Saturday, 13 May 2017

Interview with about HTTP APIs in .NET

Next week I'll be in Russia, where I'm speaking about HTTP APIs and REST at the DotNext conference in Saint Petersburg. As part of this event, I've done an interview with the Russian tech site Хабрахабр about the history and future of API development on the web and in Microsoft.NET. The interview's available on their site (in Russian), but for readers who are interested but can't read Russian, here's the original English version.

Cathedral in Saint Petersburg, Russia. / Photo by DEZALB.

Q: What kind of APIs are you designing? Where does API design fit into software development?

That’s kind of an interesting question, because I think one of the biggest misconceptions in software is that designing APIs is an activity that happens separately to everything else. Sure, there are certain kinds of API projects – particularly things like HTTP APIs which are going to be open to the public – where it might make sense to consider API design as a specific piece of work. But the truth is that most developers are actually creating APIs all the time – they just don’t realise they’re doing it. Every time you write a public method on one of your classes or choose a name for a database table, you’re creating an interface – in the everyday English sense of the word – that will end up being used by other developers at some point in the future. Other people on your team will use your classes and methods. Other teams will use your data schema or your message format.

What’s interesting, though, is that once developers realize that the code they’re working on will probably form part of an API, they tend to go too far in the other direction. They’ll implement edge cases and things that they don’t actually need, just in case somebody else might need it later. I think there’s a very fine balance, and I think the key to that balance is to be very strict about only building the features that you need right now, but to make those things as reusable and self-explanatory as you can. There’s a great essay by Pieter Hintjens, Ten Rules for Good API Design, that goes into more detail about these kinds of ideas.

The biggest API project I’m working on at the moment is a thing I’m building at Spotlight in the UK, where I work. It’s a hypermedia API exposing information about professional actors, acting jobs in film and television, and various other kinds of data used in the casting industry. We’re building it in the architectural style known as REST – and if you’re not sure what REST is, you should come to my talk at DotNext in Saint-Petersburg and learn all about it. There’s lots of different patterns for creating HTTP APIs – there’s REST, there’s GraphQL, there’s things like SOAP and RPC – but for me, the biggest appeal of REST is that I think the constraints of the RESTful style lead to a natural decoupling of the concepts and operations that your API needs to support, which makes it easier to change things and evolve the API design over time.

Q: One of the most famous applications that was "killed" by backward compatibility is IE. The problem of this browser was that it has too large number of applications for which it was required to have backward compatibility. Problem was solved by adding new application Edge, which is updatable and supports all new standards. Can you give an piece of advice on how not to get caught by that backward-compatibility trap? As an example could it be a modularity which doesn't have layers? May be there is a way to replace API with RESTful API, Service Oriented Architecture or something else?

I’ve been building web applications for a long, long time – I wrote my first HTML page a couple of years before Internet Explorer even existed, back when the only browsers were NCSA Mosaic and Erwise. It’s fascinating to look back at the history of the web, and how the web that exists today has been shaped and influenced by things like the evolution of Internet Explorer – and you’re absolutely right; one of the reasons why Microsoft has introduced a completely new browser, Edge, in the latest versions of Windows is that Internet Explorer’s commitment to backwards-compatibility has made it really difficult to implement support for modern web standards alongside the existing IE codebase.

Part of the reason why that backwards compatibility exists is that, around the year 2000, there was a massive shift in the way that corporate IT systems were developed. There are countless corporations who have bespoke applications for doing all sorts of business operations – stock control, inventory, HR, project management, all kinds of things. Way back in the 1980s and early 1990s, most of them used a central mainframe system and employees would have to use something like a terminal emulator to connect to that central server, but after the first wave of the dotcom boom hit in the late 1990s, companies realised that most of their PCs now had a web browser and an network connection, and so they could replace their old mainframe terminal applications with web applications. Windows had enormous market share at the time, and Internet Explorer was the default browser on most Windows PCs, so lots of organizations built intranet web applications that only had to work on a specific version of Internet Explorer. Sometimes they did this to take advantage of specific features, like ActiveX support; more often I think they just did it save money because it meant they didn’t have to do cross-browser testing. This happened with some pretty big commercial applications as well; as late as 2011, Microsoft Dynamics CRM still offered no support for any browser other than Internet Explorer.

So you’ve got all these companies who have invested lots of time and money in building applications that only work with Internet Explorer. Those applications aren’t built using web standards or progressive enhancement or with any notion of ‘forward compatibility’ – they’re explicitly targeting one version of one browser running on one operating system. And so when Microsoft releases a new version of Internet Explorer, those applications fail – and the companies don’t want to invest in upgrading their legacy intranet applications, so they blame the browser. So we end up with this weird situation where here in 2017, Microsoft are still shipping Internet Explorer 11, which has a compatibility mode where it switches back to the IE9 rendering engine but sends a user agent string claiming that it's IE7. Meanwhile, everyone I know uses Google Chrome or Safari for all their web browsing - but still has an IE shortcut on their desktop for when they have to log in to one of those legacy systems. .

