Build instructions from ReadMe.md don't work on Windows

[Jaykul]

1. The line source activate beakerx needs to be just activate beakerx

2. The line: ./gradlew --no-daemon :beakerx:install has to be run as administrator in order to make symlinks...

3. The line python -m beakerx.install --enable --prefix="${CONDA_PREFIX}" would need to be `python -m beakerx.install --enable --prefix="%CONDA_PREFIX%"

[Jaykul]

Here's a PowerShell version that definitely works (assuming conda and yarn are pre-installed, and that you're running elevated):

conda create -y -n beakerx python=3.5 jupyter pandas
# activate beakerx # seems to work, but doesn't leave CONDA_PREFIX set
cmd.exe /c "activate beakerx && set > beaker.env"
gc .\beaker.env | ForEach { $name,$value = $_ -split "=",2; sc Env:$name $value }

conda install -y -c conda-forge openjdk
./gradlew --no-daemon build
./gradlew --no-daemon kernelInstall
./gradlew --no-daemon :beakerx:install 
cd beakerx
pip install -e .
python -m beakerx.install --enable --prefix="${CONDA_PREFIX}"
jupyter notebook

I think this works for DOS:

conda create -y -n beakerx python=3.5 jupyter pandas
activate beakerx
conda install -y -c conda-forge openjdk
./gradlew --no-daemon build
./gradlew --no-daemon kernelInstall
./gradlew --no-daemon :beakerx:install
cd beakerx
pip install -e .
python -m beakerx.install --enable --prefix="%CONDA_PREFIX%"
jupyter notebook

[altavir]

Here is the slightly modified version that allows to skip tests and not to reset gradle daemon each time:

conda create -y -n beakerx python=3.5 jupyter pandas
# activate beakerx # seems to work, but doesn't leave CONDA_PREFIX set
cmd.exe /c "activate beakerx && set > beaker.env"
gc .\beaker.env | ForEach { $name,$value = $_ -split "=",2; sc Env:$name $value }

conda install -y -c conda-forge openjdk
./gradlew --no-daemon build kernelInstall :beakerx:install -x test
cd beakerx
pip install -e .
python -m beakerx.install --enable --prefix="${CONDA_PREFIX}"
cd ..
#jupyter notebook

I've created a file named install.ps1 and now can run it from installation directory. For now the only problem is that groovy kernel fails to build.

[altavir]

I've performed some additional searches. It seems to be this problem with changing environments: conda/conda#1519

[Jaykul]

Yes @altavir, that's exactly it.

People need to stop using cmd -- especially people who're trying to write cross-platform things.

I have Bash on Windows, and I even have PowerShell on Linux -- but CMD is dead to me 😉 I gave them a tip to do roughly the same thing I showed above to slurp up the environment changes from the cmd process...

Fwiw, my groovy kernel works and everything. The only thing that made me sad was realizing that the %%magic commands wouldn't automatically pickup all my kernels and let me mix-and-match whatever I wanted 😖

I'm working on a cross-platform PowerShell kernel for Jupyter, but I love the polyglot story of Beaker -- are there docs of how you implemented the communication? I mean, what would a Beaker-compatible kernel need to do beyond what Jupyter requires?

[altavir]

The same for me. I also have bash installed on windows, though I mostly use it for ssh only, since powershell is much more... well... powerful and concise (it also supports pretty large POSIX subset not to feel any difference). I want to use Powershell on Linux also, but I am waiting for a repository debian release.

[scottdraves]

Thanks for the report. we are stuck with what conda does but we can make it correct and document the windows workarounds at least.

[pescadorbob]

Here's a PowerShell version that definitely works (assuming conda and yarn are pre-installed, and that you're running elevated):

conda create -y -n beakerx python=3.5 jupyter pandas
# activate beakerx # seems to work, but doesn't leave CONDA_PREFIX set
cmd.exe /c "activate beakerx && set > beaker.env"
gc .\beaker.env | ForEach { $name,$value = $_ -split "=",2; sc Env:$name $value }

conda install -y -c conda-forge openjdk
./gradlew --no-daemon build
./gradlew --no-daemon kernelInstall
./gradlew --no-daemon :beakerx:install 
cd beakerx
pip install -e .
python -m beakerx.install --enable --prefix="${CONDA_PREFIX}"
jupyter notebook

I think this works for DOS:

conda create -y -n beakerx python=3.5 jupyter pandas
activate beakerx
conda install -y -c conda-forge openjdk
./gradlew --no-daemon build
./gradlew --no-daemon kernelInstall
./gradlew --no-daemon :beakerx:install
cd beakerx
pip install -e .
python -m beakerx.install --enable --prefix="%CONDA_PREFIX%"
jupyter notebook

I've tried this, but, the gradlew doesn't exist anywhere for me. Is it supposed to be executed from a particular directory or something? where are the build.gradle files we're hoping for?

[Jaykul]

gradlew is part of the source 😉