PeanutButter.RandomValueGen: the builder pattern & random generation for testing purposes

Retrieving the post... Please hold.

If the post doesn't load properly, you can check it out here: 
https://github.com/fluffynuts/blog/blob/master/20180825.md

#PeanutButter.Utils: the dictionaries

Dictionaries, HashMaps, whatever you want to call them -- they can be one of the most useful constructs in any language. In Javascript, the dictionary interface to objects makes a lot of dynamic code simpler. In fact, it was the Javascript paradigm which served as the inspiration for `PeanuButter.Utils` member: `DictionaryWrappingObject`. This class was originally made to facilitate some of the functionality in `PeanutButter.Ducktyping`: a library for duck-typing arbitrary objects onto interfaces which aren't directly related. For example, if you had an anonymous object with the correct "shape", you could duck-type it onto an interface which another part of the system requires without having to manually create your own implementation of that interface and the code to copy data / forward method calls -- all of which `PeanutButter.DuckTyping` can do, with varying amounts of flexibility to matching methods and properties, according to your needs.

But I'm not here to talk about `PeanutButter.DuckTyping` today -- as interesting as it was to write and as useful as it's proven to be in a couple of projects since then.

So, back to `DictionaryWrappingObject`: this class provides the familiar `IDictionary<string, object>` interface over any other object so you can enumerate through the properties or perform functions like querying property names without directly using reflection yourself. And, of course, you can just use very Javascript-y syntax:
```csharp
var obj = new { id = 1, name = "bob" };
var wrapper = new DictionaryWrappingObject(obj);
var name = wrapper["name"];
```
You can also construct with a case-insensitive StringComparer to make those lookups a little fuzzier:
```
var obj = new { Id = 1, Name = "Mary" };
var wrapper = new DictionaryWrappingObject(obj);
var id = wrapper["ID"];
var name = wrapper["name"];
```

Another useful dictionary construct that I stumbled across in Python is the default dictionary -- implemented in `PeanutButter.Utils`, unsurprisingly, as `DefaultDictionary`. This dictionary allows you to specify a default value to return for the case where requested keys aren't found:
```
var animalsInZoo = new DefaultDictionary<string, bool>(false);
animalsInZoo["Camel"] = true;
animalsInZoo["Panda"] = true;

//... some time later:
var haveCamels = animalsInZoo["Camel"]; // true
var haveSnakes = animalsInZoo["Snake"]; // false, not KeyNotFoundException!
```

`DefaultDictionary` can be a little smarter than having a static value for the default:
```
// set up the default dictionary such that students with a name starting
// "A" exist.
var students = new DefaultDictionary<string, bool>(
  k => k.StartsWith("A")
);
students["Anna"] = false;
students["Mary"] = true;
// ... elsewhere ...

var haveAnna = students["Anna"]; // false, explicitly set
var haveAndrew = studentds["Andrew"]; // true: default value provider
var haveMary = students["Mary"]; // true: explicitly set
var haveStewart = students["Stewart"]; // false: default value provider
```

`DefaultDictionary` is expecially useful in conjunction with `MergeDictionary`, which takes one or more other dictionaries with the same key/value types and merges them, returning the value from the first in the merge list to have a value. So we could have:
```
var config = new Dictionary<string, string>()
{
  ["host"] = "database-machine",
  ["port"] = "123"
};
var defaults = new Dictionary<string, string>(k =>
{
  switch (k):
  {
    case "host":
      return "localhost";
    case "port":
      return "3306";
    case "user":
      return "mysql";
    case "password":
      return "super-secret";
    default:
      return "";
  }
});
var final = new MergeDictionary<string, string>(
  config, defaults
);

// ... elsewhere ...
var config = new 
{
  host: final["host"],        // database-machine
  port: int.Parse(final["port"]), // 123
  user: final["user"],        // mysql (default)
  password: final["password"] // super-secret (default)
};

```

Finally, there is the `CaseWarpingDictionary`, which basically acts as a wrapper for another dictionary to change the case-sensitivity of the keys, especially useful when you want a dictionary that is a little more forgiving (case-insensitive) than the one you're working with. `CaseWarpingDictionary` can be constructed with either a boolean instructing whether or not the result is case-sensitive, or with a StringComparer, so you can, for instance, switch from `Ordinal` to `CurrentCultureIgnoreCase`.

