RiotKit-Do (RKD) usage and development manual¶

RKD is a stable, open-source, multi-purpose automation tool which balance flexibility with simplicity. The primary language is Python and YAML syntax.

RiotKit-Do can be compared to Gradle and to GNU Make, by allowing both Python and Makefile-like YAML syntax.

What I can do with RKD?

Simplify the scripts
Put your Python and Bash scripts inside a YAML file (like in GNU Makefile)
Do not reinvent the wheel (argument parsing, logs, error handling for example)
Share the code across projects and organizations, use native Python Packaging to share tasks (like in Gradle)
Natively integrate scripts with .env files
Automatically generate documentation for your scripts
Maintain your scripts in a good standard

RKD can be used on PRODUCTION, for development, for testing, to replace some of Bash scripts inside docker containers, and for many more, where Makefile was used.

Example use cases¶

Docker based production environment with multiple configuration files, procedures (see: Harbor project)
Database administrator workspace (importing dumps, creating new user accounts, plugging/unplugging databases)
Development environment (executing migrations, importing test database, splitting tests and running parallel)
On CI (prepare project to run on eg. Jenkins or Gitlab CI) - RKD is reproducible on local computer which makes inspection easier
Kubernetes/OKD deployment workspace (create shared YAML parts with JINJA2 between multiple environments and deploy from RKD)
Automate things like certificate regeneration on production server, RKD can generate any application configs using JINJA2
Installers (RKD has built-in commands for replacing lines in files, modifying .env files, asking user questions and validating answers)

Install RKD¶

RiotKit-Do is delivered as a Python package that can be installed system-wide or in a virtual environment. The virtual environment installation is similar in concept to the Gradle wrapper (gradlew)

# 1) via PIP
pip install rkd

# 2) Create project (will create a virtual env and commit files to GIT)
rkd :rkd:create-structure --commit

Getting started in freshly created structure¶

The “Quick start” section ends up with a .rkd directory, a requirements.txt and setup-venv.sh

Use eval $(setup-venv.sh) to enter shell of your project, where RKD is installed with all dependencies
Each time you install anything from pip in your project - add it to requirements.txt, you can install additional RKD tasks from pip
In .rkd/makefile.yaml you can start adding your first tasks and imports

Create your first task with Beginners guide - on YAML syntax example ¶

Check how to use commandline to run tasks in RKD with Commandline basics ¶

See how to import existing tasks to your Makefile with Importing tasks page¶

Keep learning¶

YAML syntax is described also in Tasks development section
Writing Python code in makefile.yaml requires to lookup Tasks API
Learn how to import installed tasks via pip - Importing tasks
You can also write tasks code in pure Python and redistribute those tasks via Python’s PIP - see Tasks development
With RKD you can create interactive installers - check the Creating installer wizards with RKD section

Beginners guide - on YAML syntax example¶

Where to place files¶

.rkd directory must always exists in your project. Inside .rkd directory you should place your makefile.yaml that will contain all of the required tasks.

Just like in UNIX/Linux, and just like in Python - there is an environment variable RKD_PATH that allows to define multiple paths to .rkd directories placed in other places - for example outside of your project. This gives a flexibility and possibility to build system-wide tools installable via Python’s PIP.

Environment variables¶

RKD natively reads .env (called also “dot-env files”) at startup. You can define default environment values in .env, or in other .env-some-name files that can be included in env_files section of the YAML.

Scope of environment variables

env_files and environment blocks can be defined globally, which will end in including that fact in each task, second possibility is to define those blocks per task. Having both global and per-task block merges those values together and makes per-task more important.

Example

version: org.riotkit.rkd/yaml/v1
environment:
    PYTHONPATH: "/project"
tasks:
    :hello:
        description: Prints variables
        environment:
            SOME_VAR: "HELLO"
        steps: |
            echo "SOME_VAR is ${SOME_VAR}, PYTHONPATH is ${PYTHONPATH}"

Arguments parsing¶

Arguments parsing is a strong side of RKD. Each task has it’s own argument parsing, it’s own generated –help command. Python’s argparse library is used, so Python programmers should feel like in home.

Example

version: org.riotkit.rkd/yaml/v1
environment:
    PYTHONPATH: "/project"
tasks:
    :hello:
        description: Prints your name
        arguments:
            "--name":
                required: true
                #option: store_true # for booleans/flags
                #default: "Unknown" # for default values
        steps: |
            echo "Hello ${ARG_NAME}"

rkd :hello --name Peter

Defining tasks in Python code¶

Defining tasks in Python gives wider possibilities - to access Python’s libraries, better handle errors, write less tricky code. RKD has a similar concept to hashbangs in UNIX/Linux.

There are two supported hashbangs + no hashbang:

#!python
#!bash
(just none there)

What can I do in such Python code? Everything! Import, print messages, execute shell commands, everything.

Example

version: org.riotkit.rkd/yaml/v1
environment:
    PYTHONPATH: "/project"
tasks:
    :hello:
        description: Prints your name
        arguments:
            "--name":
                required: true
                #option: store_true # for booleans/flags
                #default: "Unknown" # for default values
        steps: |
            #!python
            print('Hello %s' % ctx.get_arg('--name'))

Special variables

this - instance of current TaskInterface implementation
ctx - instance of ExecutionContext

Please check Tasks API for those classes reference.

