Pandas Columns: Bracket Indexing (df[‘x’]) Versus Dot Syntax [df.x] | by Marcin Kozak | Mar, 2024

PANDAS FOR DATA SCIENCE

Does it matter how you do it? Maybe one is faster than the other?

When using Pandas, most data scientists would go for df['x'] or df["x"] — it doesn’t really matter which one you use as long as you stick to whichever you’ve chosen. You can read more about this here:

Hence, from now on, wherever I will write df["x"], this will equally refer to df['x']. Nevertheless, there’s another option. You can also go for df.x. While it’s a less frequent option, it can improve readability, assuming that the column’s name is a valid Python identifier.¹

Does it matter which syntax you choose? This article aims to address this issue, from two most important points of view: readability and performance.

The two approaches — df["x"] and df.x — are common methods for accessing the column (here, "x") from a data frame (here, df). In the data science realm, most likely the former is more frequently used — at least my experience from a variety of data science projects suggests this.

Readability and simplicity of use

Let’s consider the methods’ advantages and disadvantages in terms of readability and simplicity:

1. df["x"]: This is the explicit method. This option allows for using columns with names that have spaces or special characters, or more generally, that are invalid Python identifiers. Thanks to this syntax, you immediately know that "x” is the name of a column. Nevertheless, this is the less readable version for eyes: when you see plenty of such code, you may have to struggle with visual clutter in front of your eyes.
2. df.x: This method provides a more concise syntax, as every time you use df.x, you save three characters. You will appreciate this especially when concise code is preferred. Using df.x, it’s like…