In this post I will walk through how to refactor a Factory, moving from a sequence of ifs to a dictionary implementation, and using delegates as a type alias for my object creation methods.
How applying Theory of Constraints helped us optimise our code
My team have been working on improving the performance our API, and identified a database call as the cause of some problems.
The team suggested three ways to tackle this problem:
- Scale up the database till it can meet our requirements.
- Introduce some light-weight caching in the application to reduce load on the database.
- Examine the query plan for this database call to find out whether the query can be optimised.
Which of these should we attempt first? There was some intense discussion about this, with arguments made in favour of each approach. What we needed was a simple framework for making decisions about how to improve our system.
This is where the Theory of Constraints (ToC) can help. Originally expounded as a paradigm for improving manufacturing systems, ToC is really useful in software engineering, both when managing projects and when improving the performance of the systems we create.
Theory of Constraints
The preliminary step in applying ToC is to identify the Goal of your system. In the case of this API, the Goal is to supply accurate data to consumers.
Now we understand the Goal of the system, we can define the Throughput of the system as the rate at which it can deliver units of that goal, in our case API responses. We can also define the Operating Expenses of the system (the cost of servers) and its Inventory (requests waiting for responses).
The next step is to identify the Constraint of the system. This is the element in the system that dictates the system’s Throughput. In a physical system, a useful heuristic is a build-up of Inventory in front of this element. In our API, our monitoring helped us pinpoint the bottleneck.
The next three steps give us a sequence of approaches for tackling the Constraint:
- First, Exploit the Constraint by finding local changes you can make to improve its performance.
- Second, Subordinate the rest of the system to the Constraint by finding ways to reduce pressure on it so it can perform more smoothly.
- Third, Elevate the Constraint by increasing the resources available to it, committing to additional Operating Expenses if necessary.
Exploitation comes first because it’s quick, cheap and local. To Subordinate you need to consider the effects on the rest of the system, but there shouldn’t be significant costs involved. Elevating the Constraint may well cost a fair amount, so it comes last on the list.
Once you have applied these steps you will either find that the Constraint has moved elsewhere (you’ve ‘broken’ the original Constraint), or it has remained in place. In either case, you should repeat the steps as part of a culture of continuous improvement. Eventually you want to see the constraint move outside your system and become a matter of consumer demand.
Applying ToC to our question
If we look at the team’s three suggestions, we can see that each corresponds to one of these techniques:
- Scaling up the database is Elevation: there’s a clear financial cost in using larger servers.
- Introducing caching is Subordination: we’re changing the rest of the system to reduce pressure on the Constraint, and need to consider questions such as cache invalidation before we make this change.
- Optimising the query is Exploitation: we’re making local changes to the Constraint to improve its performance.
Applying ToC tells us which of these approaches to consider first, namely optimising the query. We can look at caching if an optimised query is still not sufficient, and scaling should be a last resort.
In our case, query optimisation was sufficient. We managed to meet our performance target without introducing additional complexity to the system or incurring further cost.
Further Reading
Goldratt, Eliyahu M.; Jeff Cox. The Goal: A Process of Ongoing Improvement. Great Barrington, MA.: North River Press.
Concrete Inheritance and the Dependency Inversion Principle
Collaboration and DIP
We know that this code violates the Dependency Inversion Principle (DIP):
public class Atm
{
private readonly InMemoryTransactions _inMemoryTransactions;
public Atm()
{
_inMemoryTransactions = new InMemoryTransactions(new List<Transaction>());
}
public void Deposit(int amount)
{
_inMemoryTransactions.Deposit(amount);
}
}
In particular, line 3 violates the principle by creating a dependency on a specific implementation, and line 7 violates it by knowing how to construct that dependency.
We can rewrite this code to apply the DIP:
public class Atm
{
private Transactions _transactions;
public Atm(Transactions transactions)
{
_transactions = transactions;
}
public void Deposit(int amount)
{
_transactions.Deposit(amount);
}
}
Line 3 now depends on an abstraction, and we use constructor injection to supply this dependency.
Inheritance and DIP
So how about this code:
public class LoggingInMemoryTransactions : InMemoryTransactions
{
private readonly Logger _logger;
public LoggingInMemoryTransactions(List<Transaction> initialTransactions, Logger logger) : base(initialTransactions)
{
_logger = logger;
}
public override void Deposit(int amount)
{
_logger.Log($"Deposit {amount}");
base.Deposit(amount);
}
}
Well, this class depends on a Logger abstraction, which is injected in the constructor, so that dependency seems to have been satisfactorily inverted.
However, in line 1 we can see that this class inherits from InMemoryTransactions. This inheritance is also a dependency, and we’re inheriting from a concrete instance, not an abstraction.
If we then look at the constructor on line 5, we can see that it calls the base constructor; in other words, it has to know how its base class is instantiated.
