python-terraform is a python module provide a wrapper of terraform
command line tool.
terraform
is a tool made by Hashicorp, please refer to https://terraform.io/
pip install python-terraform
from python_terraform import *
t = Terraform()
return_code, stdout, stderr = t.<cmd_name>(*arguments, **options)
Note: method name same as reserved keyword like import
won't be accepted by python interpreter,
to be able to call the method, you could call cmd_name by adding _cmd
after command name, for example,
import
here could be called by
from python_terraform import *
t = Terraform()
return_code, stdout, stderr = t.import_cmd(*arguments, **options)
or just call cmd method directly
from python_terraform import *
t = Terraform()
return_code, stdout, stderr = t.cmd(<cmd_name>, *arguments, **options)
simply pass the string to arguments of the method, for example,
terraform apply target_dir
--> <instance>.apply('target_dir')
terraform import aws_instance.foo i-abcd1234
--> <instance>.import('aws_instance.foo', 'i-abcd1234')
dash to underscore
remove first dash, and then use underscore to replace dash symbol as option name
ex. -no-color --> no_color
for a simple flag option
use IsFlagged/None
as value for raising/not raising flag, for example,
terraform taint -allow-missing
--> <instance>.taint(allow_missing=IsFlagged)
terraform taint
--> <instance>.taint(allow_missing=None) or <instance>.taint()
terraform apply -no-color
--> <instance>.apply(no_color=IsFlagged)
for a boolean value option
assign True or False, for example,
terraform apply -refresh=true --> <instance>.apply(refresh=True)
if a flag could be used multiple times, assign a list to it's value
terraform apply -target=aws_instance.foo[1] -target=aws_instance.foo[2]
--->
<instance>.apply(target=['aws_instance.foo[1]', 'aws_instance.foo[2]'])
for the "var" flag, assign dictionary to it
terraform apply -var='a=b' -var='c=d'
--> tf.apply(var={'a':'b', 'c':'d'})
if an option with None as value, it won't be used
By default, stdout and stderr are captured and returned. This causes the application to appear to hang. To print terraform output in real time, provide the capture_output
option with any value other than None
. This will cause the output of terraform to be printed to the terminal in real time. The value of stdout
and stderr
below will be None
.
from python_terraform import Terraform
t = Terraform()
return_code, stdout, stderr = t.<cmd_name>(capture_output=False)
In shell:
cd /home/test
terraform apply -var='a=b' -var='c=d' -refresh=false -no-color
In python-terraform:
from python_terraform import *
tf = Terraform(working_dir='/home/test')
tf.apply(no_color=IsFlagged, refresh=False, var={'a':'b', 'c':'d'})
or
from python_terraform import *
tf = Terraform()
tf.apply('/home/test', no_color=IsFlagged, refresh=False, var={'a':'b', 'c':'d'})
or
from python_terraform import *
tf = Terraform(working_dir='/home/test', variables={'a':'b', 'c':'d'})
tf.apply(no_color=IsFlagged, refresh=False)
In shell:
cd /home/test
terraform fmt -diff=true
In python-terraform:
from python_terraform import *
tf = terraform(working_dir='/home/test')
tf.fmt(diff=True)
for apply/plan/destroy command, assign with following default value to make caller easier in python
input=False
, in this case process won't hang because you missing a variableno_color=IsFlagged
, in this case, stdout of result is easier for parsingIMHO, how terraform design boolean options is confusing.
Take input=True
and -no-color
option of apply
command for example,
they're all boolean value but with different option type.
This make api caller don't have a general rule to follow but to do
a exhaustive method implementation which I don't prefer to.
Therefore I end-up with using IsFlagged
or IsNotFlagged
as value of option
like -no-color
and True/False
value reserved for option like refresh=true