py2http 0.0.10

Declarative HTTP service entry point.

the_one_where_we_have_python_objects_and_we_want_to_make_an_http_service_out_of_them

py2http

Dispatching Python functions as http services.

Usage

Run a basic HTTP service from a list of functions.

from py2http.service import run_http_service

# Define or import functions
def add(a, b):
    return a + b

def multiply(a, b):
    return a * b

class Divider:
    def __init__(self, dividend):
        self.dividend = divident

    def divide(self, divisor):
        return self.dividend / divisor

divider_from_ten = Divider(10)

# Make a list of functions or instance methods
func_list = [add, multiply, divider_from_ten.divide]

# Create an HTTP server
run_http_service(func_list)

The HTTP server will listen on port 3030 by default.

# Test the server in a separate process
import requests

url = 'http://localhost:3030/add'
add_args = {'a': 20, 'b': 22}
requests.post(url, json=add_args).json()
# should return 42

Configuration

run_http_service and other entry points accept many configuration values to customize the handling of HTTP requests and responses. Configuration documentation is listed in the file config.yaml.

Method transformation

Expose class methods by flattening the init -> method call process.

from py2http import run_http_service
from py2http.decorators import mk_flat

class Adder:
    def __init__(self, a):
        self.a = a

    def add(self, b):
        return self.a + b

add = mk_flat(Adder, Adder.add, func_name='add')

func_list = [add]

run_http_service(func_list)

Input mapping

By default, the server will only accept JSON requests and parse the values from the request body as keyword arguments. You can define custom input mappers to perform extra handling on the JSON body or directly on the HTTP library request object, such as default injection or type mapping.

from numpy import array
import soundfile
from py2http.decorators import handle_json_req

from my_service.data import fetch_user_data

@handle_json_req  # extracts the JSON body and passes it to the input mapper as a dict
def array_input_mapper(input_kwargs):
    return {'arr': array(input_kwargs.get('arr', [])),
            'scalar': input_kwargs.get('scalar', 1)}

@handle_multipart_req  # extracts a multipart body and passes it as a dict
def sound_file_input_mapper(input_kwargs):
    file_input = input_kwargs['file_upload']
    filename = file_input.filename
    file_bytes = BytesIO(file_input.file)
    wf, sr = soundfile.read(file_bytes)
    return dict(input_kwargs, wf=wf, sr=sr, filename=filename)

def custom_header_input_mapper(req):  # takes a raw aiohttp request
    user = req.headers.get('User', '')
    return {'user': user}


def array_multiplier(arr, scalar):
    return arr * scalar

def save_audio(wf, sr, filename, storage_path='/audio'):
    target = os.path.join(storage_path, filename)
    soundfile.write(target, wf, sr)

def get_user_data(user):
    return fetch_user_data(user)

array_multiplier.input_mapper = array_input_mapper
save_audio.input_mapper = sound_file_input_mapper
get_user_data.input_mapper = custom_header_input_mapper

Output mapping

By default, the server will send the return value of the handler function in an HTTP response in JSON format. You can define custom output mappers to perform extra handling or type conversion.

from io import BytesIO
import soundfile

from py2http.decorators import send_json_resp, send_html_resp

from my_service.data import fetch_blog

@send_json_resp  # sends a JSON response
def array_output_mapper(output, input_kwargs):
    return {'array_value': output['array_value'].tolist()}

@send_html_resp  # sends an HTML response
def mk_html_list(output, input_kwargs):
    tag = input_kwargs.get('tag', 'p')
    item_list = output.split('\n')
    result = [f'<{tag}>{item}</{tag}>' for item in item_list]
    return ''.join(result)

def return_wav(output, input_kwargs):
    wf, sr = output
    wf_bytes = BytesIO()
    soundfile.write(wf_bytes, wf, sr)
    binary_result = wf_bytes.read()
    return web.Response(body=binary_result, content_type='application/octet-stream')


def array_multiplier(arr, scalar):
    return arr * scalar

def get_blog_posts(page):
    return fetch_blog(page)

def download_audio(filename):
    with open(filename) as fp:
        wf, sr = soundfile.read(fp)
    return wf, sr

array_multipler.output_mapper = array_output_mapper

get_blog_posts.output_mapper = mk_html_list

download_audio.output_mapper = return_wav

Error handling

TODO

Client generation

Py2http generates an OpenAPI specification for the server before running. You can use this document with any OpenAPI-compatible client tools. To extract the specification, you can generate a server application object before running it.

from aiohttp import web
import json
from py2http import mk_http_service, run_aiohttp_service

def add(a, b):
    return a + b

func_list = [add]

app = mk_http_service(func_list)
openapi_spec = app.openapi_spec

with open('~/openapi.json', 'w') as fp:
    json.dump(openapi_spec, fp)

web.run_app(app, port=3030)