There is a direct parallel between these two cases: each of them has a dependency on a specific concrete class, which is instantiated in a specific way.
We can rewrite this code to use composition rather than inheritance:
public class LoggingInMemoryTransactions : Transactions
{
private readonly Logger _logger;
private readonly Transactions _transactions;
public LoggingInMemoryTransactions(Logger logger, Transactions transactions)
{
_logger = logger;
_transactions = transactions;
}
public void Deposit(int amount)
{
_logger.Log($"Deposit {amount}");
_transactions.Deposit(amount);
}
}
This simple example of the Decorator pattern behaves in exactly the same way as the previous implementation, but is now only dependent on abstractions.
Unit Testing
If we try to write unit tests around our code, we will soon see the benefit of DIP.
If we try to test our first implementation of Atm, we find that our test boundary includes InMemoryTransactions, as that class is a hard-wired dependency. We can’t test the behaviour of Atm without also testing the behaviour of InMemoryTransactions. If we also have tests of InMemoryTransactions, then we may end up duplicated test scenarios.
In this case we have an additional problem: we have only defined read access to the data, so we have no way of testing this class without violating encapsulation:
using NUnit.Framework;
[TestFixture]
public class AtmShould
{
[Test]
public void Perform_deposit_transaction()
{
const int amount = 100;
var atm = new Atm();
atm.Deposit(amount);
// HELP! WHAT CAN I ASSERT AGAINST?
}
}
In the second implementation, we can write collaboration tests between Atm and Transactions, bringing our test boundary in much closer, and restricting the behaviour under test to that of Atm.
using NSubstitute;
using NUnit.Framework;
[TestFixture]
public class AtmShould
{
[Test]
public void Perform_deposit_transaction()
{
const int amount = 100;
var transactions = Substitute.For<Transactions>();
var atm = new Atm(transactions);
atm.Deposit(amount);
transactions.Received().Deposit(amount);
}
}
Similarly, to test the first implementation of LoggingInMemoryTransactions, we also have to test the behaviour it inherits from InMemoryTransactions, as one is inseparable from the other. Again, if we also have tests of the base class, then we may end up with duplicated test scenarios.
using NSubstitute;
using NUnit.Framework;
[TestFixture]
public class LoggingInMemoryTransactionsShould
{
private LoggingInMemoryTransactions _loggingInMemoryTransactions;
private Logger _logger;
private const int Amount = 100;
[SetUp]
public void SetUp()
{
_logger = Substitute.For<Logger>();
_loggingInMemoryTransactions = new LoggingInMemoryTransactions(new List<Transaction>(), _logger);
}
[Test]
public void Log_transaction()
{
_loggingInMemoryTransactions.Deposit(Amount);
_logger.Received().Log($"Deposit {Amount}");
}
[Test]
public void Store_deposit()
{
_loggingInMemoryTransactions.Deposit(Amount);
// HELP! HOW CAN I CHECK THIS WAS SAVED?
// I might be tempted to assert against the new List<Transaction>(), but this would leak implementation details.
}
}
In the second implementation, the specific behaviour of LoggingInMemoryTransactions is separated, and we can write collaboration tests independently of the behaviour of the inner implementation.
using NSubstitute;
using NUnit.Framework;
[TestFixture]
public class LoggingInMemoryTransactionsShould
{
private LoggingInMemoryTransactions _loggingInMemoryTransactions;
private Logger _logger;
private Transactions _transactions;
private const int Amount = 100;
[SetUp]
public void SetUp()
{
_logger = Substitute.For<Logger>();
_transactions = Substitute.For<Transactions>();
_loggingInMemoryTransactions = new LoggingInMemoryTransactions(_logger, _transactions);
}
[Test]
public void Log_transaction()
{
_loggingInMemoryTransactions.Deposit(Amount);
_logger.Received().Log($"Deposit {Amount}");
}
[Test]
public void Store_deposit()
{
_loggingInMemoryTransactions.Deposit(Amount);
_transactions.Received().Deposit(Amount);
}
}
Updated 26 May 2016: I’ve given code examples for the unit tests we might write in each case, and have removed DateTime createdOn from the example to avoid confusing the issue.
Updated 26 September 2017: I’ve removed a reference to an individual whose political views are not consistent with the spirit of this blog.
A strange behaviour in ReSharper’s Move Instance Method refactoring
My colleague Pedro and I were puzzling over some bizarre behaviour with ReSharper’s Move Instance Method refactoring move. This is a fairly complex move, and it makes various changes to the source and destination classes; however, in many cases, it generates code that does not compile, and seems to include wrongly qualified references.
