects.  Every notification may be computed with some relative cost.

Select and Do are common examples of operators that return active observables.

These terms are context-sensitive, as indicated by use of the word "significant".  An observable that is passive in one usage can be considered active in another.  For example, a simple projection over an observable of integers, such as adding 5 to each value, probably isn't costly enough in terms of memory or performance to care if its computation side effects are duplicated.  In contrast, a projection that calculates the factorial of each value might be costly enough in certain scenarios to consider publishing.

Conclusion

Temperature indicates the propensity of an observable to cause subscription side effects.

A subscription side effect is:

Any effect beyond an observable's subscription mechanism.

Subscription mechanisms generally consist of storing an observer in a list or subscribing to other observables, and returning a disposable for unsubscription.

Examples of subscription side effects include:  Calling Schedule, OnNext, OnError, OnCompletedGetEnumerator or MoveNext, mutating a field, creating an object in memory, running a CPU-intensive computation, sending a web request, reading a file, ending a process, formatting your C drive, or really anything else that you can think of that isn't merely the subscription mechanism.

Hot observables do not cause subscription side effects.

Cold observables do cause subscription side effects; however, we must assume that any observable with an unknown temperature is cold, and sometimes that assumption will be wrong; therefore, a more accurate definition is:

Cold observables may cause subscription side effects.

Again, we must assume that any observable with an unknown temperature is cold.

In the simplest case, an observable may be permanently hot or permanently cold, though some observables change their temperature over time so that different observers may get different temperatures; therefore:

The temperature of an observable is relative to the time of subscription.

Temperature is also relative to the kind of subscription side effect; however:

In general, the terms hot and cold cover all kinds of subscription side effects, including notifications.

Replay, PublishLast, and Publish overloads with an initialValue parameter return observables that, once connected, are cold for observers that have missed notifications yet hot with respect to any other kind of subscription side effect.

We must also consider an observable's computation side effects.  A computation side effect is:

Any effect beyond an observable's subscription and notification mechanisms that occurs for each notification.

Technically subscription side effects are initial computation side effects, although I've separated these concepts so that hot and cold only refer to subscription side effects, thus ensuring that hot is a more useful term.  As a result, the phrase computation side effects can be applied to either hot or cold observables.  I've assigned the following terms:

Passive observables do not cause any significant computation side effects.

Active observables do cause significant computation side effects.

The significance of computation side effects on performance is relative to the operation, query or application.  What is significant in one context may be insignificant in another.

Now I can finally answer the original questions from the top of the post with absolute precision and certainty.

What really makes an observable hot or cold?

Short answer:

An observable's temperature is its propensity to cause subscription side effects.

Long answer:

Hot observables do not cause subscription side effects.

Cold observables may cause subscription side effects.

In addition, there are two related concepts that may be applied to both cold and hot observables:

Passive observables do not cause any significant computation side effects.

Active observables do cause significant computation side effects.

Why should I care?

You need to be aware of whether subscribing to an observable multiple times is going to duplicate side effects; i.e., cold behavior.

When defining a query that references the same observable multiple times, you'll typically want to broadcast notifications among the operators to ensure that they observe the same exact notifications.  In other words, you want a hot observable.  A cold observable may generate different notifications for each observer.  You'll also want to broadcast other subscription side effects; e.g., given a cold observable that makes a web request each time an observer subscribes, you don't want two operators in the same query to make two separate web requests.

Alternatively, sometimes you want duplicate subscription side effects in your queries; e.g., given a cold observable that makes a web request each time an observer subscribes, applying the Retry operator means that you want to keep making web requests until one succeeds.  The very purpose of the Retry operator is to duplicate subscription side effects.  It's perhaps a fairly common mistake to pass a hot observable to Retry and then wonder why Fiddler isn't showing a second web request after the first attempt failed.

Furthermore, you must consider computation side effects to determine whether they may have a noticeable performance impact on your query.  If they do, then you'll probably want to broadcast them as well.

How can I tell if an observable is hot or cold?

There's no way to be sure from an observer's point of view.  Subscription side effects might be directly observable or they might not.

Furthermore, subscription side effects that generate notifications might be indistinguishable from the notifications that would be generated if you assumed that the observable is hot.

The only way to know for sure is to look at the documentation source code.  :)

When in doubt, it's safest to assume that an unknown observable is cold.

Temperature by Operator

(You may want to review Rx operators by category, although it's somewhat outdated as of Rx 2.0).

All of the Rx generator methods return observables that are of course cold by definition: Return, Range, Timer, IntervalGenerate and Create.

There are 4 additional generators that are kind of edge cases: Empty and Throw are cold; Start is cold too, but we'll get back to it later; Never is technically hot, but it doesn't really matter.

In general, it's safe to assume that Rx's built-in conversion operators for .NET events return hot observables; e.g., FromEvent and FromEventPattern.  In reality, there's no guarantee.  For example, you can define a custom .NET event and implement an add method that synchronously invokes each handler that it receives.  That event would convert into a cold observable!  Luckily, events generally don't do strange things like that, so it's safe to assume that they are hot.

You can be sure that Replay, Publish and PublishLast overloads that return IConnectableObservable<T> are hot until you call Connect, though after Connect it's more complicated.  Publish overloads without an initialValue parameter are always hotReplay, PublishLast, and Publish overloads with an initialValue parameter, become cold when Connect is called since they introduce a race condition.

