LINQ is a wonderful thing. It significantly enhances the expressiveness of C#, and provides lots of other benefits as well. Unfortunately, it comes with a cost.

It can hide things from you.

One of the real dangers is out-of-control time complexity. How many times will the following code perform the .Distinct() operation?

var rnd = new Random();
var values = Enumerable.Range( 1, 1000).Select(r => rnd.Next(10));

var uniqueValues = values.Announce( "Performing Distinct() operation...").Distinct();
if (uniqueValues.Count() > 2) uniqueValues.First().Dump();

If you answered ‘two’, congratulations! My output looks like this:

Performing Distinct() operation...
Performing Distinct() operation...
3


Of course, ReSharper warns you about this: “Possible multiple enumeration of IEnumerable”. ReSharper won’t catch all sins, though. I ran into something similar to this recently:

// Example data
var rnd = new Random();
var values = Enumerable.Range(1, 10000).Select(r => rnd.Next(10)).ToList();
var otherData = Enumerable.Range( 1, 10000).Select(r => rnd.Next(20)).ToList();
var counter = new object();

// Problem code
var uniqueValues = values.CountCalls(counter).Distinct();
var otherDataWhichMatchesValues = otherData.Where(od => uniqueValues.Contains(od));
otherDataWhichMatchesValues.Count().Dump();
MyExtensions.GetCallCount(counter).Dump();

That took 19 seconds to run, and made 10,000 calls to Distinct()! Can you see what’s going on? The Where() operation is testing each entry in otherData against uniqueValues – enumerating uniqueValues once for every entry in otherData – and ReSharper 8 doesn’t warn you about it! If you’re used to seeing that warning whenever you try to enumerate an IEnumerable more than once, you might be tempted to think that it won’t happen. You would, of course, be wrong.

A Distinct() operation runs in O(n) time, Where() introduces an additional O(n), and .Contains depends on the underlying data structure – which in this case, is a List. So our overall operation is running in O(n^3) – that’s a real performance killer on any non-trivial data set.

Dropping a .ToList() after the .Distinct() reduces our run time to 3 thousandths of a second, and reduces our Distinct() operations to one. Much better! (Well, at least, it seems that way for now.)

I have a habit which ReSharper doesn’t like much. I usually avoid methods like Count(predicate), and prefer to chain a Count() on the end of a Where(predicate). One of the reasons I do this is that I think it makes it clearer which calls are LINQ queries, subject to multiple evaluations, and which calls will cause evaluations. Of course, that doesn’t help if you don’t spot that .Distinct() is in the LINQ namespace in the first place!

It’s easy to forget about things like time and space complexity, but there’s a reason you learned that stuff: It’s important! Whenever you write a loop or make a LINQ call, something in the back of your mind should be thinking about how nested the current operation is, and how much time and space complexity you’re layering on top. That’s not to say that you should necessarily optimise everything, but it may help you to spot problems before they get out of control.

There are two bigger-picture lessons to learn from this.

Algorithms and Data Structures

The real root of the problem, in this case, came from not thinking through the algorithm or selecting appropriate data structures. Rather than trying to decide on an outcome first, the programmer has worked through a series of operations against the available data, until it has ended up in the right shape. You can picture the thought process:
– I need to get the entries in otherData which match entries in the values list.
– That will be slow, so I’ll call distinct on values.

The fix – adding a call to ToList() after calling distinct – has unfortunately introduced a more subtle performance bug. It works well for the test data set, but it won’t perform as well if values is sparse: if Distinct() removes a lot of duplicates, then we’ll see a performance improvement, but if there are few duplicates, the original problem the programmer was trying to fix will remain. Let’s measure.

// Example data
var rnd = new Random();
var dataRange = 10;
var values = Enumerable.Range( 1, 10000).Select(r => rnd.Next(dataRange)).ToList();
var otherData = Enumerable.Range( 1, 10000).Select(r => rnd.Next(dataRange)).ToList();
var counter = new object();

// Operation
for ( int i = 1; i < 100; i++) {
     var uniqueValues = values.CountCalls(counter).Distinct().ToList();
     var otherDataWhichMatchesValues = otherData.Where(od => uniqueValues.Contains(od));
     otherDataWhichMatchesValues.Count().Dump();
}

We’ve now introduced a variable – dataRange – which will roughly control how many duplicates we’ll get. This code roughly parallels our original, with the ToList() fix (run through numerous iterations to exaggerate the timings). As is, it completes in 0.6s, but if we change dataRange to 1000, the run-time increases to 5.4s.