Here is an example drawn from Martin Fowler’s Refactoring: Improving the Design of Existing Code:
Employee class before refactoring
using System;
namespace RefactoringExamples.ReplaceConditionalWithPolymorphism
{
public class Employee
{
private EmployeeType _employeeType;
private readonly int _monthlySalary;
private readonly int _commission;
private readonly int _bonus;
public Employee(int type)
{
Type = type;
_monthlySalary = 100;
_commission = 10;
_bonus = 20;
}
public int Type
{
get { return _employeeType.TypeCode; }
set { _employeeType = EmployeeType.TypeFrom(value); }
}
public int PayAmount()
{
return Pay();
}
private int Pay()
{
switch (_employeeType.TypeCode)
{
case EmployeeType.Engineer:
return _monthlySalary;
case EmployeeType.Salesperson:
return _monthlySalary + _commission;
case EmployeeType.Manager:
return _monthlySalary + _bonus;
default:
throw new Exception("Incorrect Employee");
}
}
}
}
EmployeeType class before refactoring:
using System;
namespace RefactoringExamples.ReplaceConditionalWithPolymorphism
{
public abstract class EmployeeType
{
public abstract int TypeCode { get; }
public static EmployeeType TypeFrom(int value)
{
switch (value)
{
case Engineer:
return new Engineer();
case Salesperson:
return new Salesperson();
case Manager:
return new Manager();
default:
throw new Exception("Incorrect Employee Code");
}
}
public const int Engineer = 0;
public const int Salesperson = 1;
public const int Manager = 2;
}
class Engineer : EmployeeType
{
public override int TypeCode => Engineer;
}
class Salesperson : EmployeeType
{
public override int TypeCode => Salesperson;
}
class Manager : EmployeeType
{
public override int TypeCode => Manager;
}
}
Employee class after refactoring
using System;
namespace RefactoringExamples.ReplaceConditionalWithPolymorphism
{
public class Employee
{
private EmployeeType _employeeType;
private readonly int _monthlySalary;
private readonly int _commission;
private readonly int _bonus;
public Employee(int type)
{
Type = type;
_monthlySalary = 100;
_commission = 10;
_bonus = 20;
}
public int Type
{
get { return _employeeType.TypeCode; }
set { _employeeType = EmployeeType.TypeFrom(value); }
}
public int MonthlySalary
{
get { return _monthlySalary; }
}
public int Commission
{
get { return _commission; }
}
public int Bonus
{
get { return _bonus; }
}
public int PayAmount()
{
return _employeeType.Pay(this);
}
}
}
EmployeeType class after refactoring
using System;
namespace RefactoringExamples.ReplaceConditionalWithPolymorphism
{
public abstract class EmployeeType
{
public abstract int TypeCode { get; }
public static EmployeeType TypeFrom(int value)
{
switch (value)
{
case Engineer:
return new Engineer();
case Salesperson:
return new Salesperson();
case Manager:
return new Manager();
default:
throw new Exception("Incorrect Employee Code");
}
}
public const int Engineer = 0;
public const int Salesperson = 1;
public const int Manager = 2;
public int Pay(Employee employee)
{
switch (System.TypeCode)
{
case EmployeeType.Engineer:
return employee.MonthlySalary;
case EmployeeType.Salesperson:
return employee.MonthlySalary + employee.Commission;
case EmployeeType.Manager:
return employee.MonthlySalary + employee.Bonus;
default:
throw new Exception("Incorrect Employee");
}
}
}
class Engineer : EmployeeType
{
public override int TypeCode => Engineer;
}
class Salesperson : EmployeeType
{
public override int TypeCode => Salesperson;
}
class Manager : EmployeeType
{
public override int TypeCode => Manager;
}
}
Notice the reference to System.TypeCode in the EmployeeType.Pay method on line 30 above; this doesn’t compile, and is most definitely not what we wanted. My hypothesis is that once ReSharper has rewritten the code, it attempts to resolve and properly qualify any symbol names; in this case, it spots TypeCode and, because this is the name of a type known to the compiler, it decides that the name needs to be appropriately qualified, adding the System namespace. It’s failing to recognise that TypeCode is actually the name of a property on the EmployeeType class, and that no qualification is necessary.
I have found three ways to work round this:
First, you can give the property a name that not already used by a type known to the compiler. If I rename TypeCode to Code, then the refactoring move works perfectly. In this case, this is probably a nice idea, as the name EmployeeType.TypeCode contains redundancy. However, in many cases the most appropriate name for a property may coincide with a type name, and renaming it will not be a good option.
The second option is to recognise that this happens, and fix it. In this case, all I need to do is remove the System namespace on line 30, and everything works correctly. I’m not a huge fan of this solution, as I like my automatic refactorings to be safe and reliable, but it may be a pragmatic choice.
The third option is to perform this refactoring by composing smaller, safe steps. This can give you finer grained control over the outcome, but at the expense of complexity.
I would be interested to hear if anyone has any further insight into what’s going on here, or if there are other techniques for overcoming it. Do let me know if you do!