## Last words
The `IDictionary<TKey, TValue>` interface is not particularly difficult to implement, but it is quite convenient to consume. And I recommend devs wanting to learn something about the internals of .NET to implement some kind of `IDictionary<TKey, TValue>` at some point in their lives. For one thing, it will give you an appreciation for the good old `Dictionary` (:

# Unit test coverage and why it matters

Good unit test coverage can help you to improve your code in ways you might not expect. I'm not talking about just chasing some mythical value, like the agile team which ascribes to the contract of "85% coverage over the project", though chasing a number (the theoretical 100% coverage, which I've only achieved in one project ever) can lead you to some interesting discoveries.

Obviously, we can have bogus coverage:

```csharp
using NUnit.Framework;
using static NExpect.Expectations;
using NExpect;

[TestFixture]
public class TestTheThing
{
  [Test]
  public void ShouldDoTheStuff()
  {
    // Arrange
    var sut = new TheThing();
    // Act
    var result = sut.DidTheStuff();
    // Assert
    Expect(result).To.Be.True();
  }
}

public class TheThing
{
  public bool DidTheStuff()
  {
    try
    {
      WriteOutSomeFile();
      MakeSomeWebCall();
      return true;
    }
    catch (WebException ex)
    {
      return false;
    }
  }
  // ... let's imagine that WriteOutSomeFile 
  // and MakeSomeWebCall are defined below...
}
```

The above test doesn't actually check that the correct file was written out -- or even that the correct web request happened. The one test above technically provides full coverage of the class (if WriteOutSomeFile and MakeSomeWebCall have no branches), but it's a bit anemic in that the coverage doesn't tell us much.

So coverage is not a number which definitively tells you that your code (or tests) are good. However, examining coverage reports (particularly the line-by-line analysis) has helped me to discover at least three classes of error:

## I found bugs I didn't know I had

When I was still at [Chillisoft](http://www.chillisoft.co.za), I'd just finished a unit of code (TDD, of course) to my satisfaction and decided to run coverage on it for interest sake. I was convinced that I'd done a good job, writing one test before each line of production code which was required. To my dismay, I found that there was one line which _wasn't_ covered. _Shame on me_, I thought, and went back to the test fixture, _where I found a test describing the exact situation that line should be handling_. Ok, so I have this test, it names the situation, but there's no coverage on the line?

_Remember: coverage reports _can_ be faulty. It's rare, but it's worthwhile re-running your reports to just make double-sure that what you're seeing is correct._

I re-ran my coverage, but that one line remained red. And the more I looked at it, the more it looked like it should actually be causing a test to fail. So I re-examine the test which is supposed to be covering it to find... I've made a mistake in that test and it's actually not running through the branch with the uncovered line. 

The fault here most likely comes down to one faulty TDD cycle where I hadn't gotten a good "red" before my "green". Still, examining the coverage report made me find the error and fix it before the code got anywhere near production. This experience is why I advocate for running coverage after reaching a point where one expects the current unit of work to be complete -- to find any holes in testing or defects in logic hidden in those holes. It's why I (convincingly) argued for all Chillisoft programmers to be granted an Ultimate license of Resharper, which has dotCover built right in to the test runner. We have coverage reports running at the CI server, but I wanted every developer to have the ability to test coverage quickly so that they could also discover flaws in their code before that code gets to production -- or even another developer's machine!

## I found dead code

Just recently, I finally got the `gulp` build tasks for `NExpect` to include coverage by default when running `npm test`. And to my dismay, `NExpect` only had about 78% coverage. Which I thought was odd, because `NExpect` was build very-much test-first: indeed, the general method of operation was to write out a test with the desired expression and then provide the code to make that expression happen. So, for example:

```csharp
Expect(someCollection).To.Contain
  .Exactly(1).Deep.Equal.To(search);
```

would have started out with most of those words red (thanks to Resharper) and they would unhighlight as I got together the class/interface linkage to make the words flow as desired. I expected coverage to be closer to 90% (I did expect some places to have been missed in lieu of ever having scrutinised coverage reports for `NExpect` before), but 78%? I had some work to do.

