Running Python Code in Jupyter Notebook

A common first assumption is that any call to a plotting function will immediately put a picture on the screen. In many environments that assumption is false. Whether a figure appears depends on where the code runs and on how that environment has been configured. This article follows a single plot from the moment your code creates it to the moment it either appears or stays hidden, so you can predict the behavior in each situation.

The wrong belief comes from confusing two separate jobs. One job is building the figure from data, which the plotting functions handle. The other job is presenting that figure to a viewer. Keeping these jobs apart is the key to understanding every behavior in this article.

matplotlib is the primary scientific plotting library in Python. It provides functions for making publication-quality visualizations such as line charts, histograms, and scatter plots. Visualizing data is a practical way to notice patterns and gain insight into an analysis, which is why a notebook-based workflow relies on it so heavily.

The two plotting calls above build a figure from the data. Nothing in those lines, by itself, says where the figure should be shown. Inside a Jupyter Notebook, that decision is controlled by a magic command, which is a special instruction recognized by the notebook environment that changes how figures are handled. Without that instruction, or outside a notebook altogether, the figure is not automatically placed on screen.

The figure can fail to appear at the decision point. Plotting succeeded and the figure exists, but the presentation step never ran. That is why the same plotting code can show a picture in one environment and show nothing in another.

Inside the Jupyter Notebook, you can show figures directly in the browser by using one of two commands: %matplotlib notebook or %matplotlib inline. Both are placed in a cell and run before the plotting code. The two commands are not interchangeable in behavior, so the choice matters.

The practical difference is what the viewer can do with the figure. Compare the two commands side by side before choosing one for a task.

The magic line runs first, then the plotting code runs after it. Changing the first line to %matplotlib inline keeps the same plotting code but produces a static figure instead.

A display magic is not the only way to show a figure. The source notes that the setup it assumes runs in a Jupyter Notebook with a display magic enabled. If you are not using the notebook, or these magic commands, you have to call plt.show to display any of the figures. The same plotting code therefore needs different handling depending on the environment.

The second misconception concerns the language version rather than the notebook. There are two major versions of Python in wide use: Python 2, more precisely 2.7, and Python 3. Python 2 is no longer actively developed. Because Python 3 contains major changes, Python 2 code usually does not run on Python 3. The belief to avoid is that code written for one version will simply run on the other.

The first line is a Python 2 style print statement. Running it under Python 3 raises a syntax error, because print is a function in Python 3 and needs parentheses. This is one example of the kind of change that stops older code from running. For new projects, the source recommends using the latest Python 3.