Update (25 May 2016): I sent JetBrains a link to this post, and within an hour they raised a bug report. Thanks to Daria at JetBrains for the speedy response!
Reinvigorating a daily stand-up by walking the board
I’ve been working with a team who had a problem with focus: when I joined them, they seemed to be busy all the time, but they were frustrated that they weren’t making progress towards their sprint goals.
This situation could be seen in microcosm at the daily stand-up meeting. In this post I’m going to describe how a simple adjustment to this meeting helped us start to improve focus, morale and productivity.
The Scrum Guide gives a template for the daily stand-up meeting (which it calls the Daily Scrum):
The Daily Scrum is held at the same time and place each day to reduce complexity. During the meeting, the Development Team members explain:
- What did I do yesterday that helped the Development Team meet the Sprint Goal?
- What will I do today to help the Development Team meet the Sprint Goal?
- Do I see any impediment that prevents me or the Development Team from meeting the Sprint Goal?
My team was indeed practising this technique, but it seemed that they often forgot the discipline of focusing on ‘helping the Development Team meet the Sprint Goal’, and the meeting often descended into yesterday-I-diddery, with each team member recounting all the things they did the day before in trivial detail.
It seems to me that in a stand-up of this format, each team member’s incentive becomes having something to say, rather than showing progress towards the Sprint Goal, and this produces an incentive to be busy, no matter how irrelevant or frankly counterproductive the tasks might be. In this team, I saw a lot of effort spent on support tasks—whether or not the issue was pressing—, a significant amount of aimless ‘refactoring’, which was essentially yak shaving, and a tendency for team members to interrupt each other for help with lower priority work. In effect, everyone starts prioritising busy work, rather than focusing on the team’s goals.
The other consequence of this approach was that the team’s board was a poor representation of our work: people would be working on tasks that weren’t visible on the board, and the stories that were on the board often didn’t move anywhere. The Scrum Master and I tried various approaches to coordinate the board and the stand-up reports, but the focus was still lacking.
I’ve previously worked in a Kanban environment, and the format of a Kanban standup is significantly different:
Standup meetings have evolved differently with Kanban. The need to go around the room and ask the three questions is obviated by the card wall. … The focus is on flow of work. The facilitator … will “walk the board.” The convention has developed work backward—from right to left (in the direction of pull)—through the tickets on the board. The facilitator might solicit a status update on a ticket or simply ask if there is any additional information that is not on the board and may not be known to the team.
(Anderson, David J. Kanban, Successful evolutionary change for your technology business)
I suggested that we try this approach for a week, and see whether it helped give us more focus. As people were concerned that we might lose sight of important work, we agreed that we would walk the board first, and then run quickly round the team to see if anything was missing.
The initial results were encouraging, and several weeks later we are still walking the board, rather than going round the team. In particular, our board now contains a great deal more information on the tasks in play, and the team have got really good at carding up even small tasks so they are visible at the next stand-up. The amount of off-plan and busy work has also dropped, and this also be a result of the focus on the tasks on the board. Perhaps my favourite development is that the tasks on the board have become much smaller: the drive to get things done is now focused on pulling small, focused tasks across the board, rather than doing busy work.
Of course this technique is no cure-all, and it took a while for the team to acquire the discipline of walking the board in order, rather than jumping in to discuss whichever task particularly excites them. However, as an incremental adaptation, I’m very pleased with its results.
A Retrospective in the Park
The other day, I facilitated a sprint retrospective in the park. The sun was shining, and we had all been working hard to complete our backlog, so it felt like a nice reward for everyone’s efforts. Holding a retrospective outdoors can also give it an energy and sense of enthusiasm that is harder to find in a small room.
I’ve run outdoor retrospectives before, and have previously followed fairly classic plans, with much arranging of index cards. This has never been a great success, as the slightest breath of a breeze can make a mess of your planning. For this retrospective, I designed a plan to avoid these problems, drawing some ideas from the Appreciative Retrospective plan.
This retrospective took an hour for a team of nine. You’ll need a pile of index cards or sticky notes, and a pen per person. Here’s how to do it:
1. Choose your location
Some people are sun lovers, whilst others, like me, burn easily and need some shade, so find a location that will work for everyone. Don’t worry too much about the state of the grass, as I suggest you conduct the retrospective standing up, if possible.
Get everyone to stand in a circle, with enough personal space for everyone, but close enough that you can hear everyone speak.
Observations: It’s nice if you can find a location where there won’t be too many distractions. We weren’t entirely successful: there was a hen party in another corner of the park, whose popping of prosecco corks and parading of an anatomically exaggerated blow-up mannequin was hard to ignore; there was also a group of male models sunning themselves noisily behind us (I’m working at a fashion company, so this is less unusual than it may sound), and three young women were smoking some interesting cigarettes upwind of us. Nonetheless, our retrospective was a success despite occasional distractions.