Consider the operations we want to do. We’re looking to build the ‘values’ dataset in the first place, and then we’re going to make many calls to .Contains() against it. While the time complexity of an insert operation on a list is O(1), Contains() is O(n). What we really want is a data structure which is O(1) for both insert and contains operations – and that’s a hash. So the fix we should really make is to change the values dataset to a HashSet, and drop the distinct operation altogether:

var rnd = new Random();
var dataRange = 10;
var values = new HashSet<int>(Enumerable.Range(1, 10000).Select(r => rnd.Next(dataRange)));
var otherData = Enumerable.Range(1, 10000).Select(r => rnd.Next(dataRange)).ToList();
var counter = new object();

for (int i = 1; i < 100; i++) {
     var otherDataWhichMatchesValues = otherData.Where(od => values.Contains(od));
     otherDataWhichMatchesValues.Count().Dump();
}

Now the run time is around 0.1s, regardless of the value of dataRange. As we can see, it’s not only faster in the sparse case, but it’s faster even than the ideal case with many duplicates – which we timed at 0.6s with 100 iterations.

I see a lot of developers who have a favourite data structure – usually a list or an array – and rarely or never make use of other data structures. If a piece of code has emerged as a target for optimisation, you may as well optimise it properly the first time around. Throwing a .ToList() onto the end of a LINQ expression is kind of a code smell: it indicates that you may not have selected the right data structure in the first place. Unless, of course, you’ve looked at the time complexities of the operations you need, and a list is a good fit: in which case, by all means ToList() it! If you’re feeling particularly lazy, want to ignore everything I’ve said so far, and roll the dice, you could even just .ToRandomCollectionType() it.

Over-reliance on tools

As a final thought, I want to caution people against over-reliance on tools. The problem is not that the developer relied on ReSharper to spot multiple enumerations; even if they’re used to spotting them without it, this is an unusual operation (which is why the ReSharper rule didn’t catch it – and yes, there is an open issue to address exactly this situation). The problem emerges when the developer not only starts to rely on ReSharper to pick up rules like this one, but starts to assume that code without ReSharper warnings is good code. That goes hand-in-hand with the fact that ReSharper isn’t always able to fix the warnings appropriately: in this case, even if ReSharper had caught the problem, its solution – to enumerate the IEnumerable to a list – wouldn’t have been the appropriate (or at least, the best) solution.

NHibernate really is a fantastic ORM… unless you use it badly. Or unless you use it kinda OK. Or unless you use it almost-but-not-quite-perfectly. Then it can be a right pain in the neck. There are a lot of things you can get wrong, and those things can cause you a world of pain.

The particular pain I dealt with recently was memory pressure.

One of the most crippling things you can do to a server is to fill up its RAM. Once your RAM is full, your OS has to start copying things out of RAM onto disk, and then back into RAM when you need to use them again. Disks are slow – much, much slower than RAM – and this is going to hurt your performance badly.

Picture this. You’ve done all the right things. Your software is built using a loosely-coupled Service-Oriented Architecture. You have a website, and it hands all sorts of tasks off to a separate service layer. You have a second service handling various data import tasks. As your load increases, it’s going to be very easy to scale horizontally: you can move your services off to separate servers, and the only thing you need to do is update a few network addresses. Once you expand beyond what you can handle with four servers (those three functions plus a separate database server), you can load-balance each function individually.

You’ve also decided to handle multiple tenants with multiple databases. This one isn’t the right decision in every situation, but there are lots of times when it makes sense, particularly if you’re holding a lot of data for each client. It makes it trivial to archive individual clients off. It makes it easy to offer different tiers of backup. It stops your row-counts from getting too high, and it isolates small clients from the performance headaches of the massive data sets maintained for larger clients.

NHibernate is going to kick you in the teeth for doing this.

We’ve been watching the problem approach for a while now. The base memory overhead for each process soared past a gigabyte some time ago. As our client list headed towards a hundred, our memory overhead headed towards two gigabytes per process. I didn’t need to run a memory profiler to know where the problem was (although I did use one to confirm my suspicions). The culprit was the NHibernate session factories. A single session factory can run towards 20 MB. With fifty clients, that means you have a full gigabyte of RAM filled with nothing but session factories. I didn’t want to have to start scaling horizontally early just because of this, and after all, this gigabyte of memory consisted of lots and lots of copies of 20 MB structures which were identical except for a single string: the database connection string. That’s horribly wasteful. (Actually, there were other differences, but we’ll get to those.) I also couldn’t start disposing of session factories once they hadn’t been used for a little while: these things take a while to construct, and we can’t let our users sit around for several seconds when they log in for the first time in a while. I needed to start re-using our session factories.