So… to go back to the original question: is there anything Microsoft could have done to avoid this trap? I think there’s a lot of things they could have done. Building IE from the ground up with a modular rendering engine, so that later versions could selectively load the appropriate engine for rendering a particular website or application. They could have made more effort to embrace the web standards that existed at the time, instead of implementing ad-hoc support for things like the MARQUEE tag and ActiveX plugins, which would have avoided the headache of having to support these esoteric features in later versions. The point is, though, none of this mattered at the time. Their focus – the driving force behind the early versions of Internet Explorer – was not to create a great application with first-class support for web standards. They were trying to kill Netscape Navigator and win market share – and it worked.

Q: Let’s imagine someone is going to introduce an API. So they collect some requirements, propose a version and gets feedback. That’s rather simple and straightforward thing. But if there are any hidden obstacles there down the road?

Always! Requirements are going to change – in fact, one of the biggest mistakes you can make is to try and anticipate those changes and make your design ‘future-proof’. Sometimes that pays off, but what mostly happens is that you end up with a much more complicated design purely because you’re trying to anticipate those future changes. Those obstacles are often things outside your control. There’s a change to the law that means you need to expose certain data in a different way. There’s a change to one of the other systems in your organization, or one of your cloud hosting providers announces that they’re deprecating a particular feature that you were relying on.

The best thing to do is to identify something simple and usable, ship it, and get as quickly as you can to a point where your API is stable, there’s no outstanding technical debt, and your team is free to move on to the next thing. That way when you do encounter one of those ‘hidden obstacles’, you have a stable codebase to use as a basis for your solution, and you have a team who have the time and the bandwidth to deal with it. And if by some stroke of luck you don’t hit any hidden obstacles, then you just move on to the next thing on your backlog.

Q: Continue with API design. We’ve released the v1.0 of our API and now v1.1 is approaching. I believe many of us noticed and or something. What are the best practices (a couple of points) you can think of that can help a developer to design a good v1.1 API in respect to v1.0?

It’s worth reading up on the concept of semantic versioning, (SemVer) and taking the time to really understand the distinction between major, minor and patch versions. SemVer says that you shouldn’t introduce any breaking changes between version x.0 and version x.1, so the most important thing is to understand what would constitute a breaking change for your particular API.

If you’re working with HTTP APIs that return JSON, for example, a typical non-breaking change would be adding a new data field to one of your resources. Clients that are using version 1.1 and are expecting to see this additional field can take advantage of it, whereas clients that are still using version 1.0 can just discard the unrecognised property.

There’s a related question about how you should manage versioning in your APIs. One very common solution is to expose URLs via routing – as opposed to – but if you’re adhering to the constraints of a RESTful system, you really need to understand whether the change in version represents a change in the underlying resource or just the representation. Remember that a URI is a Uniform Resource Identifier, and so we really shouldn’t be changing the URI that we’re using to refer to the same resource.

For example – if we have a resource We could request that resource as a JPEG (Accept: image/jpeg), or as a PNG (Accept: image/png), or ask the server if it has a plain-text representation of the resource (Accept: text/plain) – but they’re just different representations of the same underlying resource, and so they should all use the same URI.

If – say – you’ve completely replaced the CRM system used by your organization, and so “version 1” of a customer represents a record used in the old CRM system and “version 2” represents that same customer after they’ve been migrated onto a completely new platform, then it probably makes sense to treat them as separate resources and give them different URIs.

Versioning is hard, though. The easiest thing to do is never change anything

Q: .NET Core - what do you think about its API?

When .NET Core was first announced in 2015, back when it was going to be be called .NET Core 5.0, it was going to be a really stripped-down, lightweight alternative to the .NET Framework and the Common Language Runtime. That was an excellent idea in terms of making it easier to port .NET Core to different platforms, but it also left a sizable gap between the API exposed by .NET Core, and the ‘standard’ .NET/CLR API that most applications are built against.

I believe – and this is just my interpretation based on what I’ve read and people I’ve talked to – that the idea was that .NET Core would provide the fundamental building blocks. It would provide things like threading, filesystem access, network access, and then a combination of platform vendors and the open source community would develop the modules and packages that would eventually match the level of functionality offered by something like the Java Class Library or the .NET Framework. That’s a great idea in principle, but it also creates a chicken-and-egg situation: people won’t build libraries for a platform with no users, but nobody wants to use a platform that doesn’t have any libraries.

So, the decision was made that cross-platform .NET needed a standard API specification that would provide the libraries that users and application developers expected to be available on the various supported platforms. This is .NET Standard 2.0, which is already fully supported by the .NET Framework 4.6.1 and will be supported in the next versions of .NET Core and Xamarin. Of course, .NET Core 1.1 is out, and works just fine, and you can use it right now to build web apps in C# regardless of whether you’re running Windows or Linux or macOS, which is pretty awesome – but I think the next release of .NET Core is going to be the trigger for a lot of framework and package developers to migrate their projects across to .NET Core, which in turn should make it easier for developers and organizations to migrate their own applications.

