Python Engineer Beginner Mistakes

I am writing Python code since 5 years, and I have worked with around 5-6 entry-level Python programmers. There is a pattern I can see in their programming practices. My hope is by reading this article there will be lesser such mistakes.

I have worked on OpenStack for over three years, which is written in Python and has a very high code quality, so I feel I am qualified enough to write this post.

I’ll start with talking about these beginner-programmers first. Some of the people I have worked with are self-taught programmers but majority of them are junior developers working in startups in Bangalore. For self-taught Python programmers, such mistakes are understandable – Python is a pretty easy language to pick up and code, so they don’t really have a motivation to go with best practices right from the start. Even if they googled ‘Python best programming practices’, they will find a really big list of practices without any indication about which ones are the most important and absolutely must be followed right from the start. For Python beginners working in startups, I am guessing there are not enough senior Python programmers who can teach them the required best practices.

In this article I’m going to list the bare-minimum of these simple practices you should follow. Surprisingly, most of the ‘Beginner Python mistakes to avoid’ articles on the web don’t talk about the bad practices I encounter. I think they just assume such simple things are taken care of by a programmer. My experience tells otherwise.

Indentation, tabs and spaces.

Never use tabs. Only use spaces. Use 4 spaces for a tab, always.

I frequently see codes which use a mix of spaces and tabs, and mix of 2-space and 4-space indentations all in one file! If you are coming from a different language, or if you already have a notion of, say, 8 spaces for a tab, then it might be difficult and seem unnecessary at first. But I am guessing you are liking the language and you look forward to writing lots and lots of Python code in your lifespan; in such a case, just bear with me for a few weeks and it’ll come naturally to you, trust me :)

Raise specific exceptions

Don’t write like this:

try:
    # do something here
except Exception as e:
    # handle exception

Catching the parent Exception is bad. If you know what exception the code is going to throw, just except that specific exception. The most common example I have seen is when calling get() method on a Django’s model. If the object is not present, Django throws ObjectDoesNotExists exception. Catch this exception, not Exception.

Most often, you do some corrective action in the ‘handling exception’ part. If you are excepting (if there’s a word ‘excepting’) Exception, then you are heading straight into this ‘handling exception’ section no matter what exception occurred. If your database is down, the ‘DatabaseIsDown’ (I made up that name) exception will also be caught! This might make debugging extremely hard when your app grows big or you start getting a lot of users.

Trailing whitespaces

In context of Python, whitespaces means space characters. There should be no line in Python code which has spaces at it’s end.

You might think why make such a big issue of this. The problem is version/source control. You invariably will use a source control system at some point of time, for example ‘git’. When you commit trailing spaces, it becomes difficult to read the diffs (the difference between two commits/checkins). Also, removing or adding more spaces might feel okay if you develop a habit of leaving around trailing spaces. But when you will start using a source control system, this will add unnecessary noise in commits, making code review and debugging difficult. Remember, code is read much times than is written or modified.

Edge cases first. Shallow nesting.

Consider two pseudo codes. These are codes for handling a request, but you can altogether ignore that fact.

Pseudo code 1:

if request is post:
    if parameter 'param1' is specified:
        if the user can be authenticated:
            if the user can perform this action:
                # Execute core business logic here
            else:
                return saying 'user cannot perform this action'
    else:
        return saying 'parameter param1 is mandatory'
else:
    return saying 'any other method apart from POST method is disallowed'

Pseudo code 2:

if request is not post:
    return saying 'any other method apart from POST method is disallowed'

if parameter 'param1' is specified:
    return saying 'parameter param1 is mandatory'

if the user can be authenticated:
    return saying 'user cannot be authenticated'

if the user can perform this action:
    return saying 'user cannot perform this action'

# Execute core business logic here

As you can see, the second code is much easier to read. You will also notice that the second code doesn’t have a lot of nesting, which avoids common problems. Can you see that I have not written one ‘else’ section? If the core logic is several dozens of lines long, it becomes difficult to find out which else belongs to which if if we write in ‘Pseudo code 1’ style.

I am sure while starting to write, pseudo code 1 should feel normal. But with very minimal effort you can very soon get used to writing code in ‘pseudo code 2’ style.

The suggestions mentioned below are minor, probably opinionated ones. These might apply to not only Python but any programming language. I feel that these points are important enough that you should start a habit of adhering to them right from the start. It’ll help you in the long run.

