Offer: Install-Script for Linux

[fkellner]

Dear rumble developers,

I recently created a small install script which installs Java 8 (on Debian-based Systems), downloads and unpacks Apache Spark and RumbleDB and adds a REPL and JSONiq File convenience script.

I guess it could be useful to other linux users, and make installation clean and easy, with everything in one folder and an uninstall script.

Should I create a git repo with a release containing the script, and create a pull request with an edit to the documentation to point to it? Or can you integrate it in this repo? Does it even meet your standards?

Thanks in advance for any tips and/or feedback

P.S.: Here is the code:

#!/bin/bash

INSTALL_PATH=$HOME/RumbleDB
APACHE_SPARK_RELEASE=3.2.0
HADOOP_RELEASE=3.2
RUMBLEDB_VERSION=1.15.0

if [ "$1" == "--help" ]; then
	echo "Usage:"
	echo "    ./install-rumbledb.sh"
	echo "Downloads Apache Spark and RumbleDB (and installs Java 8 if not present)"
	echo "Creates a RumbleDB REPL script and a RumbleDB File execution script"
	echo "Scripts are added to PATH via an edit to ~/.bashrc"
	echo "Everything, including an uninstaller, will be stored in $INSTALL_PATH"
	exit 0
fi

echo "######### checking if we need to install Java"
if [ $(which java) ]; then
	echo "Java version is $(java --version). Version 8 is recommended, but 11 might work, too."
else
	if [ $(which apt-get) ]; then
		echo "No Java found, trying to install OpenJDK 8 JRE via apt-get"
		echo "You will need to enter your password for this"
		sudo apt-get install openjdk-8-jre
	else
		echo "No Java and no apt-get package manager found, please install Java 8"
		echo "and then return."
		exit 1
	fi
fi

echo "########## creating installation folder"
mkdir $INSTALL_PATH
cd $INSTALL_PATH

echo "########## downloading Apache Spark"
SPARK_FILENAME=spark-$APACHE_SPARK_RELEASE-bin-hadoop$HADOOP_RELEASE.tgz
wget https://dlcdn.apache.org/spark/spark-$APACHE_SPARK_RELEASE/$SPARK_FILENAME
wget https://downloads.apache.org/spark/KEYS
wget https://downloads.apache.org/spark/spark-$APACHE_SPARK_RELEASE/$SPARK_FILENAME.asc

echo "##### verifying download if gpg is installed"
if [ $(which gpg) ]; then
	echo "### importing apache spark signing keys"
	gpg --import KEYS
	echo "### verifying"
	gpg --verify $SPARK_FILENAME.asc $SPARK_FILENAME
	echo "### note:"
	echo "a warning like 'There is no indication that the signature belongs to the owner.'"
	echo "can relatively safely be ignored, to alleviate it you would need to get in"
	echo "touch with an Apache Spark developer."
else
	echo "gpg not installed, could not verify download"
fi

echo "if there is reason for concern, you have 5s to abort"

sleep 5

echo "##### unpacking file"
tar -xzf $SPARK_FILENAME

echo "########### downloading RumbleDB"
wget "https://github.com/RumbleDB/rumble/releases/download/v$RUMBLEDB_VERSION/rumbledb-$RUMBLEDB_VERSION.jar"

echo "########### creating scripts"
cat > rumble-repl << EOF
#!/bin/bash
RUMBLE_JAR_PATH=$INSTALL_PATH/rumbledb-$RUMBLEDB_VERSION.jar
SPARK_HOME=$INSTALL_PATH/spark-$APACHE_SPARK_RELEASE-bin-hadoop$HADOOP_RELEASE
JAVA_VERSION=\$(java --version)
if [ "\$1" == "--help" ]; then
	echo "Usage:";
	echo "    rumble-repl"
	echo "Invokes an interactive RumbleDB Shell using downloaded JARs and Binaries"
	echo "Rumble Jar taken from \$RUMBLE_JAR_PATH"
	echo "Spark resides at \$SPARK_HOME"
	echo "Current version of Java: \$JAVA_VERSION"
	exit 0
fi