Tram on Moscow Gate Square in Saint Petersburg. /  Photo by Dinamik.

API flexibility VS. API precision. One can design a method API so it can accept many different types of values. It’s flexibility. We also can design a method API with lots of rules on input parameters. Both ways are correct. Where is the boundary across these approaches? When should I make a “strict” API and when should I make a more “flexible” design? Don’t forget that you should take backward-compatibility into account.

By implementing an API where the method signatures are flexible, all you’re doing is pushing the complexity to somewhere else in your stack. Say we’re building an API for finding skiing holidays, and we have a choice between

DoSearch(SearchCriteria criteria)


DoSearch(string resortName, string countryCode, int minAltitude, int maxDistanceToSkiList)

One of those methods is pretty easily extensible, because we can extend the definition of the SearchCriteria object without changing the method signature – but we’re still changing the behaviour of the system, we’re just not changing that particular method. By contrast, we could add new arguments to our DoSearch method signature, but if we’re working in a language like C# where you can provide default argument values, you won’t break anything by doing that as long as you provide sensible defaults for the new arguments.

At some point, though, you need to communicate to the API consumers what search parameters are accepted by your API, and there’s lots of ways to accomplish that. If you’re building a .NET API that’s installed as a NuGet package and used from within code, then using XML comments on your methods and properties is a great way to explain to your users what they need to specify when making calls to your API. If your API is an HTTP service, look at using hypermedia and formats like SIREN to define what parameter values and ranges are acceptable.

I should add that I think within the next decade, we’re going to start seeing a whole different category of APIs powered by machine learning systems, where a lot of the conventional rules of API design won’t apply. It wouldn’t surprise me if we got an API for finding skiing holidays where you just specify what you want in natural language, and so there’s not even a method signature – you just call something like DoSearch(“ski chalet, in France or Italy, 1400m or higher, that sleeps 12 people in 8 bedrooms, available from 18-25 January 2018”) – and the underlying system will work it all out for us. Those sorts of development in machine learning are hugely exciting, but they’re also going to create a lot of interesting challenges for the developers and designers trying to incorporate them into our products and applications. 

Thanks to Alexej Sommer for taking the time to set this up (and for translating my answers into Russian – Спасибо!), and if you're at DotNext next week and want to chat about APIs, hypermedia or any of the stuff in the interview, please come and say hi!

Thursday, 27 April 2017

It Works On My Machine!

I saw on Twitter this morning that Derick Bailey is looking for people to share their own “works on my machine” stories… and halfway through filling out his survey, I decided this would probably be much more fun if I nicked his survey questions and turned them into headings in a blog post. Mainly ‘cos writing for an audience appeals to me far more than filling out survey – but Derick (and anyone else who cares?) is very welcome to use anything in this post as part of their own research.

What's typically going through your head when you say "works on my machine" to a QA person or another developer?

I think the interesting question here is actually – what did somebody say to you that caused you to respond with “it works on my machine?”

Here’s three fairly common scenarios:

Q: This code throws an exception when we run it on the staging environment…
A: It works on my machine.

Q: How are you getting on with that improvement to the search algorithm?
A: It works on my machine.

Q: Did you get anywhere with that really weird solution to the mapping problem that Chris found on StackOverflow?
A: It works on my machine…

See how in each case, there’s a sort of implicit subtext? See, I think we all understand that there’s often quite a big difference between solving a problem and delivering a solution. In almost all development scenarios, the first step is to get the code you’re working on running locally and doing the right thing on your development system – and often to do that, we have to hack things around. Running web servers as a local admin user. Granting “Everyone Full Control” of all the files in the media folder. Manually tweaking registry keys, installing DLLs, reusing credentials for APIs and external services – there’s a whole lot of stuff that has to happen as well as just writing some code, but most of the time, the code is the focus and the rest just feels like friction.

So… to answer the original question, when I tell somebody something works on my machine, I’m thinking “ok, what else, other than my own code, do we need to do to deploy the solution, close the ticket and move on?” When you’re working on spikes and prototypes, that’s a natural part of the conversation. If you’ve submitted a ticket to QA for final pre-release testing and it doesn’t work, there’s naturally a bit of tension because implicit in the conversation is the fact that somebody thinks you haven’t done your job properly – and “well it works on my machine” can come across as defensive.

Can you share a story about a time when you have said, or thought, this?

Ah, dozens. The most common example for me is when I’m making a change that spans code in 4-5 different projects, so I’ve checked them all out… in one project I’ve added a database column, in another there’s some new HTTP request routing, in the third there’s a new message queue subscriber, and then there’s the new feature code that relies on all of those changes to work properly. And it works on MY machine because locally I’ve already made all of those changes, but it get it working anywhere else, all five projects need to be reviewed, built, packaged, configured and deployed onto the same environment. Or, for another developer to work on it, they need to check out five specific branches from five specific projects and then probably run a couple of configuration scripts as well – and there’s invariable one or two little things that didn’t require any explicit configuration on my own workstation but then it turns out your teammate has enabled WebDAV publishing under Windows Programs and Features and so their IIS configuration isn’t the same as yours.