There are at least two approaches you can take here. The one I chose has two caveats: firstly, that you’re using NHibernate.Cfg.Environment.ReleaseConnections = “on_close”, and secondly that you’re not using stateless sessions at all. We’ve been moving towards ditching stateless sessions for some time anyway, because stateless sessions don’t support listeners, so the second requirement wasn’t a problem for us. The first setting is a bit more troubling, because it’s legacy behaviour: rather than letting NHibernate manage connections using one of its newer strategies, it forces NHibernate to provide a connection when a session is first opened, and use that connection for the duration of the session. This was acceptable because we were already using the legacy setting, for reasons undocumented in either code comments or our source control history. I haven’t looked into the costs and benefits of this legacy mode compared to the other strategies.

So, let’s dive into some code. First of all, you’re going to need to set your connection provider:

       cfg.SetProperty(Environment.ConnectionProvider,
                       typeof(SharedCompanyConnectionProvider).AssemblyQualifiedName);

Then, seeing as there’s no such thing as a SharedCompanyConnectionProvider, you’ll need to implement it!

        public class SharedCompanyConnectionProvider : DriverConnectionProvider
        {
            protected override string ConnectionString
            {
                get { return NHibernateSessionManager.Instance.DatabaseSettings.GetCurrentDatabaseConnectionString(); }
            }
        }

If that looks a bit scary, good. If not, let me explain. Your connection provider is no longer thread-safe! It’s relying on a singleton which serves up a connection string. This is dangerous code, and you need to be careful how you use it. (Don’t even think of using this without putting some tests around it – see later in this post.)

Now, on to wherever it is you build your sessions. Mine looks something like this:

            private static readonly object CompanySessionFactoryLockObject = new object();
            ...
            lock (CompanySessionFactoryLockObject)
            {
                var sessionFactory = NHibernateSessionManager.Instance.GetSessionFactory();
                NHibernateSessionManager.Instance.DatabaseSettings.SetCurrentDatabaseConnectionString(databaseGUID);
                ISession session = sessionFactory.OpenSession();
            }

I’ve removed a lot of the detail, but that should give you the gist of what’s going on. The key component here is the lock() line. Now that our connection provider isn’t thread-safe, we have to ensure no other threads interrupt between setting the connection string on the singleton, and creating the actual session (at which time the connection provider will provide a session with the current connection string).

The final step in the process is to make sure you have some thorough testing around what you’re doing. The risk of getting it wrong is that your session factory hands you a connection to the wrong database, and that could be very bad. I’m not going to run through the entire test setup, but it’s certainly not a unit test – this thing runs in a test suite which uses a real database instance and creates (in the case of this test) five complete databases which we’ll be accessing from various threads.

        private volatile static string _assertFailed;
        private const int NumThreadsPerDb = 2;

        [Test]
        public void HammerMultipleDatabasesSimultaneously_BehavesWell()
        {
            List<Thread> runningThreads = new List<Thread>();
            foreach (var coGuid in companyGuids)
            {
                for (int i = 0; i < NumThreadsPerDb; i++)
                {
                    var thread = new Thread(StartHammering);
                    runningThreads.Add(thread);
                    thread.Start(coGuid);
                }
            }
            while (runningThreads.Any(thread => thread.IsAlive))
            {
                if (_assertFailed != null)
                    runningThreads.ForEach(thread => thread.Abort());
                else
                    Thread.Sleep(1000);
            }
            if (_assertFailed != null) Assert.Fail(_assertFailed);
        }

        public void StartHammering( object companyGUIDObj)
        {
            // nb don't assert on a thread. We're set up to set a message into _assertFailed instead.
            var CompanyGUID = (Guid)companyGUIDObj;
            string expectedDbName = CoDatabaseNames[companyGuids.IndexOf(CompanyGUID)];
            try
            {
                Entity entity;
                using (var session = NHibernateSessionManager.Instance.GetNewSession(CompanyGUID))
                {
                    // Set up the entity with some unique data
                    session.Save(entity);
                }
                for (int i = 0; i < NumTests; i++)
                {
                    using (var session = NHibernateSessionManager.Instance.GetNewSession(CompanyGUID))
                    {
                        if (!session.Connection.ConnectionString.Contains(expectedDbName))
                            throw new Exception( "Got a connection for the wrong database!");
                        var ent = session.Get<Entity>(entity.GUID);
                        // Check some unique thing about the entity. Change it to something else for the next iteration.
                    }
                }
            }
            catch (ThreadAbortException) { }
            catch (Exception ex)
            {
                if (!ex.ToString().Contains( "ThreadAbortException"))
                    _assertFailed = ex.ToString();
            }
        }

There’s a lot going on there. The key theme is that we’re creating a bunch of threads, and each thread is assigned to a particular database. New sessions are continuously created, and then queried to ensure they contain the expected object. If the object is not found, or the session has the wrong connection string, then something has gone wrong, and the whole system isn’t behaving in a thread-safe fashion.