In addition to finding a few minor bugs that I didn't know I had (particularly with wording of failure messages in a few cases), I found that I had bits of code which theoretically should have been under test, but which weren't covered. Especially stuff like:

```csharp
[Test]
public void ComparingFloats()
{
  Expect(1.5f).To.Be.Greater.Than(1.4f)
    .And.Less.Than(1.6f);
}
```

which works as expected, but never hit the `Than` extension methods for continuations of float.

The answer became obvious upon hovering over the usages -- each `Than` was expecting to operate on values of type `double` as the _subject (actual) value_ (ie, the value being tested, in this case, `1.5f`). This is because `Expect` upcasts certain numeric types (floats to doubles, ints to longs, etc) so that the comparison code doesn't require casting from the consumer (since `NExpect` continuations hold the type of the subject all the way through, instead of downcasting to `object` and hoping for the user to provide reasonable values. There's nothing wrong (that I can tell) with this approach -- and it works well, but it _does_ mean that the `Than` extension methods expecting to operate on `float` and `int` subjects will never be used. They were dead code! So I could safely remove them. One of the ways to make code better is to remove the unnecessary bits (:

## I found holes in my api

This is again, working in `NExpect`, where, upon providing coverage for one variant of syntax, I would find that I hadn't implemented for another. For example, `NExpect` has no opinion on which of these is better:

```csharp
Expect(1).To.Not.Equal(2);
Expect(1).Not.To.Equal(2);
```

All `NExpect` does is prevent silliness like:

```csharp
Expect(1).Not.To.Not.Equal(1);
```

`NExpect` is designed around user-extensibility as one of the primary goals. As such, there are some "dangling" words, like `A`, `An`, and `Have` so
that the user can provide her own expressive extensions:

```csharp
var dog = animalFactory.CreateDog();
Expect(dog).Not.To.Be.A.Cat();
Expect(dog).To.Be.A.Dog();
Expect(dog).To.Be.A.Mammal();
```

Where the user can use Matchers or Composition to provide the logic for the `Cat`, `Dog`, and `Mammal` extension method assertions. `NExpect` doesn't actually provide extensions on these "danglers" -- they're literally just there for custom extension.

Whilst running coverage, I found that one variant of `Contain` wasn't covered, and when I wrote a test to go through all three (positive, negative, alt. negative), I found that there were missing implementations! Which I naturally implemented (:

## Using coverage to make your code better

Coverage reports like those generated by `dotCover` and the combination of `OpenCover` and `ReportGenerator` can not only give you confidence in your code and a fuzzy feeling inside at a number which shows that you do care about automated testing for your code -- they can also help you to make your code (and tests) better. And make _you_ better, going forward, because you learn more about the mistakes you make along the way.

If you want to get started relatively easily and you're in the .net world, you can use [gulp-tasks](https://github.com/fluffynuts/gulp-tasks) as a submodule in your git repo. Follow the instructions in the `start` folder and get to a point where you can run `npm run gulp cover-dotnet` (or make this your `test` script in `package.config`). This should:
- build your project
- run your tests, using `OpenCover` and `NUnit`
- generate html reports using `ReportGenerator`, under a `buildreports` folder

You can always check out [NExpect](https://github.com/fluffynuts/NExpect) to see how I get it done there (:

# What's in PeanutButter.Utils, part 2

## Metadata extensions

I just wanted to chip away at my promise to explain more of the bits in PB, so I thought I'd pick a little one (though I've found it to be quite useful): metadata extensions.

At some point, I wanted to be able to attach some arbitrary information to an object which I didn't want to extend or wrap and which some code, far down the line, would want to read. If C# was Javascript, I would have just tacked on a property:
```js
someObject.__whatDidTheCowSay = "moo";
```

But C# is _not_ Javascript. I could have maintained some global `IDictionary` somewhere, but, even though I wanted it to support a feature in [NExpect](https://github.com/fluffynuts/NExpect), where the code wouldn't have a running lifetime of any significance, it still felt like a bad idea to keep hard references to things within NExpect. The code associating the metadata has no idea of when that metadata won't be necessary any more -- and neither does the consumer.

