This is an AMP version of the article, its original content can be found here.
Hits-of-Code Instead of SLoC
Lines-of-Code (aka SLoC) is a metric with a terrible reputation. Try to google it yourself and you'll find tons of articles bad-mouthing about its counter-effectiveness and destructiveness for a software development process. The main argument is that we can't measure the progress of programming by the number of lines of code written. Probably the most famous quote is attributed to Bill Gates:
Measuring programming progress by lines of code is like measuring aircraft building progress by weight
Basically, this means that certain parts of the aircraft will take much more effort at the same time being much lighter than others (like a central computer, for example). Instead of measuring the weight of the aircraft we should measure the effort put into it... somehow. So, here is the idea. How about we measure the amount of times programmers touch the lines. Instead of counting the number of lines we'll count how many times they were actually modified—we can get this information from Git (or any other SCM). The more you touch that part of the aircraft—the more effort you spent on it, right?
I called it Hits-of-Code (HoC) and created a small tool to help us calculate this number in just one line. It's a Ruby gem, install it and run:
The number 54687 is a total number of Hits-of-Code in your code base. The principle behind this number is primitive—every time a line of code is modified, created or deleted in a Git commit, the counter increments.
The main reason why this metric is better than LoC is that it is much better aligned with the actual effort invested into the code base. Here is why.
It Always Increments
The HoC metric always goes up. Today it can not be lower than it was yesterday—just like the effort, it always increments. Lines-of-Code is not acting like this. You may have a huge code base today, but after refactoring it will become much smaller. The number of lines of code is decreased. Does it mean you are less effective? Definitely not, but the LoC metric says so, to a non-programmer. A project manager, for example, may decide that since the size of the code base stayed the same over the last month, the team is not working.
HoC doesn't have this counter-intuitive effect. Instead, HoC grows together with your every commit. The more you work on the code base, the bigger the HoC. It doesn't matter how big or small the absolute size of the your product. What matters is how much effort you put into it. That's why HoC is very intuitive and may be used as a measurement of software development progress.
Look at this 18-month graph; it shows both metrics together. I used the same Java code base of rultor, a DevOps assistant. The code base experienced a major refactoring a few months ago, as you see on the graph. I think it is obvious which metric on this graph tells us more about the efforts being invested into the product.
It Is Objective
For HoC it doesn't matter how big the absolute size of the code base, but only how big your relative contribution to it.
Let's say, you have 300K lines of code and 95% of them were copy-pasted from some third-party libraries (by the way, it is a very common and terrible practice—to keep third-party code inside your own repository). The amount of lines of code will be big, but the actual custom code part will be relatively small. Thus, the LoC metric will be misleading—it will always show 300K with small increments or decrements around it. Everybody will have a feeling that the team is working with 300K lines code base.
On the other hand, HoC will always take into account the part of code that is actually being modified. The value of HoC will be objectively correlated with the actual effort of programmers working with the code base.
It Exposes Complexity of Lines
LoC is usually criticized for its neutrality towards code complexity. An auto-generated ORM class or a complex sorting algorithm may have the same size in terms of lines of code, but the first takes seconds to write, while the second may take weeks or months. That's why lines of code is usually considered a false metric.
Hits-of-Code takes complexity into account, because the longer you work with that sorting algorithm the more modifications you make to its lines. Well, this statement is true if you use Git regularly and commit your changes frequently—that is how you tell Git about your work progress.
Finally, look at this list of open projects completed by our team over the last few years. Every project has two metrics: Lines-of-Code and Hits-of-Code. It is interesting to see how relatively small projects have very big (over a million) HoC numbers. This immediately reminds me how much time we invested into it and how old they are.
I used the HoC metric in this analysis: How much do you pay per line of code?. That post compares a traditional project that paid $3.98 per HoC and an open source one, managed by Teamed.io, that paid ¢13.
My conclusion is that this Hits-of-Code metric can be used as a tool of progress tracking in a software development project. Moreover, it can be used for estimations of team size, project budget, development schedule and so forth. Obviously, LoC can't be the only metric, but in combination with others it may greatly help in estimating, planning and tracking.