Note that in a multi-threaded test situation, you cannot just throw an exception if something goes wrong – you need to pass information about the failure to your primary thread.

One final (and important) step is to ensure the test does fail appropriately if the system doesn’t behave as expected. Remove the lock statement around your session creation code and run the test; you should see it fail. Adding the lock back in should fix it.

Agile. Kanban. Lean Startup. We have more and more ways to think about the software development process, and our industry is getting better and better at what we do. We’re wasting less time, and delivering a more customer-focused product. New businesses have great tools to build great teams. We know how to manage source code, and some people have even worked out how to manage changing data models! Issue tracking is, by and large, a solved problem, and we have tools to keep our software fully tested and behaving the way it should.

What I don’t see enough focus on these days is the code we write. Sure, we’re all doing peer reviews (right?) and we think we manage code quality through unit testing. What worries me is that, usually, this just ensures that our code works. It does nothing to really manage quality. What’s wrong with this code?

class CreditCard
{
  private Decimal _Limit;
  //...
  public bool RaiseLimit(Decimal amount)
  {
    _Limit += amount;
    //....
  }
  //...
}

This is a blatant example, but it’s a problem I see all the time. The code looks clear enough, and it’s going to work. It will build, pass your unit tests, and there’s every chance it will pass peer review – but it shouldn’t. The problem is, there is a risk that someone will do this:

Decimal oldLimit = card.Limit;
Decimal newLimit = ...;    // Code to calculate the new limit
card.RaiseLimit(newLimit);

… and that would be awful! The caller has gone to the effort of calculating the new limit and passing it to RaiseLimit(), but that’s not what RaiseLimit() expects: it expects to be passed the amount the limit should be increased by. How do we solve this? There is a simple change we can make to reduce the risk this will happen. Our improved code might look like this:

public bool RaiseLimit(Decimal increaseAmount)
{
  _Limit += increaseAmount;
  //....
}

Functionally, nothing has changed – but we’ve provided an extra clue to a developer calling our function. In these days of intelligent IDEs, coders usually get to see the method signature that they’re calling. We should endeavour to take every possible advantage of IDE features like these to streamline code creation and reduce the chance of errors.

This is one simple example from a whole spectrum of things you can do to optimise code-writing. I have wasted countless hours reading through classes trying to work out how to use them. Here’s a recent example I ran into:

class ConfiguredTask
{
  public string ConfigFolder;
  public string TaskExe;
  public void WriteMainConfig(DateRange dates, string configName) {...}
  public void WriteSecondaryConfig(DateRange dates, string configName) {...}
  public void RunTask(string path, DateRange dates) {...}
  public ResultSet GetResults(DateRange dates) {...}
  //....
}

To use this effectively, you need to know quite a bit about the internal class. For a start, you need to know that you need to write the main and secondary config files before calling RunTask() – not as obvious as you might think, as there are dozens of other public methods and properties on this class. Second, you need to know that the two functions to write config files are expecting filenames with full path information, and they need to be different, but in the same folder. Third, you need to know that RunTask() persists the results of the task both into the database – something I didn’t want – and leaves output in the folder referenced by ConfigFolder. TaskExe must contain the name of the file to run, but must not contain any path – the executable must be in the path referenced by ConfigFolder. That wasn’t even the end of it! The code to run a task used to look like this:

task.WriteMainConfig(selectedDates, @"c:\main.cfg");
task.WriteSecondaryConfig(selectedDates, @"c:\second.cfg");
task.RunTask(task.ConfigFolder, selectedDates);
ResultSet results = task.GetResults(selectedDates);

Keep in mind that, if you didn’t have an example to hand, you had to read a fair portion of the class itself to make sure you had everything right! After I was finished with it, the class looked like this:

class ConfiguredTask
{
  public string PathToExecutable;
  public string ResultsFolder;
  private void WriteMainConfig(DateRange dates) {...}
  private void WriteSecondaryConfig(DateRange dates) {...}
  public ResultSet RunTask(DateRange dates,
    bool persistResults=true, string WorkingFolder=null) {...}
  public ResultSet GetResults(DateRange dates) {...}
  //...
}

Just through a little re-factoring and a handful of code tweaks, the caller now knows to give a full path to the executable (which can now be anywhere), they don’t need to know about writing config files (it’s done during RunTask()), they know that results are persisted by default but they can override that behaviour, they know they can provide an optional working folder, and they don’t have to call GetResults() after every call to RunTask() when they want the results of that task.