\$SPARK_HOME/bin/spark-submit \$RUMBLE_JAR_PATH --shell yes --output-format json
EOF

cat > rumble-file << EOF
#!/bin/bash
RUMBLE_JAR_PATH=$INSTALL_PATH/rumbledb-$RUMBLEDB_VERSION.jar
SPARK_HOME=$INSTALL_PATH/spark-$APACHE_SPARK_RELEASE-bin-hadoop$HADOOP_RELEASE
JAVA_VERSION=\$(java --version)
if [ "\$1" == "--help" ]; then
	echo "Usage:";
	echo "    rumble-file <file>"
	echo "Execute JSONiq query taken from <file> using RumbleDB"
	echo "Rumble Jar taken from \$RUMBLE_JAR_PATH"
	echo "Spark resides at \$SPARK_HOME"
	echo "Current version of Java: \$JAVA_VERSION"
	exit 0
fi

\$SPARK_HOME/bin/spark-submit \$RUMBLE_JAR_PATH --query-path \$1 --output-format json
EOF

cat > uninstall-rumble << EOF
#!/bin/bash
if [ "\$1" == "--help" ]; then
	echo "Usage:";
	echo "    uninstall-rumble"
	echo "Uninstalls Rumble by removing $INSTALL_PATH and"
	echo "removing the PATH modification in ~/.bashrc"
	echo "If the installer installed Java, it will stay installed."
	exit 0
fi

echo "Removing Path modification from ~/.bashrc"
cp ~/.bashrc bashrc-old
cat bashrc-old | sed 's/PATH=\$PATH:$(echo $INSTALL_PATH | sed 's/\//\\\//g')\/scripts//g' > ~/.bashrc

echo "Deleting files"
cd ~
rm -rf $INSTALL_PATH
EOF

mkdir scripts
mv rumble-repl rumble-file uninstall-rumble scripts
chmod +x scripts/*

echo "########## editing ~/.bashrc to add scripts to path"
echo "PATH=\$PATH:$INSTALL_PATH/scripts" >> ~/.bashrc

echo "########## done. reloading ~/.bashrc and showing help output of scripts"
source ~/.bashrc
rumble-file --help
rumble-repl --help
uninstall-rumble --help

[ghislainfourny]

Dear @fkellner

Thank you for your contribution. It is highly appreciated. I have forwarded it to the team in order to give you feedback.

I think that the best approach would be if you could create a git repository (owned by you) with your script, and we will gladly point to it in our documentation. That way, it will be straightforward for you to maintain it without having to go through reviews every time.

Thanks again for this nice contribution!

Kind regards,
Ghislain Fourny

[thomastzhou]

Dear @fkellner

I tried it on CentOS 7 and with Oracle JDK 1.8.0_91. The script worked great for me and was very convenient overall. A small issue is that the JDK I'm using expects the flag -version instead of --version (which I believe is true for all Oracle JDKs <= 8) so it shows an error in some non-critical branches. I have not tested but I believe OpenJDK parses both flags so maybe a simple update to use the -version flag could be more robust.

Let us know if you make a repository with the script and we will happily point to it from the documentation!

[fkellner]

Dear @ghislainfourny and @thomastzhou,

thank you for the quick feedback, and sorry for the long response time!
I incorporated the flag change and published the install script here:
https://github.com/fkellner/rumbledb-install-script

Best Regards
Florian Kellner

[ghislainfourny]

Dear @fkellner

Many thanks for wrapping this up. This is very appreciated. In the next release the documentation will point to your script as an additional possibility of installation, you can see a preview here:

https://sparksoniq.readthedocs.io/en/latest/Getting%20started/

Also, in the next release the CLI will be more convenient (but in a backward compatible way so it should not break your script). Feel free to adjust your script if you wish to once the release is out:
https://sparksoniq.readthedocs.io/en/latest/CLI/

Kind regards,
Ghislain Fourny