Bors

A highly flexible and extensible service integration framework for scraping the web or consuming APIs.

Usage

1. Create your model based on the data you expect to incorporate.
2. Decide on what you want to do with your data, and add it.
3. Create or use an existing API integration library.
4. Create your root application to tie it all together.

Object Model

We use marshmallow for the underlying object schema definitions. Here's an example model:

from marshmallow import Schema, fields

class NewsItemSchema(Schema):
    """News item"""
    id = f.Str(required=True)
    url = f.Str(required=True)
    title = f.Str(required=True)
    pubDate = f.Str(required=True)
    timestamp = f.Str(required=True)
    feed_id = f.Int(required=True)
    published_date = f.Str(required=True)
    feed_name = f.Str(required=True)
    feed_url = f.Str(required=True)
    feed_enabled = f.Int(required=True)
    feed_description = f.Str(required=True)
    url_field = f.Str(required=True)
    title_field = f.Str(required=True)
    date_field = f.Str(required=True)
    feed_image = f.Str(required=True)

See the marshmallow docs for more information.

Middleware Strategies

Middleware API is implemented in the form of strategies and follows this basic layout:

"""
Simple context display strategy