I’ve taken a class which needed quite a bit of reading to work out how to use it, and turned it into a class you can use with no more information than what IntelliSense shows you as you code. The code to run a task and get the results now looks like this, instead:

ResultSet results = task.RunTask(selectedDates);

I hope you can see why this would save a future developer time! We should strive for all of our classes to be like this. The idea of ‘self-documenting code’ doesn’t quite capture this: the whole point is to not have to read the code at all. I prefer the term ‘Intuitive Interface’. As a developer, you should aim for all of your classes to have this kind of intuitive interface. Think of the questions a caller might need answered, and then answer them – in your method signature if possible, and if not, in a method comment (ideally, in a fashion that leverages the IDE you’re working in):

/// <summary>
/// If the instrument is not running, set the passed-in mode and start it running.
/// If the instrument is already running it will *NOT* change the mode.
/// </summary>
private void EnsureInstrumentIsRunning(InstrumentMode mode)
{
  // ...
}

Even in that example (which I took from a current project), the first line is really extraneous – it’s repeating information that’s already available in the method signature.

There are all sorts of things to get right and wrong when you’re building software. There’s a lot to think about when you’re kitting up your team. What OS should desktops run? What tools will we use? How much should we spend on office chairs? One of the most important decisions you’ll make is what sort of monitor your developers get, and how many.

I’m not joking about. Your developers are going to be spending a fairly large fraction of their professional lives looking at their monitors. You really want to make it a pleasant thing to do. The standard I tend to see these days is to give developers a pair of 22″ wide-screen monitors. That might sound excessive, and other staff might be resentful, but it’s really not that costly these days. A good 22″ wide-screen might cost $500, as compared to a cheap small screen which costs, say, $350. Going for twin 22″ screens means spending $1000 instead of $350. Given that a good-quality monitor will often last upwards of three years, this means you’re spending a little over $200 a year per developer.

So what do you get out of it? Plenty. First of all, your developers will quite likely be more productive. I find debugging code to be much easier when I can have an application running full-screen on one monitor while my debugging tools are on the other. Tuning CSS is easier when I can look at the context in a browser while I’m editing the file. Visual Studio is much easier to use when you can devote an entire screen to code or UI layout while still having all of your explorers, editors, and toolboxes visible on the other. Widescreens make comparing versions of code files much simpler.

At one previous job, one of my screens had a rotating stand, and I ended up turning it on its side and using it mostly for editing code. I could see a lot more context when I had the entire ‘width’ of a 22″ widescreen in height. This tended to play havoc with remote desktop sessions, however. Most annoying was that whenever my manager connected back to his machine from mine, it re-shuffled his entire desktop. Use with caution!

This advice doesn’t just apply to screens. If your developer machines are slow, they’ll end up waiting for things they do frequently. When they have to wait a lot, they’ll find things to do while they wait. You’d be amazed at how much productivity you lose if your developers alt-tab to facebook every time they hit the build button! Fast computers for your developers are vital. Consider getting machines with RAID controllers, and mirroring their drives: developers tend to spend more time customizing their workstation than any other sort of employee, and having to re-build their machine results in more lost time than just any un-committed code changes. A software developer is likely to have all sorts of settings on their IDE; they’ll have lots of little productivity tools installed, like their favourite code editor, and perhaps a little tool to help them build regexs; macros and little time-saving scripts and other things. It will cost them time, and worst of all, it will be frustrating, to have to get everything just right again. Having mirrored drives helps to reduce the chance this will happen. Your other alternative is to have a good backup and recovery system for each developer workstation, but just having RAIDed drives is probably cheaper and easier to manage.

There are lots of little things to do to make your programmers happier where they work. Remember, they spend 40-odd hours a week in front of their PC. The more pleasant it is, the happier they’ll be, and you’ll tend to get more out of them. If the chairs are uncomfortable, they’ll dread coming in to those 8-hour days, and you’ll spend more on sick leave. If the coffee in the office is terrible, your whole programming team will start leaving together two or three times a day for ‘a quick coffee up the road’. Multiply your charge-out rate by the number of programmers in your team by 1/4 (for 15 minutes) by three, and that’s what it costs you every day to not have good coffee in the office.