Then I came across [`ConditionalWeakTable`](https://msdn.microsoft.com/en-us/library/dd287757(v=vs.110).aspx) which looked very interesting: it's a way of storing data where the keys are weak references to the original objects, meaning that if the original objects are ready to GC, they can be collected and the weak reference just dies. In other words, I found a way to store arbitrary data referencing some parent object and the arbitrary data would only be held in memory until the end of the lifetime of the original object.

That's exactly what I needed.

So was born the [`MetadataExtensions`](https://github.com/fluffynuts/PeanutButter/blob/master/source/Utils/PeanutButter.Utils/MetadataExtensions.cs) class, which provides the following extension methods on all objects:

- `SetMetadata`<`T`>`(string key, object value)`
- `GetMetadata`<`T`>`(string key)`
- `HasMetadata`<`T`>`(string key)`

which we can use as follows:

```csharp
public void MethodWantingToStoreMetadata(
  ISomeType objectWeWantToStoreStateAgainst)
{
  objectWeWantToStorStateAgainst
    .SetMetadata("__whatDidTheCowSay", "moo");
}

// erstwhile, elsewhere:

public void DoSomethingInterestionIfNecessary(
  ISomeType objectWhichMightHaveMetadata)
{
  if (objectWhichMightHaveSomeMetadata
        .HasMetadata("_whatDidTheCowSay"))
  {
    var theCowSaid = objectWhichMightHaveSomeMetadata
                       .GetMetadata("_whatDidTheCowSay");
    if (theCowSaid == "moo")
    {
      Console.WriteLine("The cow is insightful.");
    } 
    else if (theCowSaid == "woof")
    {
      Console.WriteLine("That ain't no cow, son.");
    }
  }
}
```

And, of course, as soon as the associated object can be collected by the garbage collector (remembering that the reference to this object, maintained within PB, is _weak_), that object is collected and the associated metadata (if not referenced elsewhere, of course) is also freed up. This mechanism has facilitated some interesting behavior in `NExpect`, and I hope that it can be helpful to others too.

# Markdown all things

Whilst blogger.com provides a fairly good blogging platform for regular writing, I've found that it's rather painful for technical blogging. In particular, code blocks are a mission. In the past, I've wrapped code in <pre><code> ... </code></pre> and let [highlight.js](https://github.com/isagalaev/highlight.js/) do all the heavy lifting of making that actually look readable. HighlightJs has been fantastic at that, but it still hasn't been as smooth a process as I would have liked _overall_. I still tended to write the non-code parts in the WYSIWIG html editor, and had to switch the source view to work on code parts. 

When I blog, I literally want to get out the information as quickly as possible, in a readable format. I'm not here to fight with styling.

So I was quite happy to stumble across [showdown](https://github.com/showdownjs/showdown). A little Javascript in my template and suddenly I could write in possibly the simplest format ever: markdown. I had quick and easy access to simple styling elements (lists, headings, etc) as well as code blocks. All good, but not automagick out of the box. 

I thought to myself, _"I'm sure I can't be the only person who wants this"_, and _"It would be nice if that auto-bootstrapping of markdown+code could be done anywhere, not just from within my blogger template"_.

So, as is so common within the open-source world, I stand upon the very tall, very broad shoulders of [highlight.js](https://github.com/isagalaev/highlight.js/) and [showdown](https://github.com/showdownjs/showdown) to present [auto-markdown](https://github.com/fluffynuts/auto-markdown): a script you can include on any page to convert any element with the `markdown` class to be rendered as markdown. It can even be configured (script versions and code theme) via some global variables, so you don't have to fiddle with the code if you don't want to.

I trialed it with my last post and It's how I'm writing now -- I just add a shell `pre` tag with the `markdown` class and get on with the writing, without any more fighting with the html editor. As a bonus: even if the script fails for some reason (such as if the user has Javascript disabled or GitHub doesn't supply my script in time), the blog is still in a readable format: markdown.

If you're interested, follow the instructions in the [README.md](https://github.com/fluffynuts/auto-markdown). Feel free to open issues if you encounter some - for instance, I encountered some stickiness with generics in code blocks. Also, bear in mind that markdown requires html-escaping for chevrons (ie, embedding xml). 

Feel free to share it as much as you like. If you don't feel comfortable referencing my code directly, fork my repo and keep your own copy (:

Now, if only blogger's html editor had a vi mode...

