'Lil Dev

How to set up a basic coding environment on the Raspberry Pi 3B.

Step 1 - Booting Image:

download minimal Raspbian Buster image from here: https://www.raspberrypi.org/downloads/raspbian/

Raspbian Buster is the RPI equivalent of Debian Buster. It comes with Python 3.7.5 built-in, but more importantly, the Python packages that we'll download are super easy to install via pip from piwheels.

Why not Ubuntu? Or Anaconda to manage packages/dependencies?

While Ubuntu has some good, stable images for the Raspberry Pi, the aptitude repos for many Python packages are fairly old. If you don't feel the need to stay on top of the most current updates, then Ubuntu will be totally adequate.

The ports of Anaconda for the RPI, on the other hand, I find pretty disappointing. Maintenance seems pretty infrequent and the available packages are limited. The piwheels project seems much more promising for updates and attaining the general catalog of packages normally found on PyPi.

Step 1.5:

If running RPI headless:

add ssh file to boot partition by cd-ing into your SD-card directory and executing

touch ssh

This just needs to be blank file, no format necessary. More info here

Step 2 - Initial Raspbian Setup:

If running RPI headless:

ssh with pi@ipaddress

password is: raspberry

change your password with passwd command

update and upgrade: sudo apt update && sudo apt full-upgrade

Step 3 - Install OS Dependencies

install pip: sudo apt install python3-pip

numpy and general: sudo apt install sqlite3 libatlas3-base libgfortran5 git libpq5

spatial: sudo apt install libspatialindex-dev xsel xclip libxml2-dev libxslt-dev libgdal-dev

for Pillow: sudo apt install libwebpdemux2 libzstd1 libopenjp2-7 libjbig0 libtiff5 liblcms2-2 libwebp6 libwebpmux3 libjpeg-dev zlib1g-dev libzmq-dev

reactivate bash-profile: source .profile

Step 4 - Pip Install Python Packages

Python Packages:

General/SciPy/ML/Viz:

Jupyter/iPython
Pandas
NumPy
SciPy
sklearn
xgboost
matplotlib
seaborn

pip3 install Cython jupyter numpy scipy matplotlib pandas scikit-learn scikit-image seaborn

Web-Scraping:

scrapy
selenium
bs4

pip3 install requests selenium beautifulsoup4 Scrapy

NLP-related:

nltk
gensim
textblob

pip3 install nltk gensim textblob

Web-server/Database:

flask
pymongo
pscypg2

pip3 install Flask pymongo psycopg2

Geopandas 😩

This was mega annoying because the apt version of 'libproj-dev' is at 5.2.0 for Buster rather than the 6.0.0+ that 'pyproj' - a geopandas dependency - currently needs. Building PROJ from source proved too painful and buggy, so I just elected to use an older version of 'pyproj'.

pip3 install shapely fiona pyproj==1.9.6 geopandas

reactivate bash-profile: source .profile

Step 5 - Install Database Tools

Database Tools:

PostgreSQL

sudo apt install -y postgresql libpq-dev postgresql-client postgresql-client-common

Extra setup

MongoDB

sudo apt install -y mongodb

debian repo versions are very old @v2.4.14

The main complication with Mongo setups on Raspberry Pis has to do with the newer versions of Mongo needing 64-bit operating systems, while Raspbian images are currently all 32-bit operating systems. Memory issues also play some role in this.

I didn't think it worth the while, but this might allow you to get up to v3.2: http://koenaerts.ca/compile-and-install-mongodb-on-raspberry-pi/.

Redis

sudo apt install -y redis

MySQL (MariaDB)

TODO: https://pimylifeup.com/raspberry-pi-mysql/

Extras:

More ML tools:

These Deep Learning tools are probably overkill for the Pi, but I've included them just in case.

Tensorflow:

pip3 install tensorflow==1.13.1

This version is slightly old, but is confirmed to build properly. The official instructions on the Tensorflow website using Docker to build the wheel failed when it got to SciPy...

Keras:

sudo apt-get install libatlas-base-dev
sudo apt-get install gfortran
sudo apt-get install python3-h5py
sudo pip3 install keras

PyTorch:

dependencies:

sudo apt install libopenblas-dev cmake cython python3-yaml

# gdown allows you to download files from Google Drive
pip3 install gdown
gdown --id 1D3A5YSWiY-EnRWzWbzSqvj4YdY90wuXq --output torch-1.0.0a0+8322165-cp37-cp37m-linux_armv7l.whl
pip3 install torch-1.0.0a0+8322165-cp37-cp37m-linux_armv7l.whl

NOTE: this is the build for armv7 (RPI 2 & 3)

.whl file on Google Drive

for 3B+ and later: https://devtalk.nvidia.com/default/topic/1049071/jetson-nano/pytorch-for-jetson-nano-with-new-torch2trt-converter/

Extras

Docker:

curl -fsSL get.docker.com -o get-docker.sh && sh get-docker.sh

Java/Scala:

sudo apt-get install openjdk-8-jdk
wget https://downloads.lightbend.com/scala/2.12.10/scala-2.12.10.deb
sudo dpkg -i scala-2.12.10.deb

sbt:

sudo apt-get install apt-transport-https
echo "deb https://dl.bintray.com/sbt/debian /" | sudo tee -a /etc/apt/sources.list.d/sbt.list
sudo apt-key adv --keyserver hkp://keyserver.ubuntu.com:80 --recv 2EE0EA64E40A89B84B2DF73499E82A75642AC823
sudo apt-get update && sudo apt-get install sbt

limit memory usage: sudo nano /usr/bin/sbt

add SBT_OPTS=-Xmx256M to beginning

https://aknay.github.io/2017/05/09/how-to-install-scala-and-sbt-in-raspberry-pi-3.html

Node.js; npm:

curl -sL https://deb.nodesource.com/setup_10.x | sudo bash -
sudo apt install nodejs

Golang

TODO https://www.e-tinkers.com/2019/06/better-way-to-install-golang-go-on-raspberry-pi/

Rust

TODO https://www.makeuseof.com/tag/getting-started-rust-raspberry-pi/

chmod +x lil_dev/raspbian/setup.sh //add execute permission sudo lil_dev/raspbian/setup.sh -b

Basic Install Size = Full Install Size = 6.2G

OpenCV

https://linuxize.com/post/how-to-install-opencv-on-raspberry-pi/

otosky/lil_dev