I’ve seen a lot of workplaces who skimp on the little luxuries because they’re seen as ‘too costly’. You know what’s really too costly? When your programmers see photos of the office of one of their former colleagues, and the double-22″ monitors make your single 19″ look like a cage. When the espresso machine in the common room trumps the tin of International Roast you keep under the sink. When the former colleague gets a huge glass whiteboard, while you give your developers butcher paper and pencils. You know what will happen? You’ll have employees who treat their job as something to get them by until they find something better. They will join, they will hate it, and fairly soon, they will move on.

It’s simply too expensive to spend $350 per machine on monitors when you should be spending $1000.

Indulge me while I take a paragraph or two to get to the point.

I’ve worked as a Software Engineer at a lot of places now. I’ve stayed in some for many years, others only months; if there’s not anything interesting left for me to achieve, I usually start to feel like moving on. I’d like to think this is best for everyone: I get to take my ideas and experience and give a different company the benefit of them, and my (now ex-)employer gets to bring in a new employee with a different set of ideas and experience. This cross-pollination of ideas is a part of the life of software development, but there is a balance. Keep your pool of developers too static, and you risk stagnating. Turn them over too quickly, and you’ll be spending too much money on recruitment and training, and not keeping people around for long enough to see a return on that investment. Get big enough, and you can get these benefits just by letting people transfer within your company: the best of both worlds. That’s not what I want to talk about, though.

I want to talk about the baseline ideas and experiences I bring to a new company. The stuff that I just want to get out of the way so I can start actually innovating.

---

Now, not everyone needs these ideas. In fact, my ideal employer doesn’t need any of these ones. Every time I start somewhere new, these are the things that I hope are already covered, so I can get on with solving interesting problems. These are also some of the things I ask about when I’m interviewing a new employer; they’re problems that I’ve solved over and over, and I know the consequences of not solving them. If I sound a prospective employer out in and interview, I find out that they haven’t solved these problems, and it sounds like there won’t be any support for me solving them when I start, I’ll seriously think about turning down the offer. Some of them are optional; some of them are even inappropriate for some teams. Some of them are so basic that you can’t expect to build decent software at all without them. Every one of them I’ve had to introduce to an established team at some stage or other.

Source Control

This is the number one must-have for software development. When I’m working on a small, personal project which nobody else will ever touch I still use source control. It’s just too simple to set up, and the benefits are too enormous.

One of the first things I learned about source control was that Microsoft Visual Source Safe is not source control. At least, not at the sort of level you want even for a small, personal project, and certainly not at the level you need to build commercial software. I’ve suffered through it with several employers now, and there are just too many things I keep wanting to do that it doesn’t let me. Git seems to be the current darling of the open-source world; it’s free, it’s quick to get started with, and tasks like branching and merging are said to be quick and straight-forward. I haven’t worked with Git yet: I always seem to end up using the well-known staple, Subversion (and earlier in my career, it’s spiritual predecessor, CVS). UPDATE: Git is fantastic, and switching to it has had significant productivity benefits for my team.

If you haven’t solved your source control problem yet (or if you think you have, but you’re using VSS), Subversion is a good place to start. It provides all of the basics, it’s easy to set up, and it does most of the things you’re going to need to do for a small- to medium-sized team. There are plenty of hosted solutions available (with all sorts of optional add-ons), but I really think something this important is worth hosting in-house. It’s easy to do, you’re in control, and you won’t go dark every time your ISP hiccups. Do back your repository up. If you go with a hosted solution, make sure you have some way of backing up the repository yourself in case the worst happens. UPDATE: Using a distributed version control system like Git with an externally-hosted master somewhere like BitBucket or GitHub gives you all the advantages of both worlds.

What are the benefits? Well, for starters, you can see your entire development history. If you introduce a bug but can’t quite remember exactly what you changed, you can see the entire history of a single file, a whole folder, or your entire project. If someone else introduces a bug, you know who to go to to discuss a possible fix. If you really mess things up, you can roll back to a previous version. You can merge in code from multiple developers as you go, even within a single file, and you can branch the code off while you work on significant changes, and then merge it back into the trunk later. If you don’t have a decent source control solution in place, you have all the makings of multiple on-going headaches; if you do, you have a powerful tool on hand.

