README
¶
FN development kit for Python
Purpose of this library to provide simple interface to parse HTTP 1.1 requests represented as string
Following examples are showing how to use API of this library to work with streaming HTTP requests from Fn service.
Handling Hot HTTP Functions
A main loop is supplied that can repeatedly call a user function with a series of HTTP requests.
In order to utilise this, you can write your app.py as follows:
import fdk
from fdk.http import response
def handler(context, data=None, loop=None):
return response.RawResponse(
http_proto_version=context.version,
status_code=200,
headers={},
response_data=data.readall()
)
if __name__ == "__main__":
fdk.handle(handler)
Automatic HTTP input coercions
Decorators are provided that will attempt to coerce input values to Python types. Some attempt is made to coerce return values from these functions also:
import fdk
@fdk.coerce_http_input_to_content_type
def handler(context, data=None, loop=None):
"""
body is a request body, it's type depends on content type
"""
return data
if __name__ == "__main__":
fdk.handle(handler)
Working with async automatic HTTP input coercions
Latest version supports async coroutines as a request body processors:
import asyncio
import fdk
from fdk.http import response
@fdk.coerce_http_input_to_content_type
async def handler(context, data=None, loop=None):
headers = {
"Content-Type": "text/plain",
}
return response.RawResponse(
http_proto_version=context.version,
status_code=200,
headers=headers,
response_data="OK"
)
if __name__ == "__main__":
loop = asyncio.get_event_loop()
fdk.handle(handler, loop=loop)
As you can see app function is no longer callable, because its type: coroutine, so we need to bypass event loop inside
Handling Hot JSON Functions
A main loop is supplied that can repeatedly call a user function with a series of HTTP requests.
In order to utilise this, you can write your app.py as follows:
import fdk
def handler(context, data=None, loop=None):
return data
if __name__ == "__main__":
fdk.handle(handler)
Working with async Hot JSON Functions
Latest version supports async coroutines as a request body processors:
import asyncio
import fdk
async def handler(context, data=None, loop=None):
return data
if __name__ == "__main__":
loop = asyncio.get_event_loop()
fdk.handle(handler, loop=loop)
Applications powered by Fn: Concept
FDK is not only about developing functions, but providing necessary API to build serverless applications that look like nothing but classes with methods powered by Fn.
import requests
from fdk.application import decorators
@decorators.fn_app
class Application(object):
def __init__(self, *args, **kwargs):
pass
@decorators.with_fn(fn_image="denismakogon/os.environ:latest")
def env(self, fn_data=None):
return fn_data
@decorators.with_fn(fn_image="denismakogon/py-traceback-test:0.0.1",
fn_format="http")
def traceback(self, fn_data=None):
return fn_data
@decorators.fn(fn_type="sync")
def square(self, x, y, *args, **kwargs):
return x * y
@decorators.fn(fn_type="sync", dependencies={
"requests_get": requests.get
})
def request(self, *args, **kwargs):
requests_get = kwargs["dependencies"].get("requests_get")
r = requests_get('https://api.github.com/events')
r.raise_for_status()
return r.text
if __name__ == "__main__":
app = Application(config={})
res, err = app.env()
if err:
raise err
print(res)
res, err = app.traceback()
if err:
raise err
print(res)
res, err = app.square(10, 20)
if err:
raise err
print(res)
res, err = app.request()
if err:
raise err
print(res)
In order to identify to which Fn instance code needs to talk set following env var:
export API_URL=http://localhost:8080
with respect to IP address or domain name where Fn lives.
Applications powered by Fn: supply data to a function
At this moment those helper-decorators let developers interact with Fn-powered functions as with regular class methods. In order to pass necessary data into a function developer just needs to do following
if __name__ == "__main__":
app = Application(config={})
app.env(keyone="blah", keytwo="blah", somethingelse=3)
Key-value args will be turned into JSON instance and will be sent to a function as payload body.
Applications powered by Fn: working with function's result
In order to work with result from function you just need to read key-value argument fn_data:
@decorators.with_fn(fn_image="denismakogon/py-traceback-test:0.0.1",
fn_format="http")
def traceback(self, fn_data=None):
return fn_data
Applications powered by Fn: advanced serverless functions
Since release v0.0.3 developer can consume new API to build truly serverless functions without taking care of Docker images, application, etc.
@decorators.fn(fn_type="sync")
def square(self, x, y, *args, **kwargs):
return x * y
@decorators.fn(fn_type="sync", dependencies={
"requests_get": requests.get
})
def request(self, *args, **kwargs):
requests_get = kwargs["dependencies"].get("requests_get")
r = requests_get('https://api.github.com/events')
r.raise_for_status()
return r.text
Each function decorated with @decorator.fn will become truly serverless and distributed.
So, how it works?
* A developer writes function
* FDK (Fn-powered app) creates a recursive Pickle v4.0 with 3rd-party dependencies
* FDK (Fn-powered app) transfers pickled object to a function based on Python3 GPI (general purpose image)
* FDK unpickles function and its 3rd-party dependencies and runs it
* Function sends response back to Fn-powered application function caller
So, each CPU-intensive functions can be sent to Fn with the only load on networking (given example creates 7kB of traffic between app's host and Fn).
Applications powered by Fn: exceptions
Applications powered by Fn are following Go-like errors concept. It gives you full control on errors whether raise them or not.
res, err = app.env()
if err:
raise err
print(res)
Each error is an instance fn FnError that encapsulates certain logic that makes hides HTTP errors and turns them into regular Python-like exceptions.
TODOs
- generic response class
- use fdk.headers.GoLikeHeaders in http
Directories