How do you typically feel when someone says, "works on my machine," to you?

First off, I’m happy. See, I’ve worked with a very small number of developers who didn’t bother doing even this most basic validation of the work they were doing. They’d commit something, open a pull request and ask for a code review, and you’d look at what they did and think “that’s a bit weird”, so you’d wander over and say “hey, can you show me how this feature works?” – and they will actually say “I don’t know, I can’t run that project”. 

“Works on my machine” at least indicates that they’ve got all the code checked out, they’ve compiled it, and they’ve actually got it running. That’s a good start. That’s something you can work with. And at that point, it’s a great opportunity to explain things like configuration management or deployment scripts.

Can you share a story about a time when someone said this to you?

We had a problem a few weeks ago where we ported an old project from VS2010 to VS2015, and TeamCity wouldn’t build it properly. An absolute textbook example – another developer comes to me and says “well, it works on my machine; it builds fine and I can run all the tests, but TeamCity won’t run any of the unit tests and so the build keeps failing.” Turned out to be a rogue wildcard somewhere in the TeamCity build config settings that was causing it to pick up unit test DLLs from the \obj\ folders instead of \bin\. Which, of course, doesn’t happen when you’re running tests using Resharper or NCrunch, because those tools are smart enough to understand path conventions.

What are the 3 largest causes of someone saying "works on my machine"?

In my experience? The biggest one is dependencies between multiple projects. The “feature”, the unit of business value we’re trying to deliver, requires changes across several different projects and so those changes need to be coordinated and deployed together in order to run and test the new feature, and it’s very easy to miss a step when you’re trying to capture and package all of those code and configuration dependencies.

Second biggest would have to be mismatches between developer environments. We have some people running Windows 8.1, some people running Windows 10, some people working via remote desktop onto virtual machines in the cloud, and then all the various quirks of people’s individual OS configurations like the aforementioned WebDAV Authoring support.

Third? Probably data. Lookup tables, test records, and code that’s brittle because it depends on specific records existing in a particular state, and when you check out the code it doesn’t include the migration steps or SQL scripts that are necessary to set up those records.

Actually, I’m going to go for four, because the one that bites me all the time – probably once a week – is that when you add a new file to a Visual Studio solution, it doesn’t save the .csproj file by default. The new file gets added to the repo and pushed up to GitHub, but the project reference to that new file still only exists on your local machine. Sometimes it’ll crash the build; sometimes – if it’s an image or a script file or something – it’ll build, pass tests, deploy, and then fail on the test server because the new file isn’t included in the .csproj and so wasn’t included when the deployment package was built. If you think your organisation doesn’t have this problem, search your GitHub repo for commit messages including the phrase “csproj file”…

How do you combat the "works on my machine" problem?

There’s a couple of things that have definitely made a big difference to our team at Spotlight. One is setting up an internal NuGet server (we’re running Klondike), and making sure that if your project code references DLLs or any other static components, those dependencies are managed as NuGet packages. That way the first time you build the project, it’ll download all of those obscure DLLs for you instead of waiting for you to get an error message, look it up on the wiki, etc.

One is giving everybody the ability to build and deploy pre-release packages. We have TeamCity and GitHub configured so that as soon as you open a pull request, TeamCity will try and build a deployable package based on the merge head of your feature branch. This means you, the developer, can get packaged builds of your work in progress, deploy it onto one of our testing environments and see for yourself whether it’s going to work or not. Which means you get the chance to fix the bugs and configuration problems before passing it on to anyone else to review or test.

Oh, and we have something called escrow beers. If you want to introduce a new tool, dependency, language or something into one of our projects, you have to put a six-pack of beer (or a box of cookies or something similarly delicious) in escrow, in the kitchen. Put a post-it note on it saying what it’s for – and then when some poor developer is working late to get a feature out and they discover that they need to install grunt or gulp or bower or yeoman or FAKE or PSake or whatever, there’s goodies in the fridge that will help. That’s doesn’t necessarily inhibit the adoption of new tech, but having to go out and buy beer or cookies makes people stop to think about how their changes might affect their teammates, and so they’ll add some checkout scripts to get the new thing working, or document it on the wiki, or organise a demo to show everyone what they need to know. It's also funny how often somebody thinks a new tool or language is ABSOLUTELY TOTALLY AMAZING and there's no way we can possibly live without it… except it's not actually quite amazing enough to justify walking to a shop at lunchtime and buying a box of cookies.

So there you go… more than you ever wanted to know about code that works on my machine. Thanks again to Derick Bailey for the idea – and just to be clear, you’re welcome to use it, and please credit me by name if you use the information provided here in any follow-up posts or other material.

Monday, 24 April 2017

Robert M. Pirsig on "Stuckness"

Robert M. Pirsig, the author of "Zen and the Art of Motorcycle Maintenance", died today aged 88. I've read and re-read that book many times over the years. As somebody who has always found tranquillity in tinkering, I found that "Zen" evokes that meditative, transcendental state that one can achieve whilst doing mechanical maintenance better than anything I've read… and in others, it captures perfectly the awful frustration that can only be experienced when a perfectly simple job turns into a protracted bout of yak-shaving.