There’s one thing I’d like to add here, and not everyone will agree with me. I believe in small, frequent, buildable commits. I don’t like seeing a commit which touches 35 source files, has two pages of comments (which still finish with “lots of other little fixes”), and resolves a dozen bugs and nine different feature requests. You lose a lot of the benefits. Keep your commits small and granular; try to fix only one or two things each commit, and keep your repository buildable and passing all of your tests. As a bonus, your team will spend much less time on merging, particularly if they’re updating frequently as well.

Database Version Control

Closely related to source control is database version control. Many large projects end up with some sort of database back end, and that database will grow and evolve with your code. You will end up with various versions of the database lying around, and you will need a matching build of your product to work with each one. If you don’t have a good plan in place before you start putting builds out there, you’re going to create a headache for yourself.

Let me illuminate this one a little further. Lets say you’ve built a billing system. You offer a hosted solution, where each client gets their own separate virtual server running a full-blown database and your web-based product and you manage it for them, and you also support client installations, where you will generate the database on their server and allow them to both host the web product, and run their own copy of the management system. So for n clients, you have n servers, n database instances, and n builds of your system.

Now you make some changes. You need to add a table or two. Change some keys. You now have version 2, which you proudly release, along with a tool to update the database from the previous version. Not everyone updates, but you think you will be fine to support version 1 as well.

After a few bug fixes to version one, and a few minor updates to version two, both databases look a little different than they did when you built the tool to update a version 1 database to version 2. Some clients aren’t on the original version 1 or the latest bugfix build, but somewhere in between. Most of your version 2 clients are fully patched, but some of them are holding out because your latest release broke some functionality they relied on. A few haven’t patched since 2.0. You’re spending too much money supporting version 1 in its various builds, version 2 has a few builds out there, migrating any given installation from 1 to 2 now costs much more than the update licence fee, marketing is screaming at you to release version 3, and you have folders full of SQL script files, a few spreadsheets to keep track of who’s applied which patches to which databases (mostly accurate), and your DBA just took 6 weeks medical leave (citing stress).

What went wrong?

You didn’t manage your database versioning. This seems to be a common problem, and (depending on your development platform) there may not be many tools out there to help you. Believe me when I tell you that a few folders full of SQL scripts and a couple of spreadsheets are not going to cut it.

What you should do about this is very much dependent on your development platform. If you’re using Hibernate and developing in Java, your solution will be very different to someone using Ruby on Rails. I do a lot of developing in C#.Net at the moment, and we have a fairly complex NHibernate-based data layer. After looking at a few of the different tools out there, I decided to roll my own. Whenever we make a change to our data layer, we write a script to update the previous version to the current one. Whenever any of our builds accesses a database, the data layer looks at a table in that database which is automatically kept up to date with the scripts which have been run. If the versions don’t match, a client tool or website will display a message, and an administration tool will provide the option of running any pending update scripts available to that build of the tool. With a single click, we can update any database we’ve produced since I introduced the system to any newer version we have a working build for. This is similar to the system RoR uses. There are other approaches that will work, but whatever you decide to do, a well-thought-out database versioning plan will save you both headaches and money.

A Build Server

I’m a huge fan of Continuous Integration. I like small, regular commits, automated builds and unit tests, and a clean build environment. I’m currently using CruiseControl.Net, but there’s a generalised version called CruiseControl, and other products which accomplish the same task. Basically, every time somebody commits some code, your CI server checks out the latest changes, builds your system, and (optionally) runs some or all of your test suite. I like to run a cut-down version of our unit tests on every build, and schedule a longer set of tests to run overnight. This gives developers rapid feedback about the success or failure of their commits, while still ensuring builds are tested exhaustively on a regular basis.

I don’t think a continuous integration server is nearly as critical to a team’s success as having a good source control solution. I’ll even accept that CI might be overkill in some circumstances. But for any decent-sized project, I really think it’s worth having one.

Issue Tracking

All too often, companies don’t have a good issue tracking system. By which I really mean: All too often, companies use Excel as their issue tracking system.

Now Excel is actually fairly good at keeping lists of things. It will even work reasonably well at issue tracking for very small projects. It doesn’t take long to out-grow it, though, and you shouldn’t cling to it once it stops working well. There are some really good systems out there, many of them with hosted solutions, some of them free. If you’re after free and open-source, Bugzilla comes immediately to mind. Wikipedia has an extensive list of issue tracking software. There are plenty of hosts out there which offer integrated issue tracking and source control, and it turns out you get some nice benefits out of combining the two into a single solution. Google Code offers free hosting to open source projects which combines source control, issue tracking, and a documentation wiki.