Each of those operators also have overloads that accept a selector function.  The selector overloads return cold observables, though the observable passed to the selector function is similar to above.  When the selector is called, the observable passed in is hot, but when the selector returns, the observable passed in is automatically connected and starts to behave as described previously.

FromAsyncPattern and ToAsync return functions that return observables similar to PublishLast, but connected when the function is invoked.  It's safest to assume that the observable is cold, just like PublishLast.  You can think of the function as a parameterized connectable since it acts like the Connect method of IConnectableObservable<T>.

Start is similar to ToAsync and PublishLast, though it's not connectable, thus it's safest to assume that it returns a cold observable.

Task and Task<T> are themselves hot, although the ToObservable conversion returns an observable similar to PublishLast, thus it's safest to assume that ToObservable returns a cold observable.

Defer returns a cold observable, though actually it depends on whether it causes subscription side effects or not.  Typically the only reason that you wouldn't know whether it causes subscription side effects is when you use it to defer an unknown function that returns an observable.  The unknown function might not cause any invocation side effects and may return an observable that doesn't cause any subscription side effects, thus Defer may simply be passing through a hot observable.  That behavior indicates either an edge case or improper usage of the operator.

StartWith, Sample and Using return cold observables.  They all cause subscription side effects.

All of the remaining operators inherit temperatures from their source observables.  Well, maybe not all remaining operators – I may have missed a couple when I reviewed them for this post ;)

Select, Do, Aggregate and Scan may cause significant computation side effects.  (See above for a description of computation side effects.)

Technically, any operator with a delegate parameter that's invoked for some or all notifications may cause significant computation side effects.  There are many more operators like this than what I've listed above.

Remotable causes significant computation side effects, assuming that you consider transmitting notifications through the .NET remoting infrastructure to be significant in terms of its performance implications.

How can I change the temperature of an observable?

Subscription side effects:

Simple Conversions

Publish is useful for broadcasting subscription side effects.  It makes a cold observable hot.

Defer is useful for adding subscription side effects and for converting invocation side effects into subscription side effects.  It makes a hot observable or observable-returning function cold.

Operators such as Merge and Concat, among many other combining/zipping/joining operators, can add subscription side effects to hot observables by combining them with cold observables, thus returning cold observables.

Complex Conversions

Using a Subject as an observer makes a cold observable hot, though it's not a good reason to use subjects.  See my previous post on subjects for details.

Replay, PublishLast, and Publish overloads with an initialValue parameter, are useful for broadcasting subscription side effects while also replaying notifications to observers that missed them.  These operators make cold observables hot in terms of all subscription side effects except for missed notifications; therefore, they remain cold in regard to missed notifications.

Computation side effects:

Publish is useful for broadcasting computation side effects.  Whether or not you want to broadcast them depends upon their significance with respect to resource usage and/or performance and how they compare to your operation, query or application.

Operators such as Select and Do, among many others, are useful for adding computation side effects.

 


  1. Not in a mathematical or academically formal sense.  Posting the idea to my blog allows me to focus on rationalizing it and elicits proper feedback.  Though if anybody can explain it in a mathematically formal sense, then by all means please describe it in the comments.
  2. Serial behavior is described in §§4.2, 6.7 of the Rx Design Guidelines document.  Note that FromEventPattern does not enforce this contract, so it's either the raiser's responsibility to ensure that it does not raise the event concurrently or it's an observer's responsibility to apply the Synchronize operator.
  3. Obviously not in any kind of official capacity.  I just need a way to reference these new concepts now.  I'd be happy to adopt better or formal naming conventions if they already exist or will emerge in the future.
  4. Publish, PublishLast and Replay also have convenient overloads accepting a selector function that acts as a scoped query in which the original observable is hot, thus allowing callers to avoid having to deal with connection themselves.  The temperature of an observable returned by these overloads is cold because it publishes and connects as a subscription side effect.
  5. Perhaps it would be nice if Rx's implementation of Publish could somehow detect whether an observable is either Subject<T> or it's hidden by the AsObservable operator, in order to avoid introducing yet another subject into the query unnecessarily.  But that would be an internal optimization and it doesn't change the fact that publishing a hot observable probably has no noticeable performance impact in most programs.
  6. I've made all of these terms up myself.  I'll gladly replace them with formal terms if someone provides them to me in the comments.

Tags:

.NET | Rx

June 22, 2013

To Use Subject Or Not To Use Subject?

That is the question!  There appears to be some confusion on the web about whether or not Subject<T> should be used, ever.  Tis a question oft asked by developers new to Rx after reading expert recommendations to avoid subjects, due to common misuse, yet subjects seem to persist an air of mystery even for seasoned reactive developers.

First, a primer on Subjects.  Take a moment to read that.

Done?  Great.  Let's continue...

As you may already know, Erik Meijer doesn't like subjects.  End of discussion, right?  Wrong!  Erik doesn't like subjects to be used in place of purely functional alternatives.1

As Erik puts it:

[Subjects] are the "mutable variables" of the Rx world and in most cases you do not need them.

So let's enumerate the cases where we should and shouldn't use subjects.

We'll start by examining two interesting properties of subjects:

  1. They implement IObservable<T> and IObserver<T> simultaneously, which makes them suitable for both output and input, respectively.
  2. They are hot(See also: Hot and Cold Observables)

Thus to answer the question of whether we should use a subject when defining an observable (output) we must consider:

  1. the input; i.e., the source of the OnNext, OnCompleted and OnError notifications.
  2. the desired temperature of the generated observable; e.g., hot or cold.