# What's in `PeanutButter.Utils`, exactly?

`PeanutButter.Utils` is a package which pretty-much evolved as I had common problems that I was solving day-to-day. People joining a team that I was working on would be exposed to bits of it and, like a virus, those bits would propagate across other code-bases. Some people asked for documentation, which I answered with a middle-ground of `xmldoc`, which most agreed was good enough. People around me got to know of the more useful bits in `PeanutButter.Utils` or would ask me questions like "Does PeanutButter.Utils have something which can do [X]?". I kind of took the ubiquity amongst my team-mates for granted.

Fast-forward a little bit, and I've moved on to another company, where people don't know anything about the time-savers in `PeanutButter.Utils` -- and it occurs to me that that statement probably applies to pretty-much most people -- so I thought it might be worthwhile to have some kind of primer on what you can expect to find in there. An introduction, if you will. I think there's enough to break the content down into sections, so we can start with:

## Disposables
One of the patterns I like most in the .net world is that of `IDisposable`. It's a neat way to ensure that something happens at the end of a block of code irrespective of what goes on _inside_ that code. The code could throw or return early -- it doesn't matter: whatever happens in the `Dispose` method of the `IDisposable` declared at the top of a `using` block will be run. Usually, we use this for clearing up managed resources (eg on database connections), but it struck me that there were some other convenient places to use it. Most generically, if you wanted to run something simple at the start of a block of code and run something else at the end (think of toggling something on for the duration of a block of code), you could use the convienient `AutoResetter` class:
```csharp
using (new AutoResetter(
    () => ToggleFeatureOn(), 
    () => ToggleFeatureOff()))
{
    // code inside here has the feature toggled on
}
// code over here doesn't -- and the feature is 
//    toggled back off again even if the code 
//    above throws an exception.
```
It's very simple -- but it means that you can get the functionality of an `IDisposable` by writing two little lambda methods.

You can also have a variant where the result from the first lambda is fed into the second:
```csharp
using (new AutoResetter(
    () => GetCounterAndResetToZero(),
    originalCount => ResetCounterTo(originalCount)))
{
// counter is zero here
}
// counter is reset to original value here
```

Cool.

Other common problems that can be solved with `IDisposable` are:
### Ensuring mutexes / semaphores are reset, even if an exception is encountered
For this, we can use `AutoLocker`:
```csharp
using (new AutoLocker(someMutex))
{
}
using (new AutoLocker(someSemaphore))
{
}
using (new AutoLocker(someSemaphoreLite))
{
}
```

### Temporary files in tests
```csharp
using (var tempFile = new AutoTempFile())
{
   File.WriteAllBytes(
       Encoding.UTF8.GetBytes("moo, said the cow"),
       tempFile.Path
   );
   // we can run testing code against the file here
}
// file is gone here, like magick!
```
This uses the `Path.GetTempFileName()` system call by default -- so you don't have to care about where the file actually exists. Of course, there are constructor overloads to:
- create the file populated with data (string or bytes)
- create the file in a different location (not the system temp data location)
- create the file with a specific name

`AutoTempFile` also exposes the files contents via properties:
- `StringData` for string contents
- `BinaryData` for a byte[] array

There is also an `AutoTempFolder` if you want a scratch area to work in for a period of time. When it is disposed, it and all it's contents are deleted.

Similarly, `AutoDeleter` is an `IDisposable` which can keep track of multiple files you'd like to delete when it is disposed:
```csharp
using (var deleter = new AutoDeleter())
{
    // some files are created, then we can do:
    deleter.Add("C:\Some\File");
    deleter.Add("C:\Some\Other\File");
}
// and here, those files are deleted. If they can't be 
//    deleted, (eg they are locked by some process),
//    then the error is quietly suppressed. 
//    `AutoDeleter` works for folders too.
```

### Other disposables
As much as I love the `using` pattern, it can lead to some "arrow code", like this venerable ADO.NET code:
```csharp
using (var conn = CreateDbConnection())
{
    conn.Open();
    using (var cmd = conn.CreateCommand())
    {
        cmd.CommandText = "select * from users";
        using (var reader = cmd.ExecuteReader())
        {
            // read from the reader here
        }
    }
}
```
Sure, many people don't use ADO.NET "raw" like this any more -- it's just an easy example which comes to mind. I've seen far worse "nest-denting" of `using` blocks too.

