Skip to content

Docker image template for Python tools in V-FOR-WaTer.

License

Notifications You must be signed in to change notification settings

VForWaTer/tool_template_python

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

24 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

tool_template_python

Docker Image CI DOI

This is the template for a generic containerized Python tool following the Tool Specification for reusable research software using Docker.

This template can be used to generate new Github repositories from it.

How generic?

Tools using this template can be run by the toolbox-runner. That is only convenience, the tools implemented using this template are independent of any framework.

The main idea is to implement a common file structure inside container to load inputs and outputs of the tool. The template shares this structures with the R template, NodeJS template and Octave template, but can be mimiced in any container.

Each container needs at least the following structure:

/
|- in/
|  |- parameters.json
|- out/
|  |- ...
|- src/
|  |- tool.yml
|  |- run.py
  • parameters.json are parameters. Whichever framework runs the container, this is how parameters are passed.
  • tool.yml is the tool specification. It contains metadata about the scope of the tool, the number of endpoints (functions) and their parameters
  • run.py is the tool itself, or a Python script that handles the execution. It has to capture all outputs and either print them to console or create files in /out

How to build the image?

You can build the image from within the root of this repo by

docker build -t tbr_python_tempate .

Use any tag you like. If you want to run and manage the container with toolbox-runner they should be prefixed by tbr_ to be recognized.

Alternatively, the contained .github/workflows/docker-image.yml will build the image for you on new releases on Github. You need to change the target repository in the aforementioned yaml.

How to run?

This template installs the json2args python package to parse the parameters in the /in/parameters.json. This assumes that the files are not renamed and not moved and there is actually only one tool in the container. For any other case, the environment variables PARAM_FILE can be used to specify a new location for the parameters.json and TOOL_RUN can be used to specify the tool to be executed. The run.py has to take care of that.

To invoke the docker container directly run something similar to:

docker run --rm -it -v /path/to/local/in:/in -v /path/to/local/out:/out -e TOOL_RUN=foobar tbr_python_template

Then, the output will be in your local out and based on your local input folder. Stdout and Stderr are also connected to the host.

With the toolbox runner, this is simplyfied:

from toolbox_runner import list_tools
tools = list_tools() # dict with tool names as keys

foobar = tools.get('foobar')  # it has to be present there...
foobar.run(result_path='./', foo_int=1337, foo_string="Please change me")

The example above will create a temporary file structure to be mounted into the container and then create a .tar.gz on termination of all inputs, outputs, specifications and some metadata, including the image sha256 used to create the output in the current working directory.

What about real tools, no foobar?

Yeah.

  1. change the tool.yml to describe your actual tool
  2. add any pip install or apt-get install needed to the dockerfile
  3. add additional source code to /src
  4. change the run.py to consume parameters and data from /in and useful output in out
  5. build, run, rock!