YAML syntax reference¶

Let’s at the beginning start from analyzing an example.

version: org.riotkit.rkd/yaml/v1

# optional: Import tasks from Python packages
# This gives a possibility to publish tasks and share across projects, teams, organizations
imports:
    - rkt_utils.db.WaitForDatabaseTask

# optional environment section would append those variables to all tasks
# of course the tasks can overwrite those values in per-task syntax
environment:
    PYTHONPATH: "/project/src"

# optional env files loaded there would append loaded variables to all tasks
# of course the tasks can overwrite those values in per-task syntax
#env_files:
#    - .some-dotenv-file

tasks:
    :check-is-using-linux:
        description: Are you using Linux?
        # use sudo to become a other user, optional
        become: root
        steps:
            # steps can be defined as single step, or multiple steps
            # each step can be in a different language
            # each step can be a multiline string
            - "[[ $(uname -s) == \"Linux\" ]] && echo \"You are using Linux, cool\""
            - echo "step 2"
            - |
                #!python
                print('Step 3')

    :hello:
        description: Say hello
        arguments:
            "--name":
                help: "Your name"
                required: true
                #default: "Peter"
                #option: "store_true" # for booleans
        steps: |
            echo "Hello ${ARG_NAME}"

            if [[ $(uname -s) == "Linux" ]]; then
                echo "You are a Linux user"
            fi

imports - Imports external tasks installed via Python’ PIP. That’s the way to easily share code across projects

environment - Can define default values for environment variables. Environment section can be defined for all tasks, or per task

env_files - Includes .env files, can be used also per task

tasks - List of available tasks, each task has a name, descripton, list of steps (or a single step), arguments

Running the example:

Create a .rkd directory
Create .rkd/makefile.yaml file
Paste/rewrite the example into the .rkd/makefile.yaml
Run rkd :tasks from the directory where the .rkd directory is placed
Run defined tasks rkd :hello :check-is-using-linux

Extended usage - Makefile in Python syntax¶

Not only tasks can be written in Python code, but Makefile too - such makefile is called makefile.py, and placed in .rkd directory.

Example:

import os
from rkd.api.syntax import TaskAliasDeclaration as Task  # RKD API (for defining shortcuts/aliases for whole tasks lists)
from rkd.api.syntax import TaskDeclaration               # RKD API (for declaring usage of given task, importing it)
from rkd.api.contract import ExecutionContext            # RKD API (one of dependencies - context gives us access to commandline arguments and environment variables)
from rkd_python import imports as PythonImports          # group of imports (not all packages supports it, but most of them)
from rkd.standardlib.jinja import FileRendererTask       # single task
from rkd.standardlib import CallableTask                 # Basic Python callable task for a little bit advanced usage
# from .mypackage import MyTask                          # import your task from local package


def example_method(ctx: ExecutionContext, task: CallableTask) -> bool:
    os.system('xdg-open https://twitter.com/wrkclasshistory')
    return True


IMPORTS = [
    # We declare that we will use this task.
    # Declaration can take some additional arguments like args= or env=, to always append environment and/or commandline switches
    # regardless of if user used it
    TaskDeclaration(FileRendererTask()),
    # remember about the "," between tasks, it's an array/list ;)
    # TaskDeclaration(MyTask())

    TaskDeclaration(CallableTask(':read-real-history', example_method, description='Example task with simple Python code'))
]

IMPORTS += PythonImports()

TASKS = [
    # declared task-aliases. A Task Alias is a shortcut eg. ":release" that will expands to ":py:build :py:publish --username= (...)"
    # the best feature in task-aliases is that you can append and overwrite last commandline arguments, those will be added
    # at the end of the command
    Task(':release', description='Release to PyPI (snapshot when on master, release on tag)',
         to_execute=[
            ':py:build', ':py:publish', '--username=__token__', '--password=${PYPI_TOKEN}'
         ]),

    Task(':test', [':py:unittest'], description='Run unit tests'),
    Task(':docs', [':sh', '-c', ''' set -x
        cd docs
        rm -rf build
        sphinx-build -M html "source" "build"
    '''])
]

The Python syntax is very flexible
You can create your own local packages and import them here, create own advanced structure
Possibility to declare aliases and adjust TaskDeclarations for advanced usage (YAML syntax does not offer this)

Check Detailed usage page for description of all environment variables, mechanisms, good practices and more¶

Commandline basics¶

RKD command-line usage is highly inspired by GNU Make and Gradle, but it has its own extended possibilities to make your scripts smaller and more readable.

Tasks are prefixed always with “:”.
Each task can handle it’s own arguments (unique in RKD)
“@” allows to propagate arguments to next tasks (unique in RKD)

Tasks arguments usage in shell and in scripts¶

Executing multiple tasks in one command:

rkd :task1 :task2

Multiple tasks with different switches:

rkd :task1 --hello  :task2 --world --become=root

Second task will run as root user, additionally with --world parameter.

Tasks sharing the same switches

Both tasks will receive switch “–hello”

# expands to:
#  :task1 --hello
#  :task2 --hello
rkd @ --hello :task1 :task2

# handy, huh?

Advanced usage of shared switches

Operator “@” can set switches anytime, it can also clear or replace switches in NEXT TASKS.

# expands to:
#   :task1 --hello
#   :task2 --hello
#   :task3
#   :task4 --world
#   :task5 --world
rkd @ --hello :task1 :task2 @ :task3 @ --world :task4 :task5

Written as a pipeline (regular bash syntax)

It’s exactly the same example as above, but written multiline. It’s recommended to write multiline commands if they are longer.

rkd @ --hello \
    :task1 \
    :task2 \
    @
    :task3 \
    @ --world \
    :task4 \
    :task5

Importing tasks¶

Tasks can be defined as installable Python’s packages that you can import in your Makefile

Please note:

To import a group, the package you try to import need to hvve a defined imports() method inside of the package.
The imported group does not need to import automatically dependend tasks (but it can, it is recommended), you need to read into the docs of specific package if it does so

1) Install a package¶

RKD defines dependencies using Python standards.