This can be flattened out a bit with `AutoDisposer`:
```csharp
using (var disposer = new AutoDisposer())
{
    var conn = disposer.Add(CreateDbConnection());
    conn.Open();
    var cmd = disposer.Add(conn.CreateCommand());
    cmd.CommandText = "select * from users";
    var reader = disposer.Add(conn.ExecuteReader());
    // read from the db here
}
// reader is disposed
// cmd is disposed
// conn is disposed
```
`AutoDisposer` disposes of items in reverse-order in case of any disposing dependencies.


So that's part 1 of "What's in `PeanutButter.Utils`?". There are other interesting bits, like:
- extension methods to
    - make some operations more convenient
         - do conversions
         - working with `Stream` objects
         - DateTime utilities
    - facilitate more functional code (eg `.ForEach` for collections)
    - `SelectAsync` and `WhereAsync` let you use async lambdas in your LINQ
    - test and manipulate strings
- the `DeepEqualityTester`, which is at the heart of `NExpect`s `.Deep` and `.Intersection` equality testing
- `MemberExpression` helpers
- Reflection tidbits
- reading and writing arbitrary metadata for any object you encounter (think like adding property data to any object)
- Some pythonic methods (`Range` and a (imo) more useful `Zip` than the one bundled in LINQ)
- dictionaries
    - `DictionaryWrappingObject` lets you treat any object like you would in Javascript, with text property indexes
    - `DefaultDictionary` returns default values for unknown keys
    - `MergeDictionary` allows layering multiple dictionaries into one "view"
    - `CaseWarpingDictionary` provides a decorator dictionary for when the dictionary you have does indexing with inconvenient case rules

I hope to tackle these in individual posts (:

Toggle All Things!

(Note: this post was started quite a long time ago and has lived in draft mode whilst I intended to provide a concrete code example. I'm not sure when, if ever, that will come, but I think that the lessons learned are still valuable to share)

It's not an uncommon workflow: the client / company wants some features in a certain sprint and the features have to be signed off by a QA / testing team. So if we take the example of a two week sprint, what often happens is that the testers sit idle (or are tasked onto a different project) for the first week, while development happens at a furious pace, then testers are given domain to test the week's work and we enter a development freeze, where developers are unable to work ahead for fear of interrupting testing and must be on standby for any bugs picked up by QA.

Hopefully, if you're testing your own work proficiently, QA picks up nothing, or perhaps some small stuff like a misaligned label or similar. However, for that time period, you're stagnant -- which causes you to work harder when the first week of a sprint comes around again. It's not a healthy place to be in, if, for no other reason, than because it provides ammunition for the "production line" mindset and guilting you into working longer hours during the "productive" part of the sprint. And if you were on top of things, the business is also losing during the development freeze because you aren't free to forge ahead with new stories.

No-one is winning.

One possible strategy is to adopt a "feature branch per developer" strategy with a regular "merge day" where some poor soul attempts to merge the disparate work of the last two weeks. You can minimise the pain with more frequent merges during the active development phase, but whilst you're in development freeze, you're going to drift and merge hell is eventually inevitable.

One of the most useful strategies we've used to date is that of feature toggles, which, simply put, are mechanisms for turning parts of your software on or off. So here are some thoughts on that particular topic, played out over the development cycle of a SPA web application.

1. A simplistic approach: compile-time toggles (ie: releases with feature-sets)

One method is good old #if / #ifdef and compile-time constants. Bake those features in and release a version with a specified feature matrix. This works well for software where the required featureset is unlikely to change, such as for Gentoo packages, where your USE flags determine the features available to you at run-time until such time as you decide that you need more or fewer features. As the owner and maintainer of said Gentoo system, you decide what you want from your software and you compile it that way. When your requirements change, you update your flags and re-compile.

This works well when the user / tester is the developer. But not as well when the user/tester is someone else -- because then they have to contact the developer and request a rebuild and redeployment every time a feature needs to be turned on or off.

It's not an invalid strategy -- it's just not very agile. Additionally, you can't easily a automate testing (unit / integration testing) against features, because doing so requires changing compile parameters.