Related image

Of all the passages in the book, the one that has stayed with me the most is the one I've included below, on the subject of 'stuckness'. After countless evenings spent tweaking and tuning mountain bikes in my dad's garage, experiencing first-hand the frustration of a £800 mountain bike rendered completely useless by stripping the head off a 50p bolt, this passage resonated with me more than anything I think I've ever read. I still think of it frequently, normally when I find myself stuck on some hitherto inconsequential detail of a software project that's somehow managed to derail the entire team for days at a time. The book is excellent, and if you haven't read it I highly recommend it, but the passage in question is here. I hope Mr Pirsig's lawyers don't mind. :)

Stuckness. That's what I want to talk about today.

A screw sticks, for example, on a side cover assembly. You check the manual to see if there might be any special cause for this screw to come off so hard, but all it says is "Remove side cover plate" in that wonderful terse technical style that never tells you what you want to know. There's no earlier procedure left undone that might cause the cover screws to stick.

If you're experienced you'd probably apply a penetrating liquid and an impact driver at this point. But suppose you're inexperienced and you attach a self-locking plier wrench to the shank of your screwdriver and really twist it hard, a procedure you've had success with in the past, but which this time succeeds only in tearing the slot of the screw.

Your mind was already thinking ahead to what you would do when the cover plate was off, and so it takes a little time to realize that this irritating minor annoyance of a torn screw slot isn't just irritating and minor. You're stuck. Stopped. Terminated. It's absolutely stopped you from fixing the motorcycle.

This isn't a rare scene in science or technology. This is the commonest scene of all. Just plain stuck. In traditional maintenance this is the worst of all moments, so bad that you have avoided even thinking about it before you come to it.

The book's no good to you now. Neither is scientific reason. You don't need any scientific experiments to find out what's wrong. It's obvious what's wrong. What you need is an hypothesis for how you're going to get that slotless screw out of there and scientific method doesn't provide any of these hypotheses. It operates only after they're around.

This is the zero moment of consciousness. Stuck. No answer. Honked. Kaput. It's a miserable experience emotionally. You're losing time. You're incompetent. You don't know what you're doing. You should be ashamed of yourself. You should take the machine to a real mechanic who knows how to figure these things out.

It's normal at this point for the fear-anger syndrome to take over and make you want to hammer on that side plate with a chisel, to pound it off with a sledge if necessary. You think about it, and the more you think about it the more you're inclined to take the whole machine to a high bridge and drop it off. It's just outrageous that a tiny little slot of a screw can defeat you so totally.

What you're up against is the great unknown, the void of all Western thought. You need some ideas, some hypotheses. Traditional scientific method, unfortunately, has never quite gotten around to say exactly where to pick up more of these hypotheses. Traditional scientific method has always been at the very best, 20-20 hindsight. It's good for seeing where you've been. It's good for testing the truth of what you think you know, but it can't tell you where you ought to go, unless where you ought to go is a continuation of where you were going in the past. Creativity, originality, inventiveness, intuition, imagination..."unstuckness," in other words...are completely outside its domain.

We're still stuck on that screw and the only way it's going to get unstuck is by abandoning further examination of the screw according to traditional scientific method. That won't work. What we have to do is examine traditional scientific method in the light of that stuck screw.

We have been looking at that screw "objectively." According to the doctrine of "objectivity," which is integral with traditional scientific method, what we like or don't like about that screw has nothing to do with our correct thinking. We should not evaluate what we see. We should keep our mind a blank tablet which nature fills for us, and then reason disinterestedly from the facts we observe.

But when we stop and think about it disinterestedly, in terms of this stuck screw, we begin to see that this whole idea of disinterested observation is silly. Where are those facts? What are we going to observe disinterestedly? The torn slot? The immovable side cover plate? The color of the paint job? The speedometer? The sissy bar? As Poincaré would have said, there are an infinite number of facts about the motorcycle, and the right ones don't just dance up and introduce themselves. The right facts, the ones we really need, are not only passive, they are damned elusive, and we're not going to just sit back and "observe" them. We're going to have to be in there looking for them or we're going to be here a long time. Forever. As Poincaré pointed out, there must be a subliminal choice of what facts we observe.

The difference between a good mechanic and a bad one, like the difference between a good mathematician and a bad one, is precisely this ability to select the good facts from the bad ones on the basis of quality. He has to care! This is an ability about which formal traditional scientific method has nothing to say. It's long past time to take a closer look at this qualitative preselection of facts which has seemed so scrupulously ignored by those who make so much of these facts after they are "observed." I think that it will be found that a formal acknowledgment of the role of Quality in the scientific process doesn't destroy the empirical vision at all. It expands it, strengthens it and brings it far closer to actual scientific practice.