Example: Given we want to import tasks from package “rkt_armutils”.

echo "rkt_armutils==3.0" >> requirements.txt
pip install -r requirements.txt

Good practices:

Use fixed versions eg. 3.0 or even 3.0.0 and upgrade only intentionally to reduce your work. Automatic updates, especially of major versions

could be unpredictable and possibly can break something time-to-time

How do I check latest version?:

Simply install a package eg. pip install rkt_armutils, then do a pip show rkt_armutils and write the version

to the requirements.txt, or lookup a package first at https://pypi.org/project/rkt_armutils/ (where rkt_armutils is an example package)

2) In YAML syntax¶

Example: Given we want to import task “InjectQEMUBinaryIntoContainerTask”, or we want to import whole “rkt_armutils.docker” group

imports:
    # Import whole package, if the package defines a group import (method imports())
    - rkt_armutils.docker

    # Or import single task
    - rkt_armutils.docker.InjectQEMUBinaryIntoContainerTask

2) In Python syntax¶

Example: Given we want to import task “InjectQEMUBinaryIntoContainerTask”, or we want to import whole “rkt_armutils.docker” group

from rkd.api.syntax import TaskDeclaration
from rkt_armutils.docker import InjectQEMUBinaryIntoContainerTask

# ... (use "+" operator to append, remove "+" if you didn't define any import yet)
IMPORTS += [TaskDeclaration(InjectQEMUBinaryIntoContainerTask)]

Ready to go? Check Built-in tasks that you can import in your Makefile¶

Detailed usage¶

Troubleshooting¶

Output is corrupted or there is no output from a shell command executed inside of a task

The output capturing is under testing. The Python’s subprocess module is skipping “sys.stdout” and “sys.stderr” by writing directly to /dev/stdout and /dev/stderr, which makes output capturing difficult.

Run rkd in compat mode to turn off output capturing from shell commands:

RKD_COMPAT_SUBPROCESS=true rkd :some-task-here

Loading priority¶

Environment variables loading order from .env and from .rkd¶

Legend: Top is most important, the variables loaded on higher level are not overridden by lower level

Operating system environment
Current working directory .env file
.env files from directories defined in RKD_PATH

Environment variables loading order in YAML syntax¶

Legend: Top - is most important

Operating system environment
.env file
Per-task “environment” section
Per-task “env_file” imports
Global “environment” section
Global “env_file” imports

Order of loading of makefile files in same .rkd directory¶

Legend: Lower has higher priority (next is appending changes to previous)

*.py
*.yaml
*.yml

Paths and inheritance¶

RKD by default search for .rkd directory in current execution directory - ./.rkd.

The search order is following (from lower to higher load priority):

RKD’s internals (we provide a standard tasks like :tasks, :init, :sh, :exec and more)
/usr/lib/rkd
User’s home ~/.rkd
Current directory ./.rkd
RKD_PATH

Custom path defined via environment variable

RKD_PATH allows to define multiple paths that would be considered in priority.

export RKD_PATH="/some/path:/some/other/path:/home/user/riotkit/.rkd-second"

How the makefiles are loaded?

Each makefile is loaded in order, next makefile can override tasks of previous. That’s why we at first load internals, then your tasks.

Tasks execution¶

Tasks are executed one-by-one as they are specified in commandline or in TaskAlias declaration (commandline arguments).

rkd :task-1 :task-2 :task-3

task-1
task-2
task-3

A –keep-going can be specified after given task eg. :task-2 –keep-going, to ignore a single task failure and in consequence allow to go to the next task regardless of result.

Tasks development¶

RKD has two approaches to define a task. The first one is simpler - in makefile in YAML or in Python. The second one is a set of tasks as a Python package.

Creating simple tasks in YAML syntax¶

Example 1:

version: org.riotkit.rkd/yaml/v1
imports:
  - rkd.standardlib.jinja.RenderDirectoryTask

tasks:
  # see this task in "rkd :tasks"
  # run with "rkd :examples:bash-test"
  :examples:bash-test:
      description: Execute an example command in bash - show only python related tasks
      steps: |
             echo "RKD_DEPTH: ${RKD_DEPTH} # >= 2 means we are running rkd-in-rkd"
             echo "RKD_PATH: ${RKD_PATH}"
             rkd --silent :tasks | grep ":py"

  # try "rkd :examples:arguments-test --text=Hello --test-boolean"
  :examples:arguments-test:
      description: Show example usage of arguments in Bash
      arguments:
          "--text":
              help: "Adds text message"
              required: True
          "--test-boolean":
              help: "Example of a boolean flag"
              action: store_true # or store_false
      steps:
        - |
          #!bash
          echo " ==> In Bash"
          echo " Text: ${ARG_TEXT}"
          echo " Boolean test: ${ARG_TEST_BOOLEAN}"
        - |
          #!python
          print(' ==> In Python')
          print(' Text: %s ' % ctx.args['text'])
          print(' Text: %s ' % str(ctx.args['test_boolean']))
          return True

  # run with "rkd :examples:list-standardlib-modules"
  :examples:list-standardlib-modules:
      description: List all modules in the standardlib
      steps:
        - |
          #!python
          ctx: ExecutionContext
          this: TaskInterface

          import os

          print('Hello world')
          print(os)
          print(ctx)
          print(this)

          return True

Example 2:

version: org.riotkit.rkd/yaml/v1

environment:
    GLOBALLY_DEFINED: "16 May 1966, seamen across the UK walked out on a nationwide strike for the first time in half a century. Holding solid for seven weeks, they won a reduction in working hours from 56 to 48 per week "

env_files:
    - env/global.env

tasks:
    :hello:
        description: |
            #1 line: 11 June 1888 Bartolomeo Vanzetti, Italian-American anarchist who was framed & executed alongside Nicola Sacco, was born.
            #2 line: This is his short autobiography:
            #3 line: https://libcom.org/library/story-proletarian-life

        environment:
            INLINE_PER_TASK: "17 May 1972 10,000 schoolchildren in the UK walked out on strike in protest against corporal punishment. Within two years, London state schools banned corporal punishment. The rest of the country followed in 1987."
        env_files: ['env/per-task.env']
        steps: |
            echo " >> ENVIRONMENT VARIABLES DEMO"
            echo "Inline defined in this task: ${INLINE_PER_TASK}\n\n"
            echo "Inline defined globally: ${GLOBALLY_DEFINED}\n\n"
            echo "Included globally - global.env: ${TEXT_FROM_GLOBAL_ENV}\n\n"
            echo "Included in task - per-task.env: ${TEXT_PER_TASK_FROM_FILE}\n\n"

