Saving plots to memory rather than disk for python web app

Question

I'm developing a FastAPI application which uses sci-kit learn to generate some SVG files which are saved locally before being uploaded to AWS-S3 for permanent storage. However, once deployed on Heroku I realised it doesn't allow for writing to local storage.

An example of how these files are generated:

from sklearn.tree import DecisionTreeClassifier, plot_tree
import matplotlib.pyplot as plt
fig = plt.figure()
    decision_tree = plot_tree(
            pruned_clf_dt, # a decision tree made by sklearn
            filled=True,
            rounded=True,
            class_names= classNames,
            feature_names=X.columns)
    fig.savefig("example.svg", bbox_inches='tight')

Is it possible to do fig.savefig (into a variable) to save the SVG in memory or somehow save the plotted tree as an SVG into AWS-S3?

Answer 1

Answer is yes it's possible by using StringIO, though S3 requires the object to be in a string(?) like format eg:

import io
from sklearn.tree import DecisionTreeClassifier, plot_tree
import matplotlib.pyplot as plt
fig = plt.figure()
decision_tree = plot_tree(
        pruned_clf_dt,
        filled=True,
        rounded=True,
        class_names= classNames,
        feature_names=X.columns)
s = io.StringIO()
fig.savefig(s, format = 'svg', bbox_inches='tight')

svg = s.getvalue()

name = "filename.svg"

s3bucket.put_object(
        Key=name,
        Body=svg,
    )