Quickstart

• The pip-package is called ipython-cache

• The python module is called cache_magic

• The magic is called %cache

So you can run the magic by entering this into an ipython-cell:

!pip install ipython-cache
import cache_magic
%cache a = 1+1
%cache

install directly from notebook

1. open jupyter notebook

2. create new cell

3. enter !pip install cache-magic

4. execute

install into conda-environment

conda create -n test
source activate test
conda install -c juergens ipython-cache
jupyter notebook

example

%cache myVar = someSlowCalculation(some, "parameters")

This will calculate someSlowCalculation(some, "parameters") once. And in subsequent calls it restores myVar from storage.

The magic turns this example into something like this (if there was no ipython-kernel and no versioning):

try:
  with open("myVar.txt", 'rb') as fp:
    myVar = pickle.loads(fp.read())
except:
  myVar = someSlowCalculation(some, "parameters")
  with open("myVar.txt", 'wb') as fp:
    pickle.dump(myVar, fp)

general form

%cache <variable> = <expression>

Variable: This Variable’s value will be fetched from cache.

Expression: This will only be excecuted once and the result will be stored to disk.

full form

%cache [--version <version>] [--reset] [--debug] variable [= <expression>]

-v or –version: either a variable name or an integer. Whenever this changes, a new value is calculated (instead of returning an old value from the cache).

if version is ‘*’ or omitted, the hashed expression is used as version, so whenever the expression changes, a new value is cached.

-r or –reset: delete the cached value for this variable. Forces recalculation, if <expression> is present

-d or –debug: additional logging

show cache

%cache

shows all variables in cache as html-table

full reset

%cache -r
%cache --reset

deletes all cached values for all variables

where is the cache stored?

In the directory where the kernel was started (usually where the notebook is located) in a subfolder called .cache_magic

push to pypi

prepare environment:

gedit ~/.pypirc
chmod 600 ~/.pypirc
sudo apt install pandoc

upload changes to test and production:

pandoc -o README.rst README.md
restview --pypi-strict README.rst
# update version in setup.py
rm -r dist
python setup.py sdist
twine upload dist/* -r testpypi
firefox https://testpypi.python.org/pypi/ipython-cache
twine upload dist/*

test install from testpypi

pip install --index-url https://test.pypi.org/simple/ --extra-index-url https://pypi.org/simple ipython_cache --no-cache-dir --user

test installation

sudo pip install ipython_cache --no-cache-dir --user

editable import

Install into environment with -e:

!pip install -e .

reload after each change:

import cache_magic
from imp import reload
reload(cache_magic)

Alternatively (if you don’t want to install python, jupyter & co), you can use the docker-compose.yml for development:

cd ipython-cache
docker-compose up

create Conda Packet

requires the bash with latest anaconda on path

bash
mkdir test && cd test
conda skeleton pypi ipython-cache
conda-build ipython-cache -c conda-forge
anaconda upload /home/juergens/anaconda3/conda-bld/linux-64/ipython-cache-*

running tests

bash
conda remove --name test --all
conda env create -f test/environment.yml
source activate test
conda remove ipython-cache
pip uninstall ipython_cache
pip install -e .
./test/run_example.py

If there is any error, it will be printed to stderr and the script fails.

the output can be found in “test/temp”.