Explanation of examples:

“arguments” is an optional dict of arguments, key is the argument name, subkeys are passed directly to argparse
“steps” is a mandatory list or text with step definition in Bash or Python language
“description” is an optional text field that puts a description visible in “:tasks” task
“environment” is a dict of environment variables that can be defined
“env_files” is a list of paths to .env files that should be included
“imports” imports a Python package that contains tasks to be used in the makefile and in shell usage

Developing a Python package¶

Each task should implement methods of rkd.api.contract.TaskInterface interface, that’s the basic rule.

Following example task could be imported with path rkd.standardlib.ShellCommandTask, in your own task you would have a different package name instead of rkd.standardlib.

Example task from RKD standardlib:

class ShellCommandTask(TaskInterface):
    """Executes shell scripts"""

    def get_name(self) -> str:
        return ':sh'

    def get_group_name(self) -> str:
        return ''

    def configure_argparse(self, parser: ArgumentParser):
        parser.add_argument('--cmd', '-c', help='Shell command', required=True)

    def execute(self, context: ExecutionContext) -> bool:
        # self.sh() and self.io() are part of TaskUtilities via TaskInterface

        try:
            self.sh(context.get_arg('cmd'), capture=False)
        except CalledProcessError as e:
            self.io().error_msg(str(e))
            return False

        return True

Explanation of example:

The docstring in Python class is what will be shown in :tasks as description. You can also define your description by implementing def get_description() -> str
Name and group name defines a full name eg. :your-project:build
def configure_argparse() allows to inject arguments, and –help description for a task - it’s a standard Python’s argparse object to use
def execute() provides a context of execution, please read Tasks API chapter about it. In short words you can get commandline arguments, environment variables there.
self.io() is providing input-output interaction, please use it instead of print, please read Tasks API chapter about it.

Please check Tasks API for interfaces description¶

Global environment variables¶

Global switches designed to customize RKD per project. Put environment variables into your .env file, so you will no have to prepend them in the commandline every time.

RKD_WHITELIST_GROUPS¶

Allows to show only selected groups in the “:tasks” list. All tasks from hidden groups are still callable.

Examples:

RKD_WHITELIST_GROUPS=:rkd, rkd :tasks
RKD_WHITELIST_GROUPS=:rkd rkd :tasks

RKD_ALIAS_GROUPS¶

Alias group names, so it can be shorter, or even group names could be not typed at all.

Notice: :tasks will rename a group with a first defined alias for this group

Examples:

RKD_ALIAS_GROUPS=":rkd->:r" rkd :tasks :r:create-structure
RKD_ALIAS_GROUPS=":rkd->" rkd :tasks :create-structure

RKD_UI¶

Allows to toggle (true/false) the UI - messages like “Executing task X” or “Task finished”, leaving only tasks stdout, stderr and logs.

RKD_AUDIT_SESSION_LOG¶

Logs output of each executed task, when set to “true”.

Example structure of logs:

# ls .rkd/logs/2020-06-11/11\:06\:02.068556/
task-1-init.log  task-2-harbor_service_list.log

RKD_BIN¶

Defines a command that invokes RKD eg. rkd. When a custom distribution is present, then this value can different. For example project RiotKit Harbor has it’s own command harbor, which is based on RKD, so the RKD_BIN=harbor would be defined in such project.

RKD_BIN is automatically generated, when executing task in a separate process, but it can be also set globally.

Custom distribution¶

RiotKit Do can be used as a transparent framework for writing tasks for various usage, especially for specialized usage. To simplify usage for end-user RKD allows to create a custom distribution.

Custom distribution allows to:

Define custom ‘binary’ name eg. “harbor” instead of “rkd”
Hide unnecessary tasks in custom ‘binary’ (filter by groups - whitelist)
Make shortcuts to tasks: Skip writing group name, make a group name to be appended by default

Example¶

import os
from rkd import main as rkd_main

def env_or_default(env_name: str, default: str):
    return os.environ[env_name] if env_name in os.environ else default

def main():
    os.environ['RKD_WHITELIST_GROUPS'] = env_or_default('RKD_WHITELIST_GROUPS', ':env,:harbor,')
    os.environ['RKD_ALIAS_GROUPS'] = env_or_default('RKD_ALIAS_GROUPS', '->:harbor')
    os.environ['RKD_UI'] = env_or_default('RKD_UI', 'false')
    rkd_main()


if __name__ == '__main__':
    main()

$ harbor :tasks
[global]
:sh                                               # Executes shell scripts
:exec                                             # Spawns a shell process
:init                                             # :init task is executing ALWAYS. That's a technical, core task.
:tasks                                            # Lists all enabled tasks
:version                                          # Shows version of RKD and of all loaded tasks