I think the basic fault that underlies the problem of stuckness is traditional rationality's insistence upon "objectivity," a doctrine that there is a divided reality of subject and object. For true science to take place these must be rigidly separate from each other. "You are the mechanic. There is the motorcycle. You are forever apart from one another. You do this to it. You do that to it. These will be the results."

This eternally dualistic subject-object way of approaching the motorcycle sounds right to us because we're used to it. But it's not right. It's always been an artificial interpretation superimposed on reality. It's never been reality itself. When this duality is completely accepted a certain nondivided relationship between the mechanic and motorcycle, a craftsmanlike feeling for the work, is destroyed. When traditional rationality divides the world into subjects and objects it shuts out Quality, and when you're really stuck it's Quality, not any subjects or objects, that tells you where you ought to go.

By returning our attention to Quality it is hoped that we can get technological work out of the noncaring subject-object dualism and back into craftsmanlike self-involved reality again, which will reveal to us the facts we need when we are stuck.

Let's consider a reevaluation of the situation in which we assume that the stuckness now occurring, the zero of consciousness, isn't the worst of all possible situations, but the best possible situation you could be in. After all, it's exactly this stuckness that Zen Buddhists go to so much trouble to induce; through koans, deep breathing, sitting still and the like. Your mind is empty, you have a "hollow-flexible" attitude of "beginner's mind." You're right at the front end of the train of knowledge, at the track of reality itself. Consider, for a change, that this is a moment to be not feared but cultivated. If your mind is truly, profoundly stuck, then you may be much better off than when it was loaded with ideas.

The solution to the problem often at first seems unimportant or undesirable, but the state of stuckness allows it, in time, to assume its true importance. It seemed small because your previous rigid evaluation which led to the stuckness made it small.

But now consider the fact that no matter how hard you try to hang on to it, this stuckness is bound to disappear. Your mind will naturally and freely move toward a solution. Unless you are a real master at staying stuck you can't prevent this. The fear of stuckness is needless because the longer you stay stuck the more you see the Quality...reality that gets you unstuck every time. What's really been getting you stuck is the running from the stuckness through the cars of your train of knowledge looking for a solution that is out in front of the train.

Stuckness shouldn't be avoided. It's the psychic predecessor of all real understanding. An egoless acceptance of stuckness is a key to an understanding of all Quality, in mechanical work as in other endeavors. It's this understanding of Quality as revealed by stuckness which so often makes self-taught mechanics so superior to institute-trained men who have learned how to handle everything except a new situation.

Normally screws are so cheap and small and simple you think of them as unimportant. But now, as your Quality awareness becomes stronger, you realize that this one, individual, particular screw is neither cheap nor small nor unimportant. Right now this screw is worth exactly the selling price of the whole motorcycle, because the motorcycle is actually valueless until you get the screw out. With this reevaluation of the screw comes a willingness to expand your knowledge of it.

- from "Zen and the Art of Motorcycle Maintenance" by Robert M Pirsig (September 6, 1928 – April 24, 2017)

Friday, 21 April 2017

There’s a problem with the phalange!

Yesterday I was throwing together a quick Entity Framework prototype to inspect and wrangle some data held in one of our legacy databases. I’m using the Entity Framework “Code first from Database” approach, where you use the tooling to generate your initial model for you but thereafter you modify it by hand. One of the tables I’m working with here is called LookupRanges, so when I generated a bunch of entities and DbSet<> mappings, I was a bit surprised when I ended up with a class called LookupRanx in my new model.

It took a minute or two to ascertain that yes, Entity Framework had mapped my LookupRanges (plural) table name onto a class called LookupRanx. But where on earth did that ‘Ranx’ come from? My hunch here is that somebody who worked on this pluralization code remembered that the English word phalanx has the plural form phalanges – and so implemented a rule that says ‘any word ending in –anges should be singularizaed to –anx. Out of curiousity, I dug out my huge ASCII file of English words, found all the words ending in *anges, and hacked up a quick SQL script to create tables named for all these words so I could run them through EF and see what class names were generated.

Well, it gets ‘changes’ and ‘phalanges’ right – and gets literally every other case wrong.


Now this, to me, is an outstanding example of one of the biggest problems in software development… smart people like working on things that are interesting, and will frequently spend time doing something that’s interesting instead of something that’s important.

Writing a library that can singularize and pluralize English words is fascinating. It’s a never-ending problem with dozens of rules and hundreds of edge cases, and you learn a lot of weird and cool esoteric facts about language and etymology whilst you’re doing it. But in this instance, something started out as a good idea (“hey – wouldn’t it be cool if the model generator would convert plural table names to singular class names?”), and got bogged down in edge cases (“is the plural of ‘tableau’ really ‘tableaux’?”) and – in this instance – ended up with a bizarre bug because one of those so-called edge cases actually ended up breaking the default – and entirely correct – behaviour.