2. Characterise the sprint
Ask everyone to spend a couple of minutes coming up with three words to characterise the sprint; then go clockwise round the circle and ask each person to tell the team their three words.
Observations: It’s surprising how much difficulty people have sticking to three words; the important focus of this task is not the three-word limit, but getting a concise summary of the sprint.
3. Thank your neighbour
Moving anticlockwise around the team, ask each team member to thank their neighbour for something they have done during this sprint.
Observations: Our Scrum Master broke the rules by thanking the whole team for their efforts. The focus of this task is to generate a positive mood across the team, and it’s important that no one misses out on individual thanks, so I asked him to thank his neighbour for something as well.
4. Describe what went well
Hand each person three cards, and give them three minutes to write down three things that went well during the sprint. Going clockwise round the circle from a different starting point, ask each person to read out their three successes.
Observations: It’s useful for the facilitator to observe and comment on common themes, as this can help reinforce good practice.
5. Describe what could improve
Hand each person three more cards, and give them another three minutes to write down three things that could have gone even better during the sprint. Then go anticlockwise round the circle and ask each person to read out their three improvements.
6. Group the improvements
Instead of arranging the cards on a whiteboard (which isn’t practical in the park), appoint a champion for each improvement. Ask the first person to choose one of the improvements they suggested, and then get everyone else to hand this person any cards that describe a similar improvement. Keep running round the team until each team member has just one group of cards.
Observations: There’s a chance that you’ll end up with more themes than team members, in which case you’ll have to make a decision to drop some of these themes; in our case we had fewer common themes than team members, so we didn’t have to do this.
7. Select and discuss the most common themes
Rather than dot-voting, which again is impractical in the park, select the commonest themes. Ask everyone with any cards to take a step into the circle. Then ask everyone with just one card to take a step out again; then everyone with just two cards; then three, and so on until just three people are left in the inner circle.
Then have a three-minute discussion of each of these suggested improvements, with the focus of identifying at least one action per theme for the next sprint. Ensure someone is assigned to each action.
Observations: We ran slightly over the three minutes assigned to each theme, but this wasn’t a problem; if we hadn’t had a time limit, I suspect the conversation would have been much less focused.
8. Round off the retrospective
Finally, going round the circle clockwise, ask everyone to describe how they felt about the retrospective itself.
Observations: The feedback was very positive. The team had clearly enjoyed the opportunity to get out of the office, and they felt that the session had been successful: everyone was engaged and we came up with some good actions.
Feature Triggers
When we’re continuously integrating code changes, we may want to avoid prematurely exposing new behaviour to our users. A popular technique is to use Feature Toggles to turn a feature on based on a configuration setting. These are certainly useful, and are great for gradually rolling features out, but they involve introducing code into our systems that is not relevant to the system itself, and if poorly implemented can lead to us bloating our codebases with vapourware.
There is another technique that I think deserves more attention; I refer to this as Feature Triggers. The idea is to use an intrinsic part of your system to enable the new feature, rather than an extrinsic piece of configuration. The discipline lies in identifying the final change that will trigger the feature.
In his write-up of Feature Toggles, Martin Fowler describes a version of this technique that focuses on UI changes:
Release toggles are a useful technique and lots of teams use them. However they should be your last choice when you’re dealing with putting features into production.
Your first choice should be to break the feature down so you can safely introduce parts of the feature into the product. The advantages of doing this are the same ones as any strategy based on small, frequent releases. You reduce the risk of things going wrong and you get valuable feedback on how users actually use the feature that will improve the enhancements you make later.
If you really must hide a partly built feature, then the best way is to build all of it save the UI entry point and add that UI in a single release cycle. This way the non-ui code is fully integrated with everything else, but nothing is visible or used until you add the last bit at the end.
(Emphasis mine)
However, we needn’t restrict this technique to the UI. In particular, if you are practising Feature Pull, then the trigger should be the last piece of work in the chain, rather than necessarily a UI change.
Let me give an example.
I worked on a system that sold recorded music on line, and we wanted to start selling high-resolution tracks. The system had theretofore only supported a single audio format per track, usually 320 bps MP3, so the UI simply presented the customer with a ‘Add to basket’ button. Now we wanted to present them with a choice of, say, MP3 or FLAC, at different prices.
Our feature flow worked something like this:
- When a customer has a track as a FLAC in their library, then they can download and listen to it with a FLAC codec.
- When a customer has a track as a FLAC in their shopping basket, and they check out, it ends up in their library.
- When a customer on the website clicks to add a track to their basket as a FLAC, then the track ends up in their shopping basket as a FLAC.
- When a track is in the catalogue with a FLAC format, then the website shows that it can be added to basket as a FLAC.
- When we have negotiated FLAC rights for a track that is ingested, then that track ends up in the catalogue with a FLAC format.
