Adding rules to code

A typical quality check before running any models is to see if there is any data. This can happen when you are fitting a separate model for differnet user segments and some user segments may be non-existent.

In these cases, we often want the code to behave differently. To achieve this, we can use if/else statements to check if a condition is met (i.e. True) then run different pieces of code accordingly.

Basic if/else

If/else statements generally take the form of:

if CONDITION:
    CODE_TO_RUN_IF_CONDITION_IS_TRUE
else:
    CODE_TO_RUN_IF_CONDITION_IS_FALSE

Again if and else are special keywords. CONDITION needs to be something that can be cast into a boolean. This is often:

• a boolean operation, e.g. len(a) >= 10. So the output itself is already a boolean
• or a piece of data, e.g. [1, 2, 3] (to check if there is any data)

The indentation for the body of the code is important in Python. Other programming languages use {} to indicate the body of the code. This quickly fills up the script and makes it hard to read.

The simplest if/else statement only needs an if:

x = ''
if x:
    print('x is not empty')
    print(x)

The code above should print nothing. The if x: is the declaration of the if statement. Given x is not a boolean value, Python will try to cast it to a boolean, i.e. bool(x) to decide whether print(x) should run or not.

A slightly more complicated version explains how else works.

x = ''
if x:
    print('x is not empty')
    print(x)
else:
    print('x seems to be empty')

elif (else if)

It is possible to have sequential conditionals:

x = -2
if x < 0:
    print('x is bad')
elif x >= -3 and x < 5:
    print('x is okay')
else:
    print('x is good')

The above if/else is intentionally bad to demonstrate a few things:

• Python does not check if your logic makes sense, in our example with x = -2, it is unclear which line will be printed.
• Notice that only the first print() statement is ran. This is because elif and else will not be checked if the condition before is true. Therefore the ordering of your conditions matters.
• else will capture all possible leftover cases. This can lead to surprises.

Defaults in if/else

Sometimes we initialize variables (like the empty list before a loop) differently according to another variable or condition, it is then common to do

y = True
if y:
    x = 1
else:
    x = 2

Notice that x = 2 will not run at all.

However, sometimes to simplify code, it’s better to write

x = 2
if y:
    x = 1

This way it is clear that x is defaulted to 2 and will be overwritten if the condition y is True. This also avoids cases where one may forget to define x in the else clause.

An even shorter version is:

x = 2 if y else 1

If/else in list comprehension

It is common to run a loop but only process certain records that satisfy a condition.

Here is a very verbose example:

nums = [1, -3, -4, 5]

output = []
for num in nums:
    if num > 0:
        output.append(num * 10)

List comprehension can save a lot of space. All we need to do is place the condition after the loop

nums = [1, -3, -4, 5]

output = [num * 10 for num in nums if num > 0]

Notice that all the False cases will be dropped!

For readability, some people like to separate the conditional in the list comprehension

output = [num * 10
          for num in nums
          if num > 0]

Control flow in a loop

Continuing the same verbose example from above. Sometimes the code after the if statement has many lines. In that case, having a lot of code indented due to a single condition seems unnecessary. In this case we can use continue or break to rewrite the code.

nums = [1, -3, -4, 5]

output = []
for num in nums:
    if num <= 0:
        continue
    output.append(num * 10)
print(output)

continue will terminate the current iteration and move on the next iteration of the loop. break on the other hand will terminate the entire loop.

nums = [1, -3, -4, 5]

output = []
for num in nums:
    if num <= 0:
        break
    output.append(num * 10)
print(output)

Catching errors with try and except

Sometimes we know the code will break, i.e. the code throws an error message and it is not easy to capture the logic in an if/else statement. In these cases, we will try to “catch the error” before it terminates the remaining code.

Below we intentionally trigger a bug when we add a str to an int.

answer = 1

try:
    print('the answer is ' + answer)
except:
    print('something went wrong')

After running the example above, you should see the behavior when answer = "1", i.e. when no bug is triggered.

Overall, try will attempt the content in its body (i.e. the indented code), if an error occurs, it will try the code in the except section.

Best practice - catching specific errors

Catching errors seem like a good feature but sometimes the errors can go unnoticed for quite some time. To ensure you’re only capturing the type of errors you anticipated, you can catch specific errors by specifying the type after except.

answer = 1

try:
    print('the answer is ' + answer)
except TypeError:
    print('something about types went wrong')

Any TypeError triggered will only result in a print() call but any other type of error will terminate the program!

Another common practice, is to capture the error message by rewriting

answer = 1

try:
    print('the answer is ' + answer)
except TypeError as e:
    print('something about types went wrong')
    print(e)

Building error messages into my code?

It is good practice to anticipate bad input for your code so you can provide meaningful error messages to your users. For those who want to raise errors when bad inputs are passed. There’s a keyword raise for that.

input = 1

if type(input) != str:
    raise TypeError('expecting str input bur got non-str values')

So in this case we initiated a Type Error if the input is not of type str. You can see a list of built-in errors

Checking the type of variables is a common enough operation that a built-in function exists. We could rewrite the above with the function isinstance()

input = 1

if not isinstance(input, str):
    raise TypeError('expecting str input bur got non-str values')