Johan Stenberg (MSFT) johanste

Cadl for code gen

There should be no need for customizations. A "raw" service specification should be enough to generate a client library.
The cost of customizations should be proportional to the amount of customizations we are making. It must not have a steep cost cliff when we want to customize a small aspect of the generated code (a.k.a. "the cost of the first customization should be low").

	```plantuml
	@startuml
	abstract Model
	abstract StructuredInformation
	abstract Entity
	abstract Operation
	Operation : Id
	Operation : status
	abstract Job
	abstract Batch

	import json
	import typing

	import purl

	HeadersType = typing.Union[typing.Sequence[typing.Tuple[str, typing.Any]], typing.Mapping[str, typing.Any], None]
	JsonBodyType = typing.Union[typing.Mapping[str, "JsonBodyType"], str, int, float, bool, None]
	ParamType2 = typing.Union[typing.Sequence[typing.Tuple[str, typing.Any]], typing.Mapping[str, typing.Any]]
	ParamType = typing.Union[typing.Sequence[typing.Tuple[str, typing.Any]], typing.Mapping[str, typing.Any]]

	{
	"swagger": "2.0",
	"info": {
	"title": "Examples of scenarios - #1, more constrained children",
	"version": "1.0"
	},
	"paths": [],
	"definitions": {
	"GenericResource": {
	"type": "object",

	{
	"swagger": "2.0",
	"info": {
	"title": "Show x-ms-paths",
	"version": "1.0"
	},
	"x-ms-paths": {
	/* Note - fake query parameter to make the path unique. Not actually used in the API.
	You can also differentiate using a different path parameter name if the path is parameterized
	*/

	import typing


	class InfoFragment(typing.TypedDict):

	title: str
	version: str


	OptionalInAttribute = typing.TypedDict(

	{
	"swagger": "2.0",
	"info": {
	"title": "Example of polymorphic models, hypothetical document recognition style",
	"version": "1.0"
	},
	"consumes": ["application/json"],
	"produces": ["application/json"],
	"paths": {
	"/analyze": {

	class AiEventMeta(type):

	def __init__(cls, name, bases, nmspc):
	super().__init__(name, bases, nmspc)
	if not hasattr(cls, 'event_type_registry'):
	setattr(cls, 'event_type_registry', {})
	if hasattr(cls, 'EVENT_TYPE_NAME'):
	cls.event_type_registry[cls.EVENT_TYPE_NAME] = cls

	class AiEvent(metaclass=AiEventMeta):

	import carbon

	video_stream = ...

	# Module level function lets you pick a stream, "features" and go
	# (this assumes there is a default configuration for each feature/category/model)
	for event in carbon.recognize(video_stream, features=['faceDetection', 'vehicleDetection']):
	print(event)

	# Slightly more involved scenario with custom configurations and a dedicated session...