Now, each of these steps could be implemented in turn, using the principle of fake it till you make it to drive out the interface and schema changes as we went along. Crucially, we maintained existing behaviour for existing data. In the case of the UI, this meant that when a track was available in a single format, we just showed an ‘add to basket’ button with format-agnostic behaviour, whereas a track with multiple formats triggered a slightly more complex UI whose buttons had content-specific behaviour. Notice that in this case the UI was not the last piece of the puzzle: it’s behaviour was triggered by upstream data.
Instead, the entirety of this new feature was triggered by enabling the new ingestion rules—which then enabled multiple formats in the catalogue—which then enabled multiple formats on the website—which then enabled tracks to be added to the basket with specified formats—which then enabled format-aware tracks in the library—which then enabled format-aware downloads. All through this we limited the vapidity of the feature by keeping the code branching to a minimum and by releasing continuously, which meant that new code was almost immediately in use in production.
I’m sure this technique is widespread, and I think it’s worth reminding ourselves that the trigger needn’t be in the UI: by combining this technique with Feature Pull, we can often do without feature toggles and write more focused code.
On Pull, Entropy and Bottlenecks
A common scenario
Teams tend to work on features from the entry point in, pushing the behaviour through the system. In my experience this is almost always a bad idea, and I’m going to explain why, and offer some alternative approaches.
Let’s consider a fictional business: Sangeeta’s Sheet Music.
Sangeeta has focussed hitherto on selling piano music through her website and Android and iOS apps, but she’s now keen to corner the string quartet market. Now, string quartet music is sold in both score format and as parts, so there will be a certain amount of work to store information on available stock and to allow customers to choose which format they want to purchase.
Like in many successful startups, Sangeeta’s tech team has grown to some fifteen nmembers, who have been split into Catalogue, Payment, Frontend and Apps teams. She sits down with representatives from these teams, and they agree that Charlie and his team will start work on storing available formats in the Catalogue.
Two-and-a-half months later, Charlie’s team have updated the database and API schemata, and Penny and her team have enabled the shopping basket and checkout process to process different formats at different price points. At this point they approach Fatimeh from Frontend and Alejandro from Apps: “We’re nearly ready to go live with our new string quartet catalogue: we’ve got the data schema in place, our stock room is filling up with catalogue, so we just need to you to update the front end and apps, and we’re ready to start selling!”
Alejandro and his team are busy fixing a long list of bugs from the new iPhone, and won’t be able to start work for at least six weeks, but fortunately Fatimeh’s team have some capacity, so they start work. It soon becomes clear that some of the initial assumptions don’t work out on in real life: use case research shows that many orders are likely to be for all four string parts, and the shopping basket API doesn’t support this easily. Furthermore, they quickly realise that some of the catalogue doesn’t fit into the expected model: Schönberg’s String Quartet №2 has a part for soprano voice, and George Crumb’s *Black Angels* is only available in score format.
After some tedious toing and froing between the Catalogue, Payment and Frontend teams, they are happy with the functionality, but Alejandro and his team still aren’t ready implement this functionality in the apps, and Sangeeta doesn’t want to launch functionality on just the website. The feature has now been in progress for four months, yet it’s still vapourware, and the stock is sitting unsold in the stockroom.
On the face of it, there are two problems here: cross-team coordination and poorly specified behaviour. I contend that these are both results of a deeper problem: starting work in the wrong place.
Another approach
At 7digital I worked with Neil Kidd on a feature rather like this. We had observed how frequently these two problems arise, and decided we needed a different approach: we would start from the exit point and work backwards.
To do this we scripted some end-to-end User Journeys to describe the expected behaviours, stubbing out interim steps. We then worked with the responsible team to implement the final step in the process against canned data. We were then able to work backwards, working on each component in the chain to enable it to provide real rather than canned data. At each step in the process, we of course ensured that existing behaviour was unchanged, and that the new behaviour only kicked in when the previous component was ready to support it.
By starting with end-to-end User Journeys and considering the data output, we avoided nasty surprises, and our by triggering the new behaviour with data, rather than feature switches, we enabled the functionality to go live as soon as all the pieces were in place, which helps avoid producing vapourware.
I refer to this approach as Feature Pull, acknowledging the Kanban approach of pulling inventory through a manufacturing process, or work-in-progress through a development process, and I’ve been fairly vocal in encouraging colleagues to embrace it.
This concept tallies well with BDD and TDD principles, as it favours thinking about user journeys as an initial step, and testing from the desired outcome backwards.
Entropy
I wanted to find a way to illustrate this principle to colleagues outside the context of a feature, and felt a game might be the answer. I attempted to run such a game at 7digital, but the clearest lesson we drew from it was the benefit of communication, rather than anything specific to the flow of feature work.
And so at Socrates BE I ran a session in which I asked for help in coming up with a such a game.