First, phalanges is extremely unlikely to ever show up as the name of a table in an Entity Framework database model. I can think of dozens of real-world scenarios where you’d end up with a table name ending with –Ranges, –Exchanges or –Interchanges, but I’m honestly struggling to think of any remotely likely scenario where you have a Phalanges table in a SQL Server database. This is the kind of thing where the ticket or the user story probably just says ‘implement pluralization’, and then there’s no sub-prioritization or further analysis about just how much pluralization needs to be implemented. Second – it’s kind of a stupid edge case. We’re not trying to win points on University Challenge here, we’re building software. In contemporary English, phalange is an acceptable singular form, and phalanxes is an acceptable plural form. There’s no reason at all why they needed to implement support for this particular edge case. And third: if you really found yourself in a scenario where you had to map the Phalanges table to the Phalanx class, you can just rename it. It’s easy. Visual Studio has first-class support for this kind of refactoring.

And, as if that wasn’t confusing enough, there’s actually source code for an EnglishPluralizationService.cs on Microsoft’s GitHub repository. Somebody obviously had a lot of fun building this, tracking down all those bizarre little edge cases like seraph/seraphim, hippopotamus/hippopotami – but according to this implementation, the plural of phalanx is…  go on. Go and take a look.

Now, though, I’m going to eat an oranx and check my email. Using Microsoft Exchanx, of course.

Monday, 3 April 2017

The Pursuit of APIness: The Secret to Happy Code

I'll be giving a new talk at the London.NET User Group meetup here in London next Tuesday, based on an idea I've had rattling around for a decade or more now. See, it seems to me that over the course of my career, there's been a strong correlation between happy developers and successful projects. I can't think of any examples where a miserable death-march project has resulted in high-quality working software, and I can't think of too many instances where a group of happy, motivated developers has failed to deliver a working product. I've been thinking around this idea for a while, and started looking at it in terms of user experience – both the user experience that we as developers are creating for our end users, but also the 'user experience' that's being provided by the libraries, frameworks and tools that we're using to do our jobs. Here's the talk synopsis:

We spend our lives working with systems created by other people. From the UI on our phones to the cloud infrastructure that runs so much of the modern internet, these interactions are fundamental to our experience of technology - as engineers, as developers, as users - and user experiences are viral. Great user experiences lead to happy, productive people; bad experiences lead to frustration, inefficiency and misery.

Whether we realise it or not, when we create software, we are creating user experiences. People are going to interact with our code. Maybe those people are end users; maybe they're the other developers on your team. Maybe they're the mobile app team who are working with your API, or the engineers who are on call the night something goes wrong. These may be radically different use cases, but there's one powerful principle that works across all these scenarios and more. In this talk, we'll draw on ideas and insight from user experience, API design, psychology and education to show how you can incorporate this principle, known as discoverability, into every layer of your application. We'll look at some real-world systems, and we'll discuss how discoverability works with different interaction paradigms. Because, whether you're building databases, class libraries, hypermedia APIs or mobile apps, sooner or later somebody else is going to work with your code - and when they do, wouldn't it be great if they went away afterwards with a smile on their face?

If that sounds interesting (or if you think I'm completely wrong and you want to come along and heckle!), sign up at the SkillsMatter website and come along on Tuesday 11th. Hope to see you there.

Wednesday, 22 March 2017

Goodbye ECMAScript; hello UKMAScript!

The UK government has announced it will trigger Article 50 on March 29th, beginning the two-year process of the United Kingdom leaving the European Union.

That’s right - it will no longer be legal for British web developers to run ECMAScript, since the ECMAScript specification is controlled by the European Computer Manufacturers’ Association. We’re happy to announce that as of today, top engineers are starting work on a superior British programming language called UKMAScript.

UKMAScript will extend the core language specification with the following enhancements, which we believe will provide a massive boost to the UK tech industry and offset the immeasurable damage caused when all our EU colleagues and collaborators decide to exercise their freedom of movement.

  • Along with NaN and Infinity, UKMAScript will support a new primitive numeric value called MoneyForTheNhs, whose value is defined to be exactly 3.5x108 until it’s used as an argument to any function, at which point its value will be silently changed to zero after the function has returned.
  • The Math.round() method will behave as before, except Math.round(0.52) will now return Number.MAX_VALUE. Math.round(0.48) will return a new constant Number.TRAITORS and any attempt to use this value in calculations will throw a TreasonError
  • A new “illogical implication” operator !#> will be introduced. This is syntactically similar to the notion of logical implication in Boolean algebra, but designed to allow the scope of arguments to be massively exaggerated. For example, the statement (leave_eu !#> leave_customs_union && leave_eea) will implicitly bind the values of leave_customs_union and leave_eea to the value of leave_eu, despite this dependency not being expressed anywhere else in the codebase.
  • Along with null and undefined, a new language primitive brexit will be introduced. This has the special equality semantics (brexit == brexit) == undefined. typeof(brexit) will return the value “hard”, and attempting to evaluate brexit.valueOf() will throw a TreasonError.
  • UKMAScript features a new parallel programming paradigm implemented via the Referendum.Invoke() method. This causes a thread to break away from the main sequence of program control and attempt to continue execution despite no longer having access to any processing capabilities or shared resources of the host system. Note that if a thread A has called Referendum.Invoke(), any child process B attempting to call Referendum.Invoke() will be summarily ignored by process A on the grounds that it’s clearly developed a fault.