To determine whether or not to use a subject, start by asking yourself the following questions:

  1. what is the source of the notifications?  I.e., where do the values of T come from?  Who is responsible for calling OnNext?
  2. must observers share subscription side-effects or not?  I.e., does calling Subscribe execute some kind of initialization code each time it's called or does initialization occur once and its state is shared by every call to Subscribe?

Let's answer all possible combinations of these 2 questions.  I'll provide more details and examples later in this post to explain why these answers make sense.

Question 1:  What is the source of the notifications?

The source can be one of two kinds: Let's call them external and local.  An external source is any observable or event that already exists outside of your code.  A local source is when you generate an observable from your code, kind of like defining and raising a custom .NET event.  More details on these distinctions later in this post.

  1. If the source is external, then do not use a subject.
  2. If the source is local, then do use a subject.

Question 2:  Must observers share subscription side-effects or not?

  1. If not, then do not use a subject.  (Define a cold observable.)
  2. If yes, then do use a subject.  (Define a hot observable.)

Let's enumerate all possible combinations: 

  1. The source is external and I want a cold observable.
  2. The source is external and I want a hot observable.
  3. The source is local and I want a cold observable.
  4. The source is local and I want a hot observable. 

As you may have noticed, there are conflicts.

  • What if you have an external source and you do want to share subscription side-effects?
  • What if you have a local source and you do not want to share subscription side-effects?
  • What if the temperature of an external source differs from the desired temperature?

I believe these conflicts cause additional confusion among developers.  They seem to blur the line when it's appropriate or inappropriate to use subjects, though in fact the line is actually quite clear, as we'll soon see.

Let's resolve these conflicts now.

The source is external and I want a cold observable.

  1. If the source is cold, then use it as is.  You can also wrap it with Defer to add subscription side-effects.
  2. If the source is hot, then wrap it with Defer to add subscription side-effects.

The source is external and I want a hot observable.

  1. If the source is cold, then Publish it.  Alternatively, you can use PublishLast or Replay, if you need their additional behavior.
  2. If the source is hot, then use it as is.

The source is local and I want a cold observable.

  1. Define your observable with existing Rx generator methods; e.g., Return, Range, Timer, Interval, Generate or Create.

The source is local and I want a hot observable.

  1. This scenario is a potential candidate for defining your observable as Subject<T>, or even BehaviorSubject, AsyncSubject or ReplaySubject, if you need their behavior.  More details on why it's merely a "potential candidate" later in this post.

Note that Publish, PublishLast and Replay use subjects internally (via Multicast).  This isn't a coincidence.  Subscribing a subject to a cold observable broadcasts its notifications to multiple observers, thus making it hot.

So it's clear there are only two scenarios where it's correct to use subjects:

  1. The source is external and cold, and I want a hot observable.
  2. The source is local and I want a hot observable.

Let's simplify those statements.  The two scenarios where it's correct to use subjects are:

  1. Converting a cold observable into a hot observable.
  2. Generating a hot observable imperatively and statefully, without any direct external source.

Although, only in the latter scenario is it correct to use subjects explicitly.  As mentioned previously, Rx defines various operators like Publish for use in the former scenario.  Thus, only 1 scenario remains.

And so now we've answered our original questions:

When should I use a subject?  To generate a hot observable imperatively and statefully, without any direct external source.
When shouldn't I use a subject?  Everywhere else.

Perhaps that answer is a bit vague.  What do I mean by imperative and stateful?  What about local vs. external?  And why did I mention before that our final scenario is only a "potential candidate" for using a subject?

If (true == interested_in_the_answers_to_these_questions) { continue; } else { goto Conclusion; }

External Sources

Let's redefine external as quite simply a source of notifications that you do not have to generate yourself.  It comes from outside the scope of the observable that you're defining, such as a dependent object or a parameter.  It could be any number of things; e.g., an event on another object, an event on a base class, a timer callback, a Task<T> that represents the asynchronous result of reading from a file or sending a web request, an IEnumerable<T>, etc.

All of these examples have something in common: they can easily be converted into IObservable<T> using various From* and To* operators in Rx.  For example, Task<T> is converted by the ToObservable operator.  So if your external source is not already an IObservable<T>, then simply covert it using one of Rx's existing conversion operators.

Once you have an IObservable<T>, you need to determine whether it's hot or cold and whether you need it to be hot or cold.  If there's a conflict, as defined above, then use either Defer or Publish to change the temperature accordingly.  For this, there's no need to use subjects in your code.2

Finally, use LINQ to query the observable.  Again, you shouldn't use subjects here either.

The result is that you'll have a succinct observable query, controlling side-effects and meeting your specification in a declarative manner.  Subjects are simply the wrong tool for this job.  It's like hammering in a nail with a sledgehammer.  Subjects do much more than what is needed.  (Though as you'll see later in this post, in some ways subjects actually do less than what you may think, potentially introducing nasty threading bugs.)

Local / Imperative / Stateful Sources

I'm defining local as a source where you imperatively push notifications toward the outside of a stateful scope.  Another way of looking at it: a local source is external to other scopes that can't access its implementation.

Sorry if that explanation is confusing.  Let's clear it up.  Consider the following abstract requirements:

  1. You must define a hot observable.
  2. The code that generates the observable's notifications must be encapsulated by some type that you define.
  3. Your code generates notifications when callers invoke certain members of your type or when it performs certain operations.
  4. The observable must be exposed as a public property.