As it happens, the interesting part of this discussion was not the proposals for the game itself, but the last few minutes, in which we started to examine *why* a pull system is often most appropriate. We alighted upon the notion of Entropy and the idea that features in a system should flow from the point of least to the point of most entropy.
If we consider that interactions with a system can often be modelled as a decision tree that branches over time, it is not surprising that the point of most entropy is at the end of a user journey, and it makes sense to start at that point.
Theory of Constraints
This felt like a nice take-away point, but I felt there was something missing. This morning I was reflecting on these ideas again, and the answer came to me: the point of most entropy is the bottleneck in the system, and by starting at this point we are subordinating our work to the bottleneck.
Think back to Sangeeta’s Sheet Music. The most obvious bottleneck in this system is clearly Alejandro and the Apps team, as they have so much other work to do. It therefore makes sense to start work by subordinating to this bottleneck, not building up inventory by implementing features that depend on this team, but rather finding ways to enable this team to overcome its workload problems.
Once the Apps team bottleneck has been tackled, then another bottleneck appears, this time one of planning: the team needs to understand and implement all the scenarios that are needed before the minimum viable feature can go live. This may involve a fair amount of planning and conceptual work, but again there is no point in doing upstream development work until this has been tackled.
Three metaphors
So here we have three metaphors for understanding how to approach features. If you like, you can focus on Pull, and start at the exit point of your system; if your mind turns to theoretical physics, then consider the entropy of your system, and start at the point where it is greatest; and if the Theory of Constraints gets you excited, then look for the bottleneck and subordinate to it. But whatever you do, please don’t just start coding!
Experience report: running the 7digital Technical Academy
Between September 2014 and April 2015 I ran 7digital’s Technical Academy: a half-year training scheme to give a new cohort of software developers a grounding in software craft. Here are a few thoughts on how it went.
Choosing candidates
Our four participants came from diverse backgrounds: two were graduate recruits, one with a degree in Forensic Biology, the other in Physics. One internal participant was a member of our QA team, while the other joined us from the Client Relations team. Three of them were women.
We explicitly opened the external vacancies to candidates with non-Computer Science backgrounds, and this gave us access to a more diverse field of candidates. We wanted candidates who had shown an interest in software development, but didn’t demand any industry experience: programming a spreadsheet to analyse coursework data, working with Access databases in a summer job, or creating a personal website could all be evidence of this interest.
At 7digital, the technical test is to pair programme on a software kata. The usual aim is to assess the candidate’s aptitude for TDD, and to see how well they work with another person. We used the same test for our Technical Academy candidates, but were fully aware that they were unlikely to have any TDD experience. Instead, we used this as an opportunity to find out how quickly they could pick up these techniques. At the beginning of the 45-minute exercise, I took a fairly directive role as Navigator: explaining what a unit test was, how an assertion worked, and telling the candidate what to type. By the end of the session, our successful candidates had demonstrated that they had picked up all the key concepts, and were capable of writing new tests with much less direction. This aptitude for learning was the most important quality we sought in our candidates.
Joining a team
Each participant joined one of the development teams, where they were expected to pair with a more experienced developer from the beginning. In addition, we ran several learning sessions a week, and expected them to work on a personal project over the six months of the academy. We suggested that they divide their time between academy and team work roughly 50:50.
New recruits at 7digital are assigned a Mentor, who has a pastoral role, catching up with them frequently and helping them settle in. We arranged Mentors for our two new joiners, and all four participants were also assigned a Tutor, whose role was to help with with any immediate technical issues during the academy, particularly in the project work.
Learning sessions
We organised a range of learning sessions, and implemented a pull system, choosing topics to address questions that had recently arisen during the participants’ project and team work. We met every Monday morning to finalise that week’s sessions, and to start planning for the following week. Apart from keeping a list of potential topics, we didn’t plan further than two weeks ahead.
Some of the sessions were hands-on practical exercises. At the beginning of the academy, we got the participants into the habit of TDD by running mob programming kata sessions: I would sit with the four of them around a computer, and they would pass the keyboard around in 5 minute sessions. As questions arose, we would sit back from the keyboard to discuss them. I introduced programming concepts and features of the IDE (Visual Studio with ReSharper) as we went along, so they could become familiar with running tests, elemental refactoring shortcuts and so forth.
Some of the sessions were more theoretical, and we enlisted colleagues to give talks and demonstrations of topics where they had particular expertise. Initially I took responsibility for organising this, but as our participants’ confidence and familiarity with their colleagues grew, I handed responsibility for this over to them.
Near the end of the course we ran some presentation sessions: one on the SOLID Principles and a couple on Design Patterns. Each of the participants chose a principle or a pattern and gave a short presentation to the others explaining it (I pitched in and gave a presentation on the Liskov Substitution Principle so as to make up the numbers). Some of the participants were rather nervous about doing their first presentation, but the results were impressive, and it was great to see their confidence boost once we had run these sessions.