A Coding Standard

The problem with not having a coding standard is that some developers are lazy. Lazy can be good: a good lazy developer won’t write code three times when they could factor it out and write it once. Unfortunately, lazy can also be bad. An otherwise good developer who never writes any comments or documentation can cause all sorts of headaches for maintenance. A developer who isn’t lazy enough may write a dozen similar functions rather then spending a little thought to do the same task in a tenth as many lines.

You could end up with a religious war over the size of tab stops. I’ve seen this happen.

A good coding standard is a very useful tool. A long, wordy, and generally bloated coding standard, on the other hand, is mainly only useful when printed out and used as a door-stop. I won’t go into too many details here, but if you don’t have a standard at all, try to come up with a page or two, dealing with comments, member naming, namespaces, and the like. If it’s written down, you can hold people to it.

TEAM MEETINGS

I just can’t emphasize this enough. Developers often hate team meetings (“Can’t I just get on with writing code?”), but that’s usually because team meetings aren’t run well. You get three major benefits from having team meetings:

1.  You won’t end up with two developers doing the same thing.
   You only get this benefit, of course, if you talk about what you will do during the meeting. Meetings are not information-gathering sessions. They’re not show-off sessions. You’re here to plan what to do, and if you talk about doing things before you dive in, you can avoid duplicating work.
2. Developers will have a basic idea of the achievements of other team members.
   There’s no point doing lots of work getting a framework up to date if the people using the framework don’t know about it. I lied before. Meetings are show-off sessions. Just keep it short and to the point. We’re not here to listen to an hour-long lecture on the intricate object architecture your lead architect is so proud of.
3. You’ll pick up bad decisions before they hurt you, and maybe stumble across some good ones too.
   One developer might hear about something someone else is working on, and point out something you’ve already built which will help. Or just having a bunch of smart engineers in the room for ten minutes chilling out at the end of the meeting will be enough to come up with a great new idea.

I’m sure there are other benefits, too. As a team leader, you might pick up on some under-current running through the team that’s not obvious when you’re not all together. You might discover that one of your developers has heaps of experience in an area you were about to start a project in. The possibilities are endless.

Just keep them short and relevant, and always always ALWAYS head off private discussions early. Your whole team doesn’t need to listen while two developers hash out a network protocol because they don’t bother having meetings outside the general team meeting. This is the number one cause of engineers hating meetings. If you’ve got a whole-team discussion which seems to be going around in circles, appoint one or two people to investigate the issue later, and move on.

Finally, make sure someone is taking notes and distributing them. There’s no point in spending time making a decision if you’re all going to forget it. There’s no point in giving someone a task if there’s no follow-up. There’s no point in deciding what everyone will be working on for the rest of the week if everyone’s forgotten by mid-afternoon on the same day. Send out a summary email right away. Make it so short that even the laziest of developers will skim it.

Mentors

This obviously depends a little more on your team make-up, but I love having a mentor system in place where possible. Junior developers won’t develop if they don’t have someone to advise them. Don’t leave them to struggle with an architecture, trying and failing until they end up learning a bad solution: give them someone to go to for advice, so they can go straight to implementing a good solution.

I also love having periodic training seminars. Make your senior engineers run seminars. Make your junior engineers run seminars, and make your senior engineers show up to give them feedback. Provide pizza.

A Fast Deployment Process

If your product is quick and easy to deploy, people won’t screw it up. The highest-pressure deployments are often the most crucial: marketing has a huge client on the boil, they’re ready to sign a massive contract, and they were even nice enough to give you reasonable notice about the feature they simply must demo to make the sale, but the time-line is tight, you’ve barely finished the feature, and the meeting is about to start. Now is not the time to start on the intricate multi-multi-step deployment process with a dozen things that you can get wrong. Your deployment process should include backing up the previous version, a simple build installation (preferably from your clean, always-tested build server), and a rollback process for if your deployment breaks everything. It needs to deal with database versioning on the spot (and database backup too!). There’s a lot to be done there, potentially serious consequences if you mess it up, and the opportunity to look really slick to your clients when you roll updates in seamlessly, with virtually zero downtime and never a mistake made.

And more…

These ideas, and more, are the base-line things I take in to a new employer. Not every idea works for every employer, but they’re the low-hanging fruit for making life immediately better.

Once things like these are out of the way, you can get down to doing some serious innovating, knowing that the basics aren’t going to trip you up every step of the way.