[harbor]
:compose:ps                                       # List all containers
:start                                            # Create and start containers
:stop                                             # Stop running containers
:remove                                           # Forcibly stop running containers and remove (keeps volumes)
:service:list                                     # Lists all defined containers in YAML files (can be limited by --profile selector)
:service:up                                       # Starts a single service
:service:down                                     # Brings down the service without deleting the container
:service:rm                                       # Stops and removes a container and it's images
:pull                                             # Pull images specified in containers definitions
:restart                                          # Restart running containers
:config:list                                      # Gets environment variable value
:config:enable                                    # Enable a configuration file - YAML
:config:disable                                   # Disable a configuration file - YAML
:prod:gateway:reload                              # Reload gateway, regenerate missing SSL certificates
:prod:gateway:ssl:status                          # Show status of SSL certificates
:prod:gateway:ssl:regenerate                      # Regenerate all certificates with force
:prod:maintenance:on                              # Turn on the maintenance mode
:prod:maintenance:off                             # Turn on the maintenance mode
:git:apps:update                                  # Fetch a git repository from the remote
:git:apps:update-all                              # List GIT repositories
:git:apps:set-permissions                         # Make sure that the application would be able to write to allowed directories (eg. upload directories)
:git:apps:list                                    # List GIT repositories

[env]
:env:get                                          # Gets environment variable value
:env:set                                          # Sets environment variable in the .env file


Use --help to see task environment variables and switches, eg. rkd :sh --help, rkd --help

Notices for above example:

No need to type eg. :harbor:config:list - just :config:list (RKD_ALIAS_GROUPS used)
No “rkd” group is displayed (RKD_WHITELIST_GROUPS used)
There is no information about task name (RKD_UI used)

Tasks API¶

Each task must implement a TaskInterface¶

class rkd.api.contract.TaskInterface¶

configure_argparse(parser: argparse.ArgumentParser)¶: Allows a task to configure ArgumentParser (argparse)

copy_internal_dependencies(task)¶: Allows to execute a task-in-task, by copying dependent services from one task to other task

exec(cmd: str, capture: bool = False, background: bool = False) → Optional[str]¶

Starts a process in shell. Throws exception on error. To capture output set capture=True

NOTICE: Use instead of subprocess. Raw subprocess is less supported and output from raw subprocess: may be not catch properly into the logs

execute(context: rkd.api.contract.ExecutionContext) → bool¶: Executes a task. True/False should be returned as return

format_task_name(name: str) → str¶: Allows to add a fancy formatting to the task name, when the task is displayed eg. on the :tasks list

get_become_as() → str¶: User name in UNIX/Linux system, optional. When defined, then current task will be executed as this user (WARNING: a forked process would be started)

get_declared_envs() → Dict[str, Union[str, rkd.api.contract.ArgumentEnv]]¶: Dictionary of allowed envs to override: KEY -> DEFAULT VALUE

get_full_name()¶: Returns task full name, including group name

get_group_name() → str¶: Group name where the task belongs eg. “:publishing”, can be empty.

get_name() → str¶: Task name eg. “:sh”

io() → rkd.api.inputoutput.IO¶: Gives access to Input/Output object

py(code: str, become: str = None, capture: bool = False, script_path: str = None) → Optional[str]¶

Executes a Python code in a separate process

NOTICE: Use instead of subprocess. Raw subprocess is less supported and output from raw subprocess: may be not catch properly into the logs

rkd(args: list, verbose: bool = False, capture: bool = False) → str¶

Spawns an RKD subprocess

NOTICE: Use instead of subprocess. Raw subprocess is less supported and output from raw subprocess: may be not catch properly into the logs

sh(cmd: str, capture: bool = False, verbose: bool = False, strict: bool = True, env: dict = None) → Optional[str]¶

Executes a shell script in bash. Throws exception on error. To capture output set capture=True

NOTICE: Use instead of subprocess. Raw subprocess is less supported and output from raw subprocess: may be not catch properly into the logs

should_fork() → bool¶: Decides if task should be ran in a separate Python process (be careful with it)

silent_sh(cmd: str, verbose: bool = False, strict: bool = True, env: dict = None) → bool¶

sh() shortcut that catches errors and displays using IO().error_msg()

NOTICE: Use instead of subprocess. Raw subprocess is less supported and output from raw subprocess: may be not catch properly into the logs

static table(header: list, body: list, tablefmt: str = 'simple', floatfmt: str = 'g', numalign: str = 'decimal', stralign: str = 'left', missingval: str = '', showindex: str = 'default', disable_numparse: bool = False, colalign: str = None)¶

Renders a table

Parameters:	header – body – tablefmt – floatfmt – numalign – stralign – missingval – showindex – disable_numparse – colalign –
Returns:	Formatted table as string

To include a task, wrap it in a declaration¶

class rkd.api.syntax.TaskDeclaration(task: rkd.api.contract.TaskInterface, env: Dict[str, str] = {}, args: List[str] = [])¶

To create an alias for task or multiple tasks¶

class rkd.api.syntax.TaskAliasDeclaration(name: str, to_execute: List[str], env: Dict[str, str] = {}, description: str = '')¶: Allows to define a custom task name that triggers other tasks in proper order

Execution context provides parsed shell arguments and environment variables¶

class rkd.api.contract.ExecutionContext(declaration: rkd.api.contract.TaskDeclarationInterface, parent: Optional[rkd.api.contract.GroupDeclarationInterface] = None, args: Dict[str, str] = {}, env: Dict[str, str] = {}, defined_args: Dict[str, dict] = {})¶

Defines which objects could be accessed by Task. It’s a scope of a single task execution.

get_arg(name: str) → Optional[str]¶

Get argument or option

Usage:: ctx.get_arg(‘–name’) # for options ctx.get_arg(‘name’) # for arguments

Raises:	KeyError when argument/option was not defined
Returns:	Actual value or default value

get_arg_or_env(name: str) → Optional[str]¶

Provides value of user input

Usage:

get_arg_or_env(‘–file-path’) resolves into FILE_PATH env variable, and –file-path switch (file_path in argparse)

Behavior:

When user provided explicitly switch eg. –history-id, then it’s value will be taken in priority. If switch –history-id was not used, but user provided HISTORY_ID environment variable, then it will be considered.

If no switch provided and no environment variable provided, but a switch has default value - it would be returned.

If no switch provided and no environment variable provided, the switch does not have default, but environment variable has a default value defined, it would be returned.