Finally, we arranged a weekly reading group. During the first three weeks I led the group in a discussion of chapters 13–15 of Eliyahu Goldratt’s *The Goal*, which give the participants an accessible introduction to the Theory of Constraints, which underlies much of the Kanban methodology. After this, they started reading material on code design, and soon told me that they were happy to continue these sessions without me. The approach of discussion one or two chapters each week seemed to work very successfully, as they could engage properly with the issues.
The project
The project was a key focus of the participants’ learning. We sought ideas from colleagues across the company, and then let the participants choose a project. The person who had suggested the project acted as client, and was expected to be available to answer any questions, and to come along to demos to see how the work was progressing. Tutors were also expected to help out and give guidance.
We emphasised iterative development in the projects, encouraging our participants to get a walking skeleton running and deployed to a server on the network through our CI server (TeamCity) as soon as possible. Only then did we suggest they start fleshing out functionality. We also stressed the importance of doing TDD from the start. This approach front-loaded many of the difficulties in the projects: participants found it hard work to get going, but once they had a deployed project, they found they could pick up speed in their work.
We arranged frequent demos in which the participants could show off the progress they were making in their projects, and get feedback from their Clients and the rest of the academy.
Unlike in their team work, where pair programming was usual, much of the project work was done solo. There was a fantastic moment of revelation about a month before the final demonstrations, when it became clear that several of the participants were struggling with technical issues, and were worried whether they would be able to complete their projects in time. I suggested that rather than working solo, they work together in rotating pairs to solve these problems. At our next catch-up they reported that they had solved the problems and made more progress in those short pairing sessions than they had in the previous weeks’ solo work. For our participants to learn first-hand the value of pair programming less than six months into their careers was wonderful.
What worked
- We found fantastic recruits and excellent internal candidates.
- Pull learning was very successful.
- The projects gave a good context for learning.
- The participants developed an excellent team spirit and ability to self-organise.
- The focus on XP and Software Craft skills equipped the participants with skills that are lacked by many developers with many more years’ experience: we created four exceptional software developers.
What didn’t work as well
- Having a Client for each project worked better in theory than in practice; some didn’t even attend the demos.
- Participants sometimes had difficulty getting a good balance between team work and Technical Academy work, and there were weeks in which their projects got no attention.
Would I do it again?
In a heartbeat. It was an amazing experience, and seeing the quality of our graduates is incredibly gratifying. I have now left 7digital, but there is a new Technical Academy kicking off as I write this, led by one of last year’s cohort. I wish them the very best of luck!
Edited on 20 Nov 2015 at 15:04 to mention the presentation sessions.
Edited on 26 Sept 2017 at 16:31 to remove a reference to an individual whose political views are not consistent with the spirit of this blog.
Watch out for exceptions in NUnit TestFixture constructors
Don’t do anything that might throw an exception in a TestFixture constructor: your tests will be ignored, rather than failing.
My colleague Chris and I have been playing around with polymorphous testing, taking advantage of the fact that you can pass parameters to an NUnit TestFixture like this:
[TestFixture(parameter1)]
[TestFixture(parameter2)]
public class BazTests{
public BazTests(Blah parameter){
...
}
}
We wanted to use these parameters to pass in a collaborator, which would then enable us to run the same tests against different implementations of the same interface. However, we immediately encountered difficulties, as the TestFixture attribute will only take parameters that are compile-time constants, which rules out directly passing in a collaborator class. If you are adding parameters to a Test, you can use the TestDataSource attribute to pass in more complex arguments; however, no such attribute is available for TestFixtures.
To get round this, we created a static factory that would then create our dependency according to the enum value passed in:
[TestFixture(TestSetting.Foo)]
[TestFixture(TestSetting.Bar)]
public class BazTests{
private ITestHelper _testHelper;
public BazTests(TestSetting testSetting){
_testHelper = TestHelperFactory.Create(testSetting);
}
}
However, some of our implementations of ITestHelper require a connection string, and when it is not present, a NullReferenceException is thrown. We would have expected some of our tests to fail because of this, but instead they were just ignored, and as TeamCity disregards ignored tests, this meant that our builds were treated as successful.
It suspect the tests are ignored because exceptions in the test constructor are not sent to TeamCity as errors; rather the exception must happen in one of the attribute-decorated methods.
The code gave us the correct failures once we had rewritten it like this:
[TestFixture(TestSetting.Foo)]
[TestFixture(TestSetting.Bar)]
public class BazTests{
private TestSetting _testSetting;
private ITestHelper _testHelper;
public BazTests(TestSetting testSetting){
_testSetting = testSetting;
}
[TestFixtureSetUp]
public void TestFixtureSetUp(){
_testHelper = TestHelperFactory.Create(_testSetting);
}
}