Now let's apply these requirements to what we've already learned earlier: 

  1. Since we're defining a hot observable, Subject<T> is certainly a candidate...
  2. We've established a scope for what is external: any source defined by another type and referenced by ours is external to our type.
  3. We must write code to push notifications (input); therefore, our source is local.  We have no external source to convert or to generate notifications for us.
  4. Our code must be able to reference something in order to push notifications, thus our type must be stateful.  We'll store a reference to "something" in a field.
  5. The field will be returned by a public property (output).
  6. Thus we need a type that represents both input and output, and is hot.

How should we model these requirements?

Scroll down for the answer:
.
.
.
.
.
.
.
.

Answer: Define a .NET event and expose it as an observable public property using FromEventPattern!

Ha, I bet you didn't see that one coming, did you?

Yes, alternatively we could use Subject<T>.  The conditions are perfect.  If we don't actually need or want a .NET event, then we should use a subject; however, if we already have an event defined, then there's no need to duplicate state and "raise" two similar events.  The conversion approach is probably better in that case.

Let's compare the similarity of subjects to .NET events in general:

  1. Events are also hot.
  2. An event's delegate shares the same scope (encapsulation) within our type as our proposed observable.
  3. We also write code to raise events (input).
  4. Our code must be able to reference the delegate in order to raise events, thus our type must be stateful.  We store a reference to the delegate in a field.
  5. The field will be returned by a public event (output).
  6. A multicast delegate represents both input and output, and is hot.

Subject<T> is kind of like Rx's implementation of an observable "event".  It broadcasts notifications to multiple observers, like raising an event for multiple event handlers.  It is the stateful component of Rx as it can be stored in a field or a variable and mutated (invoked) imperatively.  It is hot, so subscribing does not cause any side-effects (and subscribers may have already missed notifications, which is a common symptom of events and hot observables alike.)

It's correct to use Subject<T> to replace local .NET events, though .NET events are still useful sometimes due to their metadata and first-class language/tooling support.  It all depends on context, so consider this carefully before replacing all of your .NET events with subjects.  You may want to examine each event on a case-by-case basis.  And keep in mind that you can always expose an event as an observable via FromEventPattern, or perhaps it's best to leave that decision up to the consumers of your class and simply expose a .NET event.

Subjects actually have an additional related use case that events cannot cover: If we shrink our scope from type to method and apply the same logic as before, then we'll discover that subjects may still apply though events do not.

Let's specify similar method-scoped requirements:

  1. You must define a hot observable.
  2. The code that generates the observable's notifications must be within a method that you define.
  3. Your code generates notifications when it performs certain asynchronous operations within the method.
  4. The observable must be exposed as the return value of the method.

And again, let's apply these requirements to what we've already learned earlier: 

  1. Since we're defining a hot observable, Subject<T> is certainly a candidate...
  2. We've established a scope for what is external: any source passed into our method and any fields on the method's defining type are external to our method.
  3. We must write code to push notifications (input); therefore, our source is local.  We have no external source to convert or to generate notifications for us.
  4. Our code must be able to reference something in order to push notifications, thus our method must be stateful.  We'll store a reference to "something" in a local variable.
  5. The variable will be returned from the method (output).
  6. Thus we need a type that represents both input and output, and is hot.

How should we model these requirements?

Scroll down for the answer:
.
.
.
.
.
.
.
.

Answer 1: Write an anonymous event (a.k.a., anonymous method, delegate or lambda expression) and convert it using FromEvent!

Ha, did I trick you again?

You thought for sure this time I'd recommend using Subject<T>.  You'd also be correct, we could use Subject<T>.  The conditions may be perfect, though as you'll see in a moment there are better alternatives, even better than this one.

Alright, that answer was actually a joke, I don't recommend that particular approach at all.  It was meant to get you thinking about alternatives to subjects in places where you may think they're absolutely necessary.

But out of curiosity, how would that have worked anyway?

An event is backed by a delegate; it's the delegate that has all of those important properties (e.g., input, output, hot) that we discussed earlier.  For example, we could start by declaring a delegate as our observer.  By assigning the delegate to a local variable we can invoke it imperatively like a method, which is similar to calling OnNext on an observer.  We can use Task.Factory.StartNew to execute an async computation that invokes the delegate.  Finally, we can convert our delegate into an observable via the FromEvent method, which gives us our return value.

NOTE:  I don't recommend this approach, as mentioned previously.  For one thing, it doesn't support OnError or OnCompleted.  It doesn't support cancellation either, but given that we're generating a hot observable perhaps it doesn't really matter.  It's also a bit strange anyway.

using System;
using System.Reactive.Linq;
using System.Threading.Tasks;

namespace Rx.Labs
{
    static class ToUseSubjectOrNotToUseSubject
    {
        public static void Main()
        {
            IObservable<string> xs = Generate();

            using (xs.Take(1).Subscribe(Console.WriteLine))
            using (xs.Take(2).Subscribe(Console.WriteLine))
            using (xs.Take(3).Subscribe(Console.WriteLine))
            using (xs.Take(4).Subscribe(Console.WriteLine))
            {
                Console.ReadLine();
            }
        }

        static IObservable<string> Generate()
        {
            Action<string> observer = _ => { };    /* We could use null, but then at every invocation 
                                                    * we'd have to copy to a local and check for null.
                                                    */

            Task.Factory.StartNew(() =>
                {
                    Console.WriteLine("Invocation side-effect.");

                    Task.Delay(TimeSpan.FromSeconds(1)).Wait();  // Simulate work

                    observer("Generating");
                    observer("a");
                    observer("hot");
                    observer("observable.");
                });

            return Observable.FromEvent<string>(
                eh => observer += eh,
                eh => observer -= eh);
        }
    }
}

Output:

Invocation side-effect.
Generating
Generating
Generating
Generating
a
a
a
hot
hot
observable.