UKMAScript ships with no standard library or runtime, but the UKMAScript language committee assures us that platform vendors are lining up to deliver first-class support for the new language.

To further promote the popularity of UKMAScript, the only alternative permitted once Article 50 has been invoked is a new language called LABOUR, which takes many of the core language principles of COBOL-64 but is only accessible using the GNU/Corbyn compiler. This compiler has a tremendously exciting installation routine but then doesn’t actually do anything other than occasionally create internal process deadlocks for no reason.

Tuesday, 7 March 2017

It's a bug! It's a feature! It's… a limitation of the fundamental design of your test framework?

As some of you probably know, I'm a big fan of NCrunch. When I'm coding in C#, NCrunch gets a CPU core and a whole screen to itself (yes, I don't really write code on a system that looks like this) and sits there quietly running all my tests, all the time, and telling me the second I break anything.

I'm also a big fan of testing things that are as close to production behaviour as you can. Unit tests are great for informing the design of your components, but without integration testing you can't be sure they're actually going to work when you stick them together.

So on my current project, there's a suite of unit tests using FakeItEasy and assertions, and then a suite of integration tests that connect to the live API, follow the various hypermedia links, throw assorted JSON objects at the PUT and POST endpoints to see how they respond, and then call DELETE to clean up when they're done. And, just to keep us honest, we've got a post-deploy step in our Octopus Deploy script that will actually run the integration test suite as part of the deployment process, and roll the whole thing back if any of the tests fail. Another small step on the road to truly continuous deployment.

Anyway. Last week, I push a release to our dev environment, and a whole load of tests fail. Which is weird, because it worked on MY machine. And it worked on my machine when I pointed my local codebase at the database in the dev environment. And – here's the fun part – it worked on my machine when I pointed the entire test suite at the dev environment. So I start eliminating variables. One of the first things I pick up on is that my local test runner is NCrunch, whereas the post-deploy step is using nunit-console. So I run the local integration tests using nunit-console and – bang. Failures. Which is good, because I know what's causing the weirdness, but weird, because tests are supposed to either pass or fail regardless of what test runner you're using.

So I dig a little deeper, and I end up with what looks to me like a bug in NCrunch. See, we're using the TestCaseSource attribute to generate test cases for the API tests, and – because all we need is a bunch of different JSON objects – we're just spinning up new anonymous objects and passing them in as test cases.

Here's two anonymous objects:

var testCase1 = new { forenames = null, surname = "Batman" }
var testCase2 = new { forenames = String.Empty, surname = "Batman" }

What I noticed is that if you generate these two test cases, NCrunch will only see them as a single test – which I assumed was because their ToString() representations are equal, because null and String.Empty both return String.Empty when you ToString() them in this situation. So I opened a post about it on the NCrunch forums, even going so far as to suggest using GetHashCode() when enumerating test names, and got this really interesting response from Remco Mulder, the NCrunch lead developer:

Tests must be uniquely identifiable between execution and discovery runs. This isn't important for a tool like the nunit console runner where a test can be discovered and executed within the same process call (and thus identified by its memory address), but for a tool like NCrunch, there's no way to run the test or collect data from it without this. As you've identified, generated tests with a null parameter and an empty string will return the same result under .ToString(), so NCrunch can't tell them apart.

The only way to solve this is to change the design of your code. Try using the NUnit .SetName() method to give each of your generated tests a distinctive name.

Unfortunately .GetHashCode() is not a reliable solution to this problem as this method is not designed to generate the same identifier across different processes. This method returns different results under x86 vs x64, and under .NET Core it will actually return an entirely different result for each process. Because your code is responsible for generating the tests, the problem can only be solved within your own code.

I thought this was a really interesting insight into how a tool like NCrunch has to deal with situations that an in-process test runner like nunit-console will probably never encounter. It also turns out I’d dismissed that very warning a few weeks earlier – when it cropped up in response to an unrelated issue which produced the same symptons – and sure enough, after clicking the “Show all hidden warnings” button on the NCrunch toolbar, the warning popped back up – along with a very detailed explanation of what was causing it:


Plus, I had no idea that NUnit has a TestCaseData interface with a SetName() method on it, which gives a much nicer way of presenting these test cases in both NCrunch and NUnit. I've ended up with something akin to:

public static IEnumerable TestData() {
  foreach (var data in new[] { null, String.Empty }) {
    var testCase = new { forenames = data, surname = "Batman" };
    var json = JsonConvert.SerializeObject(testCase);
    yield return new TestCaseData(testCase).SetName(json);

Oh, and if you're interested, the deployment failures were because of a weird validation rule that treats null as missing, which is fine, but String.Empty as an empty string which violates a string length constraint. Which is wrong, and now the API doesn't do it any more. This is just another reason why integration testing is a good idea. So there you have it – a bug that wasn’t a bug, a crash-course in how NUnit and NCrunch actually work behind the scenes, and a TIL for naming your NUnit tests explicitly. Happy Friday.