2. Better: static values in the code

Another version of this is compiling with static / constant values and branching in code. This allows testing against the different features (especially if you're using static values instead of constants). This alleviates the automated testing conundrum but still leaves the human testing department out in the cold. They will still contact you for a new build containing the required feature matrix.

Rebuilding and redeploying to satisfy a required feature matrix becomes impractical the more features you enable and the more agile you'd like to be. Suppose you think feature [X] is ready, but 1/2 an hour before sign-off, testing finds a defect. Now you have to rush to provide a deployment package without [X]. In addition, testing may not yet have had the time to prove that there are no unintended side-effects from not including feature [X].

3. Even better: run-time toggles determined by app / web configuration

By now you've realised that the toggles should be configurable and they have been pushed them into your web config (for example). Great! Toggles can be changed fairly quickly, without requiring a re-deployment. The problem is that you still need someone fairly technical (and who has access to the staging machines) to update the enabled feature-set.

My experience has been that if you can make the toggling of features relatively quick and painless, you can offload the decision on which features are in an accepted release to someone in testing or management. You don't want to waste your time and resources on fighting for feature inclusion / exclusion -- you want to focus on making new things, solving new problems! So whilst feature toggles in your web config are better than compile-time, you could push this just a little further for a lot of win.

4. Best: run-time toggles determined per user interaction

In the case of a web system, this would mean "per web request". Imagine if the testers had the ability to test just the stuff they were confident in, and, if they had time, could toggle on a beta feature and test if they have capacity. Imagine if, when a flaw is found in the implementation or design of a feature, it didn't have to hold up the entire release -- just toggle it off, continue testing other features and sign off what is available.

Our implementation was system-wide, re-evaluated per web request, but you could even push this further with per-user configuration included in request headers. Practically, this may be more effort than it's worth.

Features of good feature toggles

• It should be easy to toggle features and difficult to end up with a set of conflicting behaviors from the application. 
• It should be easy to determine the feature-set state of the application and difficult to be confused about interactions which cause experienced defects. I'm suggesting that if you have old feature [X] and new feature [Y], which replaces [X], there shouldn't be two toggles ("enable [X]" and "enable [Y]") -- whilst this technically solves the problem, you're expecting testers to understand the repercussions of the ability of enabling both. Instead, you should strive towards well-defined feature toggles which explain what they do and provide only two well-known states (the "off" and "on" states) which cannot be confused by interaction with other parts of the system.
• It should be as frictionless as possible to develop against (adding features and consuming them) to encourage using the framework so that the agility of feature toggles is experienced as often as possible. More toggles are better than fewer toggles, especially for non-inter-dependent features.
• Toggles should be easy to remove as well: once QA is satisfied with a feature, the toggle should be removed along with any branches of logic which disabled the feature.

This can all be achieved, and it's really not that tricky. Here are the steps we went through:

A. First pass: development freeze and the rush around sign-off time sucks. We need to be able to add new features which can be easily disabled!

So, first off, we define a class / interface which has boolean toggles on it. And we get the web app to put persist into our storage (sql, document db, whatever) on startup, adding toggles which are missing. We save three versions of this configuration, for three proposed feature-sets: "development", "staging", "production" and we let annotations on the configuration class determine default toggle states for features.

For example, we create a feature toggle with the feature defaulted on for "development" and off for "staging" and "production". In this way, developers on the team default to seeing the current work (and interacting with it) and the feature isn't inadvertently introduced to QA or the production servers. Of course, the moment QA is ready, they can manually enable the feature. Production only gets the feature enabled when a deployment to production is explicitly made with a script to enable the feature.

We allow selection of a feature set ("development", "staging", "production") in the web config, because this is unlikely to change on a deployment target, so probably won't need human interaction (let deployment scripts take care of this) and surface these toggles to the user via a simple UI. The UI itself is toggled with a feature toggle, so the production site doesn't have it at all -- no way for a user to enable a feature they shouldn't or disable a feature which should be there.

This process also makes adding / removing feature toggles trivial for other developers on the system.

Consuming toggles, server-side (back-end developer):
We could make the decision about defined feature classes to inject by manipulating the IoC container -- inject implementation "A" of interface "I" or, if the toggle is enabled, inject implementation "B". This may work with differing degrees of efficiency according to your IoC container and you can end up with two implementations which are incredibly similar, but which differ only in one minor aspect.