When the –switch has default value (user does not use it, or user sets it explicitly to default value), and environment variable SWITCH is defined, then environment variable would be taken.

From RKD 2.1 the environment variable names can be mapped to any ArgParse switch.

Below example maps “COMMAND” environment variable to “–cmd” switch.

Raises:	`MissingInputException` – When no switch and no environment variable was provided, then an exception is thrown.

get_env(name: str, switch: str = '', error_on_not_used: bool = False)¶: Get environment variable value

Interaction with input and output¶

class rkd.api.inputoutput.IO¶

Interacting with input and output - stdout/stderr/stdin, logging

capture_descriptors(target_files: List[str] = None, stream=None, enable_standard_out: bool = True)¶: Capture stdout and stderr from a block of code - use with ‘with’

critical(text)¶: Logger: critical

debug(text)¶: Logger: debug

err(text)¶: Standard error

errln(text)¶: Standard error + newline

error(text)¶: Logger: error

error_msg(text)¶: Error message (optional output)

h1(text)¶: Heading #1 (optional output)

h2(text)¶: Heading #2 (optional output)

h3(text)¶: Heading #3 (optional output)

h4(text)¶: Heading #3 (optional output)

info(text)¶: Logger: info

info_msg(text)¶: Informational message (optional output)

is_silent() → bool¶: Is output silent? In silent mode OPTIONAL MESSAGES are not shown

opt_out(text)¶: Optional output - fancy output skipped in –silent mode

opt_outln(text)¶: Optional output - fancy output skipped in –silent mode + newline

out(text)¶: Standard output

outln(text)¶: Standard output + newline

print_group(text)¶: Prints a colored text inside brackets [text] (optional output)

print_line()¶: Prints a newline

print_opt_line()¶: Prints a newline (optional output)

print_separator()¶: Prints a text separator (optional output)

success_msg(text)¶: Success message (optional output)

warn(text)¶: Logger: warn

Storing temporary files¶

class rkd.api.temp.TempManager(chdir: str = './.rkd/')¶

Manages temporary files inside .rkd directory Using this class you make sure your code is more safe to use on Continuous Integration systems (CI)

Usage:: path = self.temp.assign_temporary_file(mode=0o755)

assign_temporary_file(mode: int = 493) → str¶

Assign a path for writing temporary files in RKD workspace

Note: The RKD is executing the finally_clean_up() at the end of each task

Usage:

try:: path = RKDTemp.assign_temporary_file_path() # (…) some action there
finally:: RKDTemp.finally_clean_up()

finally_clean_up()¶

Used to clean up all temporary files at the end of the code execution

TaskExecutor is running this method after each finished task

Working with YAML files¶

Makefile.yaml has checked syntax before it is parsed by RKD. A jsonschema library was used to validate YAML files against a JSON formatted schema file.

This gives the early validation of typing inside of YAML files, and a clear message to the user about place where the typo is.

YAML parsing API¶

Schema validation is a part of YAML parsing, the preferred way of working with YAML files is to not only parse the schema but also validate. In result of this there is a class that wraps yaml library - rkd.yaml_parser.YamlFileLoader, use it instead of plain yaml library.

Notice: The YAML and schema files are automatically searched in .rkd, .rkd/schema directories, including RKD_PATH

Example usage:

from rkd.yaml_parser import YamlFileLoader

parsed = YamlFileLoader([]).load_from_file('deployment.yml', 'org.riotkit.harbor/deployment/v1')

FAQ¶

FileNotFoundError: Schema “my-schema-name.json” cannot be found, looked in: [‘…/riotkit-harbor’, ‘/…/riotkit-harbor/schema’, ‘/…/riotkit-harbor/.rkd/schema’, ‘/home/…/.rkd/schema’, ‘/usr/lib/rkd/schema’, ‘/usr/lib/python3.8/site-packages/rkd/internal/schema’]

The schema file cannot be found, the name is invalid or file missing. The schema should be placed somewhere in the .rkd/schema directory - in global, in home directory or in project.

rkd.exception.YAMLFileValidationError: YAML schema validation failed at path “tasks” with error: [] is not of type ‘object’

It means you created a list (starts with “-“) instead of dictionary at “tasks” path.

Example what went wrong:

tasks:
    - description: first
    - description: second

Example how it should be as an ‘object’:

tasks:
    first:
        description: first

    second:
        description: second

API¶

class rkd.yaml_parser.YamlFileLoader(paths: List[str])¶

YAML loader extended by schema validation support

YAML schema is stored as JSON files in .rkd/schema directories. The Loader looks in all paths defined in RKD_PATH as well as in paths provided by ApplicationContext

find_path_by_name(filename: str, subdir: str) → str¶: Find schema in one of RKD directories or in current path

load(stream, schema_name: str)¶: Loads a YAML, validates and return parsed as dict/list

load_from_file(filename: str, schema_name: str)¶: Loads a YAML file from given path, a wrapper to load()

Creating installer wizards with RKD¶

Wizard is a component designed to create comfortable installers, where user has to answer a few questions to get the task done.

Concept¶

User answers questions invoked by ask() method calls
At the end the finish() is called, which acts as a commit, saves answers into .rkd/tmp-wizard.json by default and into the .env file (depends on if to_env=true was specified)
Next RKD task executed can read .rkd/tmp-wizard.json looking for answers, the answers placed in .env are already loaded automatically as part of standard mechanism of environment variables support

Example Wizard¶

from rkd.api.inputoutput import Wizard

# self is the TaskInterface instance, in Makefile.yaml it would be "this", in Python code it is "self"
Wizard(self)\
    .ask('Service name', attribute='service_name', regexp='([A-Za-z0-9_]+)', default='redis')\
    .finish()

Service name [([A-Za-z0-9_]+)] [default: redis]:
    -> redis

Example of application that is using Wizard to ask interactive questions

https://github.com/riotkit-org/rkd-coop/raw/master/docs/demo.gif

Using Wizard results internally¶

Wizard is designed to keep the data on the disk, so you can access it in any other task executed, but this is not mandatory. You can skip committing changes to disk by not using finish() which is flushing data to json and to .env files.

Use wizard.answers to see all answers that would be put into json file, and wizard.to_env to browse all environment variables that would be set in .env if finish() would be used.

Example of loading stored values by other task¶

Wizard stores values into file and into .env file, so it can read it from file after it was stored there. This allows you to separate Wizard questions into one RKD task, and the rest of logic/steps into other RKD tasks.

from rkd.api.inputoutput import Wizard

# ... assuming that previously the Wizard was completed by user and the finish() method was called ...

wizard = Wizard(self)
wizard.load_previously_stored_values()

print(wizard.answers, wizard.to_env)

API¶

class rkd.api.inputoutput.Wizard(task: TaskInterface, filename: str = 'tmp-wizard.json')¶

ask(title: str, attribute: str, regexp: str = '', to_env: bool = False, default: str = None, choices: list = [], secret: bool = False) → rkd.api.inputoutput.Wizard¶

Asks user a question

Usage:

wizard = Wizard(self) wizard.ask(‘In which year the Spanish social revolution has begun?’,

attribute=’year’, choices=[‘1936’, ‘1910’])

wizard.finish()

finish() → rkd.api.inputoutput.Wizard¶: Commit all pending changes into json and .env files

input(secret: bool = False)¶: Extracted for unit testing to be possible easier

load_previously_stored_values()¶: Load previously saved values

Good practices¶

Do not use os.getenv()¶

Note: Only in Python code

The ExecutionContext is providing processed environment variables. Variables could be overridden on some levels eg. in makefile.py - rkd.api.syntax.TaskAliasDeclaration can take a dict of environment variables to force override.

Use context.get_env() instead.

Define your environment variables¶

Note: Only in Python code

By using context.get_env() you are enforced to implement a TaskInterface.get_declared_envs() returning a list of all environment variables used in your task code.

All defined environment variables will land in –help, which is considered as a task self-documentation.

Use sh(), exec(), rkd() and silent_sh()¶

Using raw subprocess will make your commands output invisible in logs, as the subprocess is writting directly to stdout/stderr skipping sys.stdout and sys.stderr. The methods provided by RKD are buffering the output and making it possible to save to both file and to console.

Do not print if you do not must, use io()¶

rkd.api.inputoutput.IO provides a standardized way of printing messages. The class itself distinct importance of messages, writing them to proper stdout/stderr and to log files.

print is also captured by IO, but should be used only eventually.

Process isolation and permissions changing with sudo¶

Alternatively called “forking” is a feature of RKD similar to Gradle’s JVM forking - the task can be run in a separate Python’s process. This gives a possibility to run specific task as a specific user (eg. upgrade permissions to ROOT or downgrade to regular user)

Mechanism¶

RKD uses serialization to transfer data between processes - a standard pickle library is used. Pickle has limitations on what can be serialized - any inner-methods and lambdas cannot be returned by task.

To test if your task is compatible with running as a separate process simply add --become=USER-NAME to the commandline of your task. If it will fail due to serialization issue, then you will be notified with a nice stacktrace.

Technically the mechanism works on the task executor level, it means that process isolation is independent of the programming language as whole task’s execute() is ran in a separate process, even if task is declared in YAML and has Bash steps.

Permissions changing with sudo¶

YAML syntax allows to define additional attribute become, that if defined then makes whole task to execute inside a separate Python process ran with sudo.

Additionally the RKD commandline supports a per-task parameter --become

Future usage¶

The mechanism is universal, it can be possibly used to sandbox, or even to execute tasks remotely. Currently we do not support such features but we do not say its impossible in the future.

Docker entrypoints under control¶

RKD has enough small footprint so that it can be used as an entrypoint in docker containers. There are a few features that are making RKD very attractive to use in this role.

Environment variables¶

Defined commandline --my-switch can have optionally overridden value with environment variable. In docker it can help easily adjusting default values.

Task needs to create an explicit declaration of environment variable:

def get_declared_envs(self) -> Dict[str, ArgumentEnv]:
    return {
        'MY_SWITCH': ArgumentEnv(name='MY_SWITCH', switch='--switch-name', default=''),
    }

def execute(self, ctx: ExecutionContext) -> bool:
    # this one will look for a switch value, if switch has default value, then it will look for an environment variable
    ctx.get_arg_or_env('--my-switch')

Arguments propagation¶

When setting ENTRYPOINT ["rkd", ":entrypoint"] everything that will be passed as docker’s CMD will be passed to rkd, so additional tasks and arguments can be appended.

Tasks customization¶

It is a good practice to split your entrypoint into multiple tasks executed one-by-one. This gives you a possibility to create new makefile.yaml/py in any place and modify RKD_PATH environment variable to add additional tasks or replace existing. The RKD_PATH has always higher priority than current .rkd directory.

Possible options:

Create a bind-mount volume with additional .rkd/makefile.yaml, add .rkd/makefile.yaml into container and set RKD_PATH to point to .rkd directory
Create new docker image having original in FROM, add .rkd/makefile.yaml into container and set RKD_PATH to point to .rkd directory

Massive files rendering with JINJA2¶

:j2:directory-to-directory is a specially designed task to render JINJA2 templates recursively preserving a directory structure. You can create for example templates/etc/nginx/nginx.conf.j2 and render ./templates/etc into /etc with all files being copied on the fly.

All jinja2 templates will have access to environment variables - with templating syntax you can define very advanced configuration files

Privileges dropping¶

Often in entrypoint there are cache/uploads permissions corrected, so the root user is used. To migrate the application, to run the webserver the privileges could be dropped.

