The first part of the article can be found here.
Highload is a separate matter. In practice, you have to sacrifice a lot of things including good architecture and understandable code for high productivity. Sometimes you may want to cry when you look at what your cozy project turned to after optimization. But what can you do? Instead, program execution time reduced twice. Sometimes you don’t have a choice.
Projects using third-party libraries
When I talk about third-party libraries some my associates start looking at me understandingly. I see in their faces – they suffered in the same way. In real conditions of the big project you will hardly develop absolutely all functionality. There are usually quite common tasks that have been already solved by someone. In this situation it’s reasonable to use a ready-made solution than to reinvent the wheel. This is it, but sometimes it appears that authors of this ready solution are not good programmers. Their project solves its main task, but it’s developed…not really professionally. And integrating it into your project is… not really easy. This circumstance makes you add terrible kludges spoiling your fine architecture. But this is an operational need since it’s not cost-effective to implement this functionality on your own. (I will tell you in secret, several times I couldn’t bear it and developed my own library instead of using the third-party one. But this is an exception, not a rule.)
Project of newbie developers
You can often hear that an average developer should know this and that. It’s assumed that a developer knows some programming language (if we talk about OOP-language, he should at least know polymorphism and inheritance), he can understand complex syntax constructions, basic platform mechanisms, he knows common patterns (when he sees a class named Visitor, he will immediately understand the definite system part), he easily reads comments in English and can do much more things. But before becoming an average developer everyone was a newbie developer. And nowadays there are a lot of people who begin to study this science. And this is ok that they don’t know and can’t do a lot of things. This is a usual situation when some novice developers form a team and start developing a project. They learn many new tricks while developing the project. Of course they will do a lot of things in improper way. And you shouldn’t expect the opposite. It would be good if senior colleagues help with different issues: how to re-write code in a better way, what books to read, etc. But you shouldn’t specify the same requirements as to senior staff. Junior developers’ project may have lots of deviation from “how it should be” as they are just studying. Advice and recommendations are good while the requirement to write an ideal architecture from the first time is not that good.
Sometimes I have to create demo projects to show some great features to my associates. It can be language, engine, library or something else they haven’t seen before. As a rule, the project should be detailed and simple with lots of comments. In some cases there are tens of lines of comments for a single line of code and this is ok. You don’t write the perfect code, you just use the code as illustration. At the same time the code can be very bad and execute for too long. It doesn’t matter as we have a different purpose to demonstrate some technology to the public.
In this project, we don’t show anything to anyone. We want to understand some new interesting feature by ourselves. Let’s say we examine some algorithm. It’s ok to write 10 versions of the algorithm side by side. Probably, versions will be in different languages. It will be ok not to care about the agreement on names (that is good to follow when creating a real project) and to name the same things in the same way – by names set in the book. It doesn’t matter what’s the common practice in this or that language. We examine the algorithm; we don’t care about such things right now. If you are going to show the result to someone else you should work over the code, but this is another story. And while you are on the research stage your main purpose is the research, not the perfect code. Of course, you can combine these things, but this is not obligatory. The main thing is not to substitute the academic purpose with the perfect code purpose.
Many developers can say that it is necessary to get used to write good code everywhere. But local code has some specific, you don’t need to show it to anyone, you don’t need to report to anyone, you can be guided by your own ideas in development. There can appear a lot of bad intermediate code in the workflow. You can play with the platform, make some experiments. You can develop in the way convenient for you. You can create some data dump regardless of the general architecture and just save it to a local file from the most unacceptable place. It’s possible to write any amount of comments in any language, if this is more convenient for you.
But remember that the situation changes when the creative process ends and you need to show the results to others (send local developments to the central repository). You need to clear the code in this case. All the experiments, kludges and excessive comments should be removed. Respect developers who will have to understand your writings.
The main purpose of such projects is to quickly create some features to get the idea how they will look like. This is a reasonable approach. Let’s say we have 5 different implementations: we will chuck together a common concept of each variant. After that you will be able to examine all approaches on live samples and choose the one that will be used in the main project. It’s important to understand prototyping tasks. You don’t need to clear this code, you don’t need to write it ideally. I am always impressed by people who criticize a prototype with the following statements: “this variable could be named in a more understandable manner” or “it would be good to move this interface button 1 pixel to the left”. Does it matter how the variable is named? This is a prototype, back off. Such discussion is reasonable for a finished project, but it has no sense for a prototype.
I remember that once I and my friends decided to make a birthday present to a good man. We created a Java project which represented in OOP his life, friends and other interesting things he interacts with. The program really worked, it was possible to execute some funny commands from the console. As to the source code, absolutely all names (classes, methods, variables, etc.) were written in Russian (Java allows such tricks). Javadoc was also written in Russian and didn’t contain any useful information. Logic was implemented in the simplest way. We used the simplest algorithms instead of the complicated ones. Architecture wasn’t too pretty; we even didn’t try to think over it.
And the present worked out, though we haven’t used any good practice of writing the perfect code. And the matter is in absolutely different purposes of the project.
Once again I would like to draw your attention to main ideas. If you want to become a good programmer, you should develop, learn to write better code, work to improve yourself. You should try to write as clear and good code as possible. But you need to understand that your code will never be ideal in a big project. You shouldn’t admit bad code; write it in the way that you won’t be ashamed of. At the same time don’t forget about the project purpose since in most cases the perfect code is not the purpose, but means to reach it. It’s not necessary to intentionally write bad code, but show no fanaticism about the perfect code as well. Remember about your purposes and situation. It’s not that easy to write good code. Compare your effort spent for code improvement and the result it will get. If you read a good article about the good code, don’t extract separate pieces of advice that you will use everywhere without a second thought. Pay attention to the context of the given advice. Think of the situation described in the article. Think when it’s appropriate to use known patterns and when it’s not. And in general, think more, it’s very useful in programming. And everything will be ok for you.
P.S. You can find description of other project types and how perfect code deals with it in the part 1 of the article.
Perfect code and real projects: part 2Andrey Akinshin
The first part of the article can be found here.