Answer 2 (Recommended):  Define an observable using an existing Rx generator method, then apply Publish with RefCount.

Rx provides various generators that make it easy to create an observable from thin air; e.g., Return, Range, Timer, Interval, Generate, and also Create.  If none of the former generators apply, then fallback to CreateCreate is the Swiss Army knife of observable generation.  I suspect that it also has similar performance characteristics to Subject<T>.  (See the Thread Safety section below for details.)

A particularly useful approach is to write an async iterator using a special overload of Create:

Rx 2.0 (Preferred):    Create<TResult>(Func<IObserver<TResult>, Task>)
Rx 1.1 Experimental:   Create<TResult>(Func<IObserver<TResult>, IEnumerable<IObservable<object>>>)

Generators are better than subjects because they allow you to remain functional, though Create does seem quite similar to using a subject.  Create lets you take more of a declarative approach, which is hopefully easier to understand and maintain than code that uses a subject.  Note that a primary benefit of Create is that it supports cancelation and auto-detachment, though it's not very useful when you need a hot observable.

Generators aren't the whole story either.  They return a cold observable, but our specification was to return a hot observable.

That's where Publish is useful, as described earlier in this post.  Publish converts a cold observable into a hot observable; however, it returns IConnectableObservable<T>, which requires a call to Connect.  You can return IConnectableObservable<T> from your method, just make sure that you change the return type of your method to match so that callers have the opportunity to call Connect themselves.

Alternatively, you can call Publish then RefCount to get an IObservable<T> that is shared among multiple consecutive observers.  Note that this isn't truly a hot observable - it's more like a warm observable.  RefCount makes a single subscription to the underlying observable while there's at least one observer of your query.  When your query has no more observers, changing the reference count to 0, the underlying subscription is disposed.  If another observer subscribes to your query later, moving the reference count from 0 to 1 again, then RefCount makes a new subscription to the underlying observable, causing subscription side-effects to occur again.

Note that it's bad practice to call Connect yourself in this scenario.  If you're tempted to do so, perhaps because RefCount doesn't meet your needs, then you should strongly consider returning a cold observable instead and leave it up to callers to decide whether Publish should be used at all.

Also beware to avoid closures when using Create.3  Closing over a local variable that you intend to mutate may cause unexpected behavior since every observer will share the same variable.  This may seem fine when generating a hot observable, and perhaps it is, though RefCount generates a warm observable, as mentioned previously.

Here's a correct example that uses Create, Publish and RefCount instead of a subject

using System;
using System.Reactive.Concurrency;
using System.Reactive.Linq;
using System.Threading.Tasks;

namespace Rx.Labs
{
    static class ToUseSubjectOrNotToUseSubject
    {
        public static void Main()
        {
            IObservable<string> xs = Generate();

            using (xs.Take(1).Subscribe(Console.WriteLine))
            using (xs.Take(2).Subscribe(Console.WriteLine))
            using (xs.Take(3).Subscribe(Console.WriteLine))
            using (xs.Take(4).Subscribe(Console.WriteLine))
            {
                Console.ReadLine();
            }
        }

        static IObservable<string> Generate()
        {
            return Observable.Create<string>(observer =>
                TaskPoolScheduler.Default.Schedule(() =>
                {
                    Console.WriteLine("Subscription side-effect.");

                    Task.Delay(TimeSpan.FromSeconds(1)).Wait();  // Simulate work

                    observer.OnNext("Generating");
                    observer.OnNext("a");
                    observer.OnNext("hot");
                    observer.OnNext("observable.");

                    observer.OnCompleted();
                }))
                .Publish()
                .RefCount();
        }
    }
}

Output:

Subscription side-effect.
Generating
Generating
Generating
Generating
a
a
a
hot
hot
observable.

Thread Safety

Subject<T>, Observable.Create and ObserverBase<T> do not make the same guarantees that are made by Rx generator methods, for performance reasons.  They allow callers to violate some of Rx's contracts for IObserver<T>, though they do ensure some contracts automatically regardless.

The contracts to which I'm referring are defined in the Rx Design Guidelines document.

Section 4.2 basically states that notifications must be pushed serially for Rx operators to behave properly.  But more generally, this should be a safe assumption everywhere.  When you subscribe to an observable, you don't expect to observe concurrent notifications.  Imagine coding subscriptions against this assumption but actually receiving concurrent notifications at runtime.  Take a look at some of your existing code.  What could happen?  For example, if you subscribe on a UI thread then you may get a threading context exception.  If you subscribe on some other thread instead, then maybe you'll get some strange threading bugs or maybe you'll get an exception, if you're lucky.

All Rx generator methods, excluding Create, ensure that this contract is satisfied.  Subject<T> and ObserverBase<T> do not.  For instance, calling OnNext concurrently on a Subject<T> will generate concurrent notifications.4

Section 4.1 states the Rx grammar as follows:

OnNext* (OnCompleted | OnError)?

All Rx operators and generator methods ensure that this contract is satisfied.  ObservableBase<T>, ObserverBase<T>, Create and Subject<T> actually ensure this contract as well, but only for serial notifications.  If a caller doesn't satisfy the §4.2 contract, then it seems that these APIs do not automatically ensure the §4.1 contract is satisfied either, due to lock-free race conditions.  For instance, it seems to me that calling a subject's OnNext and OnError methods simultaneously on different threads may cause an observer to receive OnError followed by OnNext.  But if you call OnError followed by OnNext on the same thread, then the call to OnNext is correctly ignored.

Rx offers Synchronize to satisfy these contracts for a given subject.  It offers this Synchronize method to do the same for observers, in general.  And of course, just because you can violate the contracts doesn't mean that you necessarily will.  But you do have to be aware of them when using these APIs, especially considering the kinds of bugs that may arise.  Threading bugs originating from within Rx operators, due to the misuse of a subject somewhere within a query, can be very difficult to diagnose.

So when you find yourself invoking OnNext, OnError or OnCompleted on a subject or within Create, make sure that you're satisfying Rx's contracts.

The takeaway is that when a subject isn't necessary due to the presence of an external source or an appropriate generator method, it's best not to use a subject in the first place so you can avoid having to invoke it imperatively and risk introducing subtle threading bugs in your queries.

Correct Examples

As shown in this post, there's only 1 particular set of conditions in which it's absolutely appropriate to use subjects.  Here are some examples.

Reactive Property

A common example of when it's correct to use a subject is to represent a sequence of property changes for a particular property, exposing it with AsObservable to hide its identity.  It's similar to implementing the PropertyChanged event, but only raising it for a particular property.  Instead of raising an event from within the property's setter, you'd call subject.OnNext(value).

Subject<T> works fine, though more commonly BehaviorSubject<T> is used instead because it stores the latest value of the property and pushes it immediately to new observers.  If your program is highly reactive, then you may find that you don't even need to keep a backing field for the property since BehaviorSubject<T> encapsulates it.  To set the value, call subject.OnNext(value).  Unfortunately, there's no way to extract the value imperatively (at least not until Rx exposes a Current property), so you may find that a backing field is still necessary unless your program is highly reactive.  Subscribing to the subject is the natural way of extracting the current value and all subsequent changes.

using System;
using System.Reactive.Linq;
using System.Reactive.Subjects;

namespace Rx.Labs
{
    class SubjectAsProperty
    {
        public static void Main()
        {
            var foo = new Foo();

            Console.WriteLine("Normal property usage:");

            Console.WriteLine(foo.Number);
            foo.Number += 8;
            Console.WriteLine(foo.Number);

            Console.WriteLine("Reactive property usage:");

            using (foo.NumberObservable.Subscribe(n => Console.WriteLine("Observer A: {0}", n)))
            {
                foo.Number = 1;
                foo.Number = 2;

                using (foo.NumberObservable.Subscribe(n => Console.WriteLine("Observer B: {0}", n)))
                {
                    foo.Number = 3;
                }

                using (foo.NumberObservable.Subscribe(n => Console.WriteLine("Observer C: {0}", n)))
                {
                    foo.Number = 4;
                }
            }

            Console.WriteLine(foo.Number);
            Console.ReadKey();
        }
    }

    class Foo
    {
        public int Number
        {
            get
            {
                return number.FirstAsync().Wait();
            }
            set
            {
                number.OnNext(value);
            }
        }

        public IObservable<int> NumberObservable
        {
            get
            {
                return number.AsObservable();
            }
        }

        private readonly BehaviorSubject<int> number = new BehaviorSubject<int>(42);
    }
}

Output:

Normal property usage:
42
50
Reactive property usage:
Observer A: 50
Observer A: 1
Observer A: 2
Observer B: 2
Observer A: 3
Observer B: 3
Observer C: 3
Observer A: 4
Observer C: 4
4

Cross-Cutting APIs

It's common to use a subject to represent cross-cutting concerns in a reactive manner.  For example, imagine exposing a property on a purely memory-bound logging class, returning an observable that pushes a notification whenever something is logged, such as entering and exiting methods (possibly via AOP).  This could be useful for real-time performance analysis, reactive heuristic-based adjustments, post-mortem debugging and even real-time debugging.  You could have one observer writing notifications to disk and one writing them to a GUI console, while another adjusts algorithms at runtime based on performance heuristics.  ReplaySubject<T> could be used to buffer notifications in memory and replay them to new observers, if necessary.

Another example would be a sequence of alerts exposed as an observable property on an Alert class.  Various parts of the program could send alerts concurrently by calling Alert.Unimportant(...) or Alert.RequiresUserInteraction(...), for example.  One observer could send emails while another displays toast notifications.

using System;
using System.Reactive.Linq;
using System.Reactive.Subjects;

namespace Rx.Labs
{
    class AlertLab
    {
        public static void Main()
        {
            var toast = from alert in Alert.All.OfType<ToastAlert>()
                        select alert.Message;

            var email = Alert.All.OfType<EmailAlert>();

            using (toast.Subscribe(a => Console.WriteLine("Important: {0}", a)))
            using (email.Subscribe(a => Console.WriteLine("Email \"{0}\" with \"{1}\".", a.To, a.Message)))
            {
                new ToastAlert("Hello Console!").Send();
                new EmailAlert("Hello Fred!", "fred@example.com").Send();
                new ToastAlert("Goodbye!").Send();
            }
        }
    }

    public abstract class Alert
    {
        public static IObservable<Alert> All
        {
            get
            {
                return alerts.AsObservable();
            }
        }

        public string Message
        {
            get
            {
                return message;
            }
        }

        private static readonly ISubject<Alert, Alert> alerts = Subject.Synchronize(new Subject<Alert>());
        private readonly string message;

        protected Alert(string message)
        {
            this.message = message;
        }

        public void Send()
        {
            alerts.OnNext(this);
        }
    }

    public class ToastAlert : Alert
    {
        public ToastAlert(string message)
            : base(message)
        {
        }
    }

    public class EmailAlert : Alert
    {
        public string To { get; private set; }

        public EmailAlert(string message, string to)
            : base(message)
        {
            this.To = to;
        }
    }
}

Output:

Important: Hello Console!
Email "fred@example.com" with "Hello Fred!".
Important: Goodbye!