Solutions:

In YAML syntax each task have a possible field to use: become: user-name-here
In Python class TaskInterface has method get_become_as() that should return empty string or a username to use sudo with
In commandline there is a switch --become=user-name-here that can be used with most of the tasks

Built-in tasks¶

Shell¶

Provides tasks for shell commands execution - mostly used in YAML syntax and in Python modules.

:sh¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib.shell	rkd.standardlib.shell.ShellCommandTask	pip install rkd== SELECT VERSION

Executes a Bash script. Can be multi-line.

Notice: phrase %RKD% is replaced with an rkd binary name

Example of plain usage:

rkd :sh -c "ps aux"
rkd :sh --background -c "some-heavy-task"

Example of task alias usage:

from rkd.api.syntax import TaskAliasDeclaration as Task

#
# Example of Makefile-like syntax
#

IMPORTS = []

TASKS = [
    Task(':find-images', [
        ':sh', '-c', 'find ../../ -name \'*.png\''
    ]),

    Task(':build', [':sh', '-c', ''' set -x;
        cd ../../../
    
        chmod +x setup.py
        ./setup.py build
        
        ls -la
    '''])
]

:exec¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib.shell	rkd.standardlib.shell.ExecProcessCommand	pip install rkd== SELECT VERSION

Works identically as :sh, but for spawns a single process. Does not allow a multi-line script syntax.

Class to import: BaseShellCommandWithArgumentParsingTask¶

Creates a command that executes bash script and provides argument parsing using Python’s argparse. Parsed arguments are registered as ARG_{{argument_name}} eg. –activity-type would be exported as ARG_ACTIVITY_TYPE.

IMPORTS += [
    BaseShellCommandWithArgumentParsingTask(
        name=":protest",
        group=":activism",
        description="Take action!",
        arguments_definition=lambda argparse: (
            argparse.add_argument('--activity-type', '-t', help='Select an activity type')
        ),
        command='''
            echo "Let's act! Let's ${ARG_ACTIVITY_TYPE}!"
        '''
    )
]

Technical/Core¶

:init¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib	rkd.standardlib.InitTask	pip install rkd== SELECT VERSION

This task runs ALWAYS. :init implements a possibility to inherit global settings to other tasks

:tasks¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib	rkd.standardlib.TasksListingTask	pip install rkd== SELECT VERSION

Lists all tasks that are loaded by all chained makefile.py configurations.

Environment variables:

RKD_WHITELIST_GROUPS: (Optional) Comma separated list of groups to only show on the list
RKD_ALIAS_GROUPS: (Optional) Comma separated list of groups aliases eg. “:international-workers-association->:iwa,:anarchist-federation->:fa”

:version¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib	rkd.standardlib.VersionTask	pip install rkd== SELECT VERSION

Shows version of RKD and lists versions of all loaded tasks, even those that are provided not by RiotKit. The version strings are taken from Python modules as RKD strongly rely on Python Packaging.

CallableTask¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib	rkd.standardlib.CallableTask	pip install rkd== SELECT VERSION

This is actually not a task to use directly, it is a template of a task to implement yourself. It’s kind of a shortcut to create a task by defining a simple method as a callback.

import os
from rkd.api.syntax import TaskDeclaration
from rkd.api.contract import ExecutionContext
from rkd.standardlib import CallableTask


def union_method(context: ExecutionContext) -> bool:
    os.system('xdg-open https://iwa-ait.org')
    return True


IMPORTS = [
    TaskDeclaration(CallableTask(':create-union', union_method))
]

TASKS = []

class rkd.standardlib.CallableTask(name: str, callback: Callable[[rkd.api.contract.ExecutionContext, rkd.api.contract.TaskInterface], bool], args_callback: Callable[[argparse.ArgumentParser], None] = None, description: str = '', group: str = '', become: str = '', argparse_options: List[rkd.api.contract.ArgparseArgument] = None)¶

Executes a custom callback - allows to quickly define a short task

configure_argparse(parser: argparse.ArgumentParser)¶: Allows a task to configure ArgumentParser (argparse)

execute(context: rkd.api.contract.ExecutionContext) → bool¶: Executes a task. True/False should be returned as return

get_become_as() → str¶: User name in UNIX/Linux system, optional. When defined, then current task will be executed as this user (WARNING: a forked process would be started)

get_declared_envs() → Dict[str, str]¶: Dictionary of allowed envs to override: KEY -> DEFAULT VALUE

get_group_name() → str¶: Group name where the task belongs eg. “:publishing”, can be empty.

get_name() → str¶: Task name eg. “:sh”

:rkd:create-structure¶

Package to import	Single task to import	PIP package to install	Stable version
rkd.standardlib	rkd.standardlib.CreateStructureTask	pip install rkd== SELECT VERSION

Creates a template structure used by RKD in current directory.

API for developers:

This task is extensible by class inheritance, you can override methods to implement your own task with changed behavior. It was designed to allow to create customized installers for tools based on RKD (custom RKD distributions), the example is RiotKit Harbor.

Look for “interface methods” in class code, those methods are guaranteed to not change from minor version to minor version.

class rkd.standardlib.CreateStructureTask¶

Creates a RKD file structure in current directory

This task is designed to be extended, see methods marked as “interface methods”.

configure_argparse(parser: argparse.ArgumentParser)¶: Allows a task to configure ArgumentParser (argparse)

execute(ctx: rkd.api.contract.ExecutionContext) → bool¶: Executes a task. True/False should be returned as return

get_group_name() → str¶: Group name where the task belongs eg. “:publishing”, can be empty.

get_name() → str¶: Task name eg. “:sh”

get_patterns_to_add_to_gitignore(ctx: rkd.api.contract.ExecutionContext) → list¶

List of patterns to write to .gitignore