Alternatively, components of the system could, perhaps, request feature toggle configuration via IoC injection branch subtly as needed. We make sure that anything which needs feature toggles has a per-web-request lifetime (as does the feature toggles injectible) so that the web app doesn't have to be restarted to bring a toggle into effect

Consuming toggles, client-side (front-end developer):

The current feature toggles are exposed as a calculated stylesheet where disabled features literally have their display set to "none !important". So when the page loads, incomplete features are not available for interaction. On feature toggle, this stylesheet can be reloaded to get instant UI feedback of toggled features without even a page reload.

Toggling toggles (user):

Initially, toggling a feature simply toggles the boolean in storage and re-requests the featureset stylesheet, immediately updating the UI. Calls to the api get a new instance of the feature toggles entity and can change their behavior accordingly.

Typically, a "big hammer" approach works well: a controller action returns null or an empty collection when the toggle is off or calls into defined logic to produce a calculated result when the feature is enabled. Sometimes, a finer chisel is called into play -- some component further down the logic chain subtly changes behavior based on the toggle.

B. New features may replace older ones -- we need a toggle to turn one bit on and simultaneously turn another off

We already have this capability at the server -- we can branch code according to an injected feature toggles matrix configuration object -- but at the client, we are just hiding UI elements which expose disabled features.
Thus we added in more client-side logic: we expose a Javascript blob in a generated js file with the feature toggle matrix and emit an event when this is re-downloaded.
To ease developer consumption, a framework function is provided to tuck away the complexity of listening for the correct event.
Also, whilst we've already had calculated stylesheets to toggle incomplete features off, we can now add in calculated stylesheets to toggle fallback features (ie, legacy feature) back on again.

Winning at features

The net result of employing this tactic has meant that we've managed to pull out far in front of the testing team which used to drive deadlines. Simply put, we're able to work on sprint [N+1] whilst testing is signing off sprint [N], such that we've essentially ended up a sprint ahead and had the breathing-room to address technical debt -- a win for everyone!

Feature toggles provided:

• A way to alleviate stress for testers and developers: if a feature isn't deemed finished by sign-off time, it's simply toggled out for deployment
• A way to alleviate stress for developers: whilst features for this sprint are under human testing, developers don't have to sit idle -- they can move forward with the next set of requirements.
• A way to alleviate stress for project managers: it's far easier to say "we managed to get 4/5 of the required features deployed" than "we couldn't deploy because one feature couldn't be signed off"

Interactive, per-request feature toggles meant that:

• Testers could test at their own pace, enabling features on their environment when they were ready and disabling if they thought that those features were interacting negatively with others or if they considered them incomplete at sign-off time for the sprint.
• Testers can preview a feature during development to give early feedback to the developers if they have capacity
• "What-if" conversations could be had: "what if feature [X] was enabled for the users without [Y], which appears to be a little dodgy right now?"
• There was no rush at deployment time for getting "just the right build" to deploy: simply deploy the latest and toggle off the features which aren't considered fully-baked.

NExpect hits 100 releases!

It's probably not such a big deal, but it feels like it to me: NExpect (GitHub, Nuget) has hit its 100th release! Whilst the reason for the 100th release was trivial (adding support to assert the existence of keys by any case in a case-insensitive dictionary), it feels like some kind of milestone.

Of course, it wouldn't be here without the fantastic contributions from Cobus Smit as well as the input from everyone who uses it (even the developers on my team who had it foisted upon them!).

NExpect isn't done with its evolution though. I have a trello board that I probably should convert to GitHub issues. And I welcome bug reports and requests which align with the ethos of the project:

• Expect(NExpect).To.Be.Readable();
  • Because code is for co-workers, not compilers. And your tests are part of your documentation.
• Expect(NExpect).To.Be.Expressive();
  • Because the intent of a test should be easy to understand. The reader can delve into the details when she cares to.
• Expect(NExpect).To.Be.Extensible();
  • Because I can't predict every use-case. I believe that your assertions framework should enable expressive, readable tests through extension.

Personally, I'm finding NExpect to be a pleasure to use. But perhaps I'm a little biased (: