Configuration tool for configuration files


License
MIT
Install
pip install cfgen==0.2.5

Documentation

Build Status PyPI version

cfgen

This tool allows you to generate a config file from a template and source configuration files. Variables in the source configuration files are evaluated every time the tool is executed which allows you to resolve them dynamically using system shell.

Problem

You need to run a build tool build that requires a configuration file build.cfg.

[git]
branch = bug-111

Every time you switch between branches you need to modify this file. Not fun when you do it several times in a day.

Solution

In this particular case you need to create cfgen.metaconfig

current_branch = git rev-parse --abbrev-ref HEAD

build.cfg.template

[git]
branch = {{ current_branch }}

Run $ cfgen build.cfg The tool will evaluate contents of cfgen.metaconfig executing git rev-parse --abbrev-ref HEAD in the system shell and then write build.cfg using build.cfg.template as a template. These files can be commited into git so you need to create them only once.

Installation

$ sudo pip install cfgen

The fist command installs the python module, the seconde one installs cfgen script.

Usage

$ cfgen --help

Features

metaconfig file hierarchy

cfgen looks for metaconfig files starting from the root directory down to the current one. It merges content with last variable defenition winning. For instance if you run cfgen build.cfg from /home/me/git/myproject it will look for cfgen.metaconfig in /home, /home/me, /home/me/git, /home/me/git/myproject in that order. It will loadd and merge them all. While merging, the variables loaded later override the same variables loaded earlier. This allows you to set a global variable with some default value and override it in a subdirectory.

template file hierarchy

Similarly a template file is looked for however only the last one is used.

Jinja2 templates

Template files are Jinja2 templates.

context aware evaluation

The variables defined in metaconfig files can be used as environment variables in the followng defenitions:

current_branch = git rev-parse --abbrev-ref HEAD
binary_name = echo "myapp_${current_branch}.jar"

shell evaluation

Since variables are evaluated with system shell you can use all shell features. Here is an example of setting aws_profile variable from user input:

aws_profile = /bin/bash -e -o pipefail -c 'read -p "Enter AWS profile: " aws_profile && echo $aws_profile'

-o pipefail will make sure cfgen fails if error occurs when using pipes

s3_object_version = /bin/bash -e -o pipefail -c "aws s3api head-object --bucket '${s3_bucket}' --key '${s3_key}' --output=json --query='VersionId' | tr -d '\"' "

caching

You can cache results of evaluation. This is useful when evaluation is slow or requires user input. List variable names (one per line) you want to cache in cfgen.caching. For these variables cfgen executes evaluation on the first run, caches the results in .cfgen.cache and uses the cached result in subsecquent runs.