New project: from private to public

Earlier in the year I wrote about using systemd in the context of setting up a service for temperature logging and visualisation on a Raspberry Pi. Whilst the project has been stable for a long time, I’m now releasing it publically. Here it is. Key features include:

• Logs temperature using a Raspberry Pi and DS18B 1-wire device
• Queries the Dark Sky weather API
• Writes temperature readings and weather data to a local database
• Streams temperature readings and weather data to Adafruit feeds, which can be used to build a dashboard

Hopefully the instructions are easy to use and the various features can be turned off as needed

From private to public

In many ways keeping a private git repository can foster bad habits. My delay to make it public was mainly due to hard coding API keys: fine for local use but a liability in code meant for sharing. In the end I worked out how to separate my private keys from the code but by the time that happened they were in the git history for all the world to see.

Scrubbing git history

It is possible to selectively scrub git history. This Stack Overflow post details how to purge a file from git history; the github faq does the same. However, it gets a little messy when the file in question is the main file containing code within the repository.

In the end the benefit of retaining a git history without my main file didn’t outweigh the cost of going through the steps and risking either corrupting the repo or revealing the keys. I just made a separate public repo.

Hiding keys

In the same spirit, I opted for a simple approach to hide the keys. I wrote a file called api.txt and populated each line with a key:value pair. I then added the filename to .gitignore to ensure I didn’t add it to a commit, thereby nullifying the whole point of the exercise. I then used the code below to load it into a dictionary object within python.

# Test for reading api key text file

from os import listdir
from os.path import isfile, join
print(listdir())

api = {}
with open('api.txt', 'r') as file:
    for line in file:
        (key, value) = line.split()
        api[key] = value
file.close()

A better way?

Both R and Python understand environment variables. For R we can use the .Renviron file, populating it with key:value pairs such as MY_KEY=12345. We can recall this variable in code using Sys.getenv("MY_KEY").

For Python we can do a similar thing in a few different ways. Anaconda has its own method; there’s also decouple:

from decouple import config

API_KEY  = config('MY_KEY')

The only difference with the R method is that the file that stores the keys should be called .env.

Another method would be to use the os package to set and get variables. Obviously the setting of variables would happen in a script not added to the repository.

import os

# Set environment variables
os.environ['MY_KEY'] = '12345'

# Get environment variables
KEY = os.getenv('MY_KEY')