Make commit messages lengthy

This is a fairly general trend I see. The commit messages such people generally write are just a few words long. Most of the beginners have started using git as version control system. In git, such people have gotten used to using git commit -m "your commit message here" command. If you use this command, I am sure you won’t even feel like writing a commit message more than what fits in one line :). Do you know that you can just type git commit and it will open up an editor (vim or nano) to write the commit message? Start using this and stop using -m flag directly on command line. Your colleagues and your future self will be thankful to you :).

I generally prefer writing a short description in the first line of commit message, then leave a blank line, and then write a paragraph after that to explain in more detail if required. The first line is the most important. Try to describe as much as possible concisely in this single line. But don’t ever let this line go above 100 characters, preferably 80. See this commit message as an example:

Don't allow access to POST /v1/purchases API without authentication

It was revealed during testing that the above mentioned API is accessible
even when accessed without authentication. This is a security risk. This
commit fixes it. One 'if' condition added to reject API call if the user is
not authenticated.

You might feel that it’s a lot of effort to write such log git commit messages. But you can realise it’ll not take more than 2 minutes to write this. Much less than how much time you spent writing code corresponding to this commit :)

Again, remember that commit messages are important part of documentation, and are extremely important for debugging later.

Do only one thing in a commit

If you are fixing three issues, make sure you create three different commits for each one, and never one big commit to include three bug fixes. Even if each of these bugfixes are just a couple of lines. If you are doing a feature work, then it’s okay to put all the feature work in one single commit, but for a bugfixes, create a separate commit for each one. As you can see, you should get comfortable with the idea that a commit can be as short as one line and as large as couple of hundred of lines. Let go of the feeling which your brain has about number of lines in a commit, that it feels right only when there are about 50-100 lines in a commit. :)

If you are using git, take use of git add -p to add specific chunks of code in git, instead of adding a complete file into git like you’re used to, by doing git add file1.py. It’s a pretty powerful tool. Spend 15 mins to understand it and you’ll go ‘why didn’t I know about this till now!’

I am realizing I can actually write ‘Git beginner mistakes’ as a separate blog :). But anyway, continuing..

Spaces after comma and colon

Just make this a habit. It improves your code readability.

Example 1 - bad:

def create_user(name=name,height=height,weight=weight):
    pass

Example 2 - good:

def create_user(name=name, height=height, weight=weight):
    pass

Example 3 - bad:

request_dict={'name':name,'height':height}

Example 4 - good:

request_dict = {'name': name, 'height': height}

Write a lot of comments and docstrings

If something is not abvious from the code, you should write a comment about it. Also, make a habit of writing docstrings to functions/methods, even if it is just a single line. Example:

def create_user(name, height=None, weight=None):
    '''Create a user entry in database. Returns database object created for user.'''
    # Logic to create entry into db and return db object

Write TODOs literally everywhere

While writing code, you realize it can be improved, but you don’t want to improve it right now as it is not that important thing at this moment. What should you do? Just create a TODO! I find this a very good compromise for smaller tasks, instead of creating an issue in an issue/bug tracker. TODOs can be written for improving on docstrings (TODO(rushiagr): add more description here), a more optimised version of code (e.g. TODO(rushiagr): can be optimised by using list comprehensions instead of the ‘for’ loop), or a refactoring (TODO(rushiagr): create a common method of this thrice-duplicated code). The parenteses specify who wrote the TODO, which can be helpful if somebody wants to fix the TODO but doesn’t know who to contact for more information. But you can leave this part out too.

I use TODOs so much that I created a Vim abbreviation for it, so that as soon as I type #t followed by a space, it auto-completes it to # TODO(rushiagr): . If you’re using Vim, you can do by adding this line to your .vimrc:

:abbreviate #t # TODO(rushiagr):

Parting words

My architect at Reliance used to say this: ‘You should not be able to tell by looking at code who wrote that code’. What he meant was, one should not be able to introduce his/her specific style of writing code into code base. Code should be written using a convention generally accepted by the whole of the team. It might not completely include everything of the best-practices of that language, but your team should decide what programming practices can’t be compromised.

That’s all folks! :) Do let me know what you feel about this blog by commenting. Do you know something which you too encountered a lot which is not present in this article? Or you disagree with something I wrote?

w