Geek's Code Corner

Recently I found myself using R scripts from others and having to deploy my own R code on other machines. The majority of these had a hard coded path to either the script or the input data somewhere in the code. This becomes an issue when code needs to run on different machines, so here is a quick solution to this issue. While I prefer Python over R, I do need to use R from time to time as some packages have no equivalent in Python (yet). One issue with R I seem to be running into quite often is that depending on how you run a script, the working directory may or may not be set to that script鈥檚 directory. Especially those that like to open their .R files in RStudio without creating a proper project (like me) are likely to run into issues where the working directory isn鈥檛 necessarily the folder the script is in. A quick-and-dirty solution is to use setwd() to set the working directory in the beginning of the script to your local folder. However, if that script needs to be run by someone else on their machine, that path needs to be changed. The Read more…

Home Assistant is running great! It is time to add some more devices to my setup, and since all groundwork to add Zigbee devices using Zigbee2MQTT was done previously (check out this post) it makes sense to add more devices that support this protocol. For this post I picked up two Aqara sensors which can measure temperature, humidity and air pressure in a small form factor. Pairing new Zigbee devices First we need to pair Zigbee2MQTT to the new sensors, to do this we’ll have to log onto the Raspberry Pi and edit the file /opt/zigbee2mqtt/data/configuration.yaml. Find the line that says permit_join: false and change this to permit_join: true. Then restart Zigbee2MQTT using the commands below. sudo nano /opt/zigbee2mqtt/data/configuration.yaml sudo systemctl restart zigbee2mqtt Now Zigbee2MQTT is ready to pair with new devices. To pair a sensor, press the button on top. Once you press the button it will blink once indicating it is pairing and upon connecting it will blink twice. Now go back into /opt/zigbee2mqtt/data/configuration.yaml and give your device a name, this is done by adding a friendly_name tag to the device as shown below in my configuration file. … devices: ‘0x00158d0003495e50’: friendly_name: Aqara Cube ‘0x00158d00068b6c2e’: friendly_name: TMP Sensor Read more…

Are you tired of fumbling with your Nintendo 3DS charging cable? Do you wish you could use a modern, reversible USB Type-C connector instead? With a few common tools and a little know-how, you can add a USB Type-C port to your Nintendo 3DS charging cradle in no time! In this blog post, we’ll show you how to modify your charging cradle to accept a USB Type-C connector, so you can charge your device with ease. We’ll cover everything from the materials you’ll need to the step-by-step process of drilling and soldering. This process is similar to adding a USB-C port to a Raspberry Pi Pico, which we covered in a previous post. The Nintendo 3DS charging cradle is a good candidate for this modification as there is plenty of room on the inside to add a USB-C breakout board and wire it up to the charging PCB. What you need Before we begin, let’s make sure you have all the tools and components required for this soldering project. Here’s what you should have at your workspace: A 3DS charging cradle Electrical wire – You’ll need two pieces (preferably black and red), which can be repurposed from e.g. an old Read more…

With Google’s Cloud Functions you can make simple microservices that can perform a task on Google’s infrastructure. This allows some functionality of an app to be moved to the cloud. This can be useful in a number of cases; for instance the Dashboard from a previous post. When running scripts on a platform like the Kindle Paperwhite, you are somewhat limited to which tools and packages you can use easily. For instance installing additional python packages would be a non-trivial step, as is installing other software. Therefore, the script makes use of vanilla regular expressions to parse html instead of a more specialized package like BeautifulSoup. However, when more complex processing is required, re-implementing everything would be quite cumbersome. So here a cloud function could be a great solution. Each function can be called through an HTTP request, similarly as most web APIs. The script will then be run on Google’s hardware and the results are passed back in a response. As an example I’ll set up a Cloud Function that will grab citation data from a Google Scholar profile, as Scholar has no API, the only way to get citation statistics is to parse them out of the HTML Read more…

With the Agent Based Model (ABM) from the previous post we can simulate any number of Gwent players, with known skill and play rate, playing throughout a season. Using this model we’ll try to get a firmer grasp on the ladder system and test whether this rewards skill or grinding games… Climbing Pro Ladder Skill vs Grind The ABM mimics a population of agents playing Gwent for an entire season with different innate skill and time to play. Though unlike with actual people we can change the parameters of the model to examine what we want to test. So in the first model we’ll disable any kind of learning, where agents would get better as they play more games, to see only the impact of playing more games on the peak MMR score (of a single faction). To do this a model was run with these parameters: 8000 players 100 steps resulting in players playing 25-170 games per season no learning players don’t get better by playing more games initial skill ranges from 1200-2700 distributed similarly as chess players After running the model with these parameters, we can group players by skill (in bins of 200 ELO) and amount of Read more…

When three or more groups of samples are compared (e.g. Using ANOVA/Tukey HSD or Kruskal-Wallis/Dunn), you often see results shown as a boxplot, with lines highlighting which groups are significantly different. In Python there is no single package to do this quickly, though by combining scikit-posthocs with statannotations similar plots can be generated with relative ease. Here wego over the code involved step by step. Statistical tests comparing three or more groups are typically done in two steps. The first test will check if there are any statistical difference between the groups, the second test will then tell you which groups are different. The second test is referred to as a post hoc test. While there are many combinations of tests to choose from, common combinations are Kruskal-Wallis followed by a post hoc Dunn test (non-parametric) and ANOVA with Tukey Honest Significant Differences (parametric). Though note that the code below can easily be adapted for other tests. All code from this post can be found on GitHub and Binder. The data The iris dataset, with measurements of petal and sepals of three species of flowers will be a nice way to test some of these test. The code below will Read more…