Conclusion

If you've got anything that can be converted into an observable, such as a Task or Task<T>, an event, or an enumerable, then convert it using a From* or To* operator such as these or this, or if you've already got an observable, then use it.  Now consider temperature conflicts, change the temperature if necessary via Defer or Publish, and finally query it.  Don't use subjects.

Else if you don't have an observable or anything that can be converted into one:

If you require a cold observable, then use an existing Rx generator method; e.g., Return, Range, Timer, Interval, Generate or Create.  Don't use subjects.

Else if you require a hot observable:

If the scope of your observable is a method (i.e., it's entirely encapsulated by a method), then use an existing Rx generator method.  You can also apply Publish and RefCount, but consider relaxing the conditions to simply return a cold observable.  Don't use subjects.

Else if the scope of your observable is a type (e.g., it's exposed as a public property and backed by a field):

If you need to define a similar event or a similar event already exists, then convert the event into an observable like the first case.  Don't use subjects.

Else if you don't need to define a similar event and no similar event already exists, then use a subject.

In summary:

When should I use a subject?

When all of the following are true:

  • you don't have an observable or anything that can be converted into one.
  • you require a hot observable.
  • the scope of your observable is a type.
  • you don't need to define a similar event and no similar event already exists.

Why should I use a subject in that case?

Because you've got no choice!

 


  1. At least that's my interpretation of Erik's comments.  I find it to be really good advice in practice when working with Rx.
  2. Perhaps in very advanced and rare scenarios you may find a custom ISubject<T> implementation to be useful, in which case it's safe to use Multicast rather than Publish.
  3. Which happens to be an easy-to-miss bug when implementing custom operators in Rx.  If you need to return a cold observable, as generally operators do, and you must use a closure, then wrap the closure and the call to Create in a call to Defer to ensure that the closure isn't shared across multiple subscriptions.
  4. In some advanced and rare scenarios you may want to elide this contract and permit concurrent notifications through a subject, for performance reasons, though you should do so only if you can guarantee that none of Rx's query operators will be applied to the subject and all subscribers are thread-safe.

Tags:

.NET | Patterns | Rx

June 04, 2013

Bye Google Reader, Hello Windows RSS Platform

With the impending demise of iGoogle and Google Reader, which have been set to my web browser's home page for several years now, I had recently come to the realization that I was going to open IE one day and see an error message where iGoogle used to be.  All of my subscribed RSS feeds would be gone.

I did some research to see what kinds of replacements were out there.  For me, RSS feeds provide topics of particular interest that may be worthy of exploring on the web, and my web browser's home page is the gateway to the web; therefore, I wasn't interested in using feed-reader software that runs outside of my browser.  Feeds contain hyperlinks that lead me into the browser, so I might as well read my feeds right on the home page.

It seemed appropriate to begin my research by revisiting the Feeds tab in IE 10.  I had thought perhaps there have been improvements that could make it a viable replacement for Google Reader.  It didn't take long before I realized why I had chosen Google Reader several years ago and never switched to IE's built-in RSS feed support.  Alas, since I have no choice now but to change, I decided to research deeper into this feature of IE and see whether it may be worth using.  There were a few things I really liked about it:

  1. IE automatically detects feeds on web pages and lets you subscribe to them easily via a button on its Command Bar.
  2. IE opens feeds in a special reader allowing you to easily search, sort and filter.  It also provides a link to quickly subscribe or unsubscribe.
  3. IE lets you easily organize your feeds into folders.

I'm sure you'd agree that the integrated experience is a huge plus; however, IE also has a few negatives:

  1. IE doesn't provide a "What's New?" kind of view; i.e., an aggregated view containing all of the latest unread items among all feeds.
  2. IE doesn't provide a way to quickly view a list of items in a selected feed without showing their contents.
  3. IE doesn't provide a way to mark a single item as read/unread.

So it seems that IE has a great story for subscribing to and organizing feeds, though it really lacks in support for interacting with feeds.  The latter is extremely important to me, which is why I had originally chosen iGoogle as my home page for its Google Reader widget.

But the story doesn't end there.  Digging deeper I discovered that IE builds its RSS features on top of a public Windows API called the Windows RSS Platform.   IE merely provides a GUI containing only a subset of the functionality offered by this platform.  That was great news!  I decided to write my own web page against this API.  I could set it as my home page and interact with my RSS feeds easily, similar to the Google Reader experience, while taking advantage of the integration and organizational features built-in to IE.

These were my primary goals:

  1. Implement a stand-alone HTML page.  No server, no external scripts, no external style sheets, no external images.
  2. Target HTML5 in IE 10 (desktop mode only) for Windows 8.
  3. Touch input support.
  4. Use JavaScript to interact with the Windows RSS Platform APIs via Window's built-in ActiveX automation.
  5. Display 3 synchronized panels:
    • Tree view shows all feeds organized into folders, similar to the Feeds tab in IE.
    • Aggregate view containing all of the latest unread items; i.e., "What's New?"
    • Content viewer for the selected item.

