mkdir my_project
cd my_project
vim setup.py
try:
from setuptools import setup, find_packages
except ImportError:
from distutils.core import setup, find_packages
setup(
name="project_name",
version="0.1.0",
packages=find_packages(),
install_requires=["mistral", "yaql"],
entry_points={
"mistral.yaql_functions": [
"random_uuid = my_package.sub_package.yaql:random_uuid_"
]
}
)
Publish the random_uuid_ function in the entry_points section, in the mistral.yaql_functions namespace in setup.py. This function will be defined later.
Note that the package name will be used in Pip and must not overlap with other packages installed. project_name may be replaced by something else. The package name (my_package here) may overlap with other packages, but module paths (.py files) may not.
For example, it is possible to have a mistral package (though not recommended), but there must not be a mistral/version.py file, which would overlap with the file existing in the original mistral package.
yaql and mistral are the required packages. mistral is necessary in this example only because calls to the Mistral Python DB API are made.
For each entry point, the syntax is:
"<name_of_YAQL_expression> = <path.to.module>:<function_name>"
stevedore will detect all the entry points and make them available to all Python applications needing them. Using this feature, there is no need to modify Mistral’s core code.
A package folder is directory with a __init__.py file. Create a file that will contain the custom YAQL functions. There are no restrictions on the paths or file names used.
mkdir -p my_package/sub_package
touch my_package/__init__.py
touch my_package/sub_package/__init__.py
That function might have context as first argument to have the current YAQL context available inside the function.
cd my_package/sub_package
vim yaql.py
from uuid import uuid5, UUID
from time import time
def random_uuid_(context):
"""generate a UUID using the execution ID and the clock"""
# fetch the current workflow execution ID found in the context
execution_id = context['__execution']['id']
time_str = str(time())
execution_uuid = UUID(execution_id)
return uuid5(execution_uuid, time_str)
This function returns a random UUID using the current workflow execution ID as a namespace.
The context argument will be passed by Mistral YAQL engine to the function. It is invisble to the user. It contains variables from the current task execution scope, such as __execution which is a dictionary with information about the current workflow execution such as its id.
Note that errors can be raised and will be displayed in the task execution state information in case they are raised. Any valid Python primitives may be returned.
The context argument is optional. There can be as many arguments as wanted, even list arguments such as *args or dictionary arguments such as **kwargs can be used as function arguments.
For more information about YAQL, read the official YAQL documentation.
curl https://bootstrap.pypa.io/get-pip.py | python
pip install --upgrade setuptools
cd -
pip install .
The function name in Python random_uuid_ does not matter, only the entry point name random_uuid does.
my_workflow:
tasks:
my_action_task:
action: std.echo
publish:
random_id: <% random_uuid() %>
input:
output: "hello world"
After any new created functions or any modification in the code, re-run pip install . and restart Mistral.
While developing, it is sufficient to add the root source folder (the parent folder of my_package) to the PYTHONPATH environment variable and the line random_uuid = my_package.sub_package.yaql:random_uuid_ in the Mistral entry points in the mistral.yaql_functions namespace. If the path to the parent folder of my_package is /path/to/my_project.
export PYTHONPATH=$PYTHONPATH:/path/to/my_project
vim $(find / -name "mistral.*egg-info*")/entry_points.txt
[entry_points]
mistral.yaql_functions =
random_uuid = my_package.sub_package.yaql:random_uuid_