Jupyter Notebook for ISIS/Slang
If you want to start an ISIS session with a jupyter notebook this should be straight forward on the cluster. For this you need to load two modules via the module system
module load conda slang-jupyter
Once this is done you need to access the conda environment via
conda activate
Here you can start the jupyter server, but make sure that the shell where you start the server has all needed modules loaded (e.g., if you need localmodels, for example tbnew, make sure it is activated. If you never actively loaded a module to use ISIS, you are probably good).
Start the server
jupyter notebook
now you should be greeted with the usual page from the notebooks where you can open notebooks etc. The folder were you started the server is the root folder and you can not go above this level. If you want to start a ISIS notebook, just create a new notebook and select isis (there is also a slang kernel, that is not the same!). Having the notebook opened you should be able to run commands in the cells as usual. The notebook acts as one document, so functions and variables defined in one cell are accessible in another (provided the cell with the definitions was executed first). For details about the notebook interactions check the manual of jupyter itself.
Embedding images
Currently the kernel (that is the part of the software that translates your input to isis) has no built in feature to display images. If you issue a plot command and you are locally at the machine a normal pgplot will open just as it is with the isis shell. The simplest way to include an image in the notebook is to plot into a png file
variable pg = open_plot("my-cool-plot.png/png"); % this opens a plot to a png file, any plot command afterwards will draw to this file
xlabel("X-axis");
ylabel("Y-axis");
plot([0:4:#100], sin([0:4:#100]));
oplot([0,4], [0,1]);
close_plot(pg); % this is important, if you don't close the file, any additional plot command will create a new file with appended counter
The created png can be included in the notebook with a markdown cell
where the text in square brackets is the description, this can be left empty (but the brackets are necessary).
This way you can also use xfig or tikz to create pdfs or png and embed them accordingly.
Remote Access
If you want to work on a notebook remotely then this is not directly possible for various reasons but mainly because this would allow everyone to execute arbitrary code on our cluster (and, in case this was not clear, we don't want that). There is, however, a way to forward the server via an ssh tunnel between your computer (let's say you use a laptop) and the cluster.
To do so, you first have to get the information about where the server is running
jupyter notebook list # make sure the conda environment is activated to have access to jupyter
This will show you a line similar to
http://localhost:8888/?token=2b9af3b8973be0b3270d0cf93941969db373bd70bfe195ea :: /home/stierhof
where the first part is the address where the server is accessible (locally) and the second is the root folder where it is running. Now to access the server on your laptop, create the tunnel with
ssh -NL 8888:localhost:8888 <user>@<computer>
This command will not return and seemingly do nothing! You need to keep the shell where you started the tunnel open, if you close it the tunnel will be closed too. In this example the number 8888 is the port number from the listed notebook above, if you are shown a different number you also have to change it in the ssh command (but 8888 is the default port). Once the ssh tunnel connects you local port with the port on the cluster you can paste the full address as shown by the jupyter notebook list command into a browser and access the server from there in the usual way.
Disclaimer
The software that enables isis to be used in a jupyter notebook is not very well tested yet. If you find any problems, or better yet, have solutions to problems or ideas for features, please let me know - Jakob Stierhof