I'm happy to report that my project was a success.  I worked on it all last weekend.  I even added several usability features that will hopefully make it appealing to others.  It relies heavily on jQuery, which is minified and embedded into the file.

To the right, you'll see a screenshot of the page in touch-screen mode.

Editing and organizational features weren't required thankfully due to IE's existing integration; e.g., the feed button on the Command Bar and the Feeds tab in the so-called Favorites Center provide all that is needed to subscribe and unsubscribe feeds, organize feeds into folders and configure feed update intervals and cache settings.  Once you've subscribed to a feed in IE, my page will automatically show it and aggregate the latest items into the center view.  To update the page, simply refresh it by pressing F5.

Download the finished product (as is, without warranty, use at your own risk, etc. etc.) from the following link.  It's a single minified HTML file named "rss.htm", approx. 108 KB in size.  Feel free to link to this blog post, but please do not link to the file directly and do not host the file publicly.

(EDIT: This file is now part of an open source project named RubyFeed.  I've changed the URL below to point to the v1.0 release of RubyFeed, which corresponds to the exact version of the file that was previously linked here.  However, I recommend downloading the latest release instead.)

https://rubyfeed.codeplex.com/releases/view/129236

(MD5 = e7d93aab06276614b917ea68369c1ac9)

Please let me know if you find it useful.  If you have any ideas for new features or would like to report bugs then let me know in the comments of this post.  I'm also considering publishing the project as open source if anyone's interested in contributing to it.

Installation

No installation is required.  If you're running Windows 8 and IE 10 then all of the components that you need are already installed on your computer.

The simplest way to get started is to copy the HTML file anywhere onto your computer and then double-click it (assuming that IE is configured as your default web browser).  You can also set it as your home page using the file:/// protocol.  For example, if you copy the file to your C: drive then you could set your home page to file:///c:\rss.htm.

WARNING: Since the page uses ActiveX you may have to deal with some security restrictions in IE, which may cause annoying prompts or, in the worst case scenario, IE will entirely block the page from loading.  Unfortunately, I haven't found a way to avoid the prompts when loading the file directly from the local file system with the file:/// protocol.

Alternatively, the best solution for me was to host the page locally and relax some of the security requirements in the Local Intranet security zone.  I was able to set my homepage to http://localhost/ and I'm no longer prompted with any security warnings.  Since I'm already running IIS locally, this was a piece of cake - it was as simple as copying the file to my root web directory and renaming it to default.htm.  Hunting down the appropriate security settings in IE was a bit more complicated, so I embedded the instructions directly into the file itself.  Open the page and when prompted with an ActiveX security warning simply choose to disallow the control to load; after a second or two the page will provide step-by-step instructions on how to disable the prompts.  Note that you must not disable scripts from running; otherwise, the instructions won't be shown.

Configuration

The page references a favicon.ico file in the same directory, though it's not required to actually exist.  If it does exist, then IE will display the icon when you browse to the page.  Feel free to create your own icon if you want.  Name it favicon.ico and copy it to the same directory as the HTML page.

The page also contains several configurable settings.  Open the file in a text editor such as notepad and look at the first JavaScript block near the top of the file.  It defines several variables that you can change to suit your needs, just don't delete any of them.

Happy RSS surfing!

Tags: ,

newsgroups | Web | WWW

November 23, 2012

Visual Studio Settings Switcher

I just published the first release of my Settings Switcher extension for Visual Studio 2012 on CodePlex.

Yesterday, I was looking for a quick solution to switch between Visual Studio settings files (.vssettings) without using the Tools > Import and Export Settings… dialog.  The dialog’s great, but it’s a bit slow to use a wizard every time I want to switch settings.  Sometimes I like to jump back and forth between different solutions quickly during the day, like when I’m working on various open source projects with different code formatting requirements.  I wanted a quicker way to load settings after opening a solution.

I found this blog post, which had a link to a tip on Sara Ford’s blog showing how easy it is to create a command button in the IDE to switch between settings by writing a simple macro.

Unfortunately, Visual Studio 2012 doesn’t support macros.

So I decided to create a Visual Studio extension.  The original premise was simply to show a drop-down list containing all of the .vssettings files and when one is selected those settings are applied.  That turned out to be a useful feature of the extension; however, while I was developing it I realized that I could add a few more useful features.  For example, I added a button to format every code file in the solution according to the currently selected settings.

But the primary feature, IMO, is that when you close a solution or exit Visual Studio, the extension saves a reference to the current settings file in the solution’s user options file (.suo).  The next time that you open that solution, the associated settings are automatically applied by the extension.  “Automate everything” is my motto :)

This extension should be useful to open source developers.  Typically we work on projects across multiple teams, with different code formatting requirements.  The next time you fork a project you can simply open the solution, edit your settings in the normal Tools > Options… dialog to meet the requirements of the team, then export the settings to a new file using the Export Current Settings button on the Settings Switcher toolbar provided by the extension.  And that’s it!  Every time you open that particular solution, its associated settings are applied automatically.

Project coordinators should also consider checking in a .vssettings file to their repositories.  This will make it easier for interested developers to set up their environment for working on your project.  A developer would simply need to copy the .vssettings file that you provided to the directory that Visual Studio uses to export settings files on their system (Tools > Options > Environment > Import and Export Settings).  Then follow the instructions above to associate the settings with the solution.

I’ve got some additional features that I plan on adding in the future, time permitting.  Some of them are very community-oriented.  We’ll see how that goes.

Feedback is appreciated.  Feel free to request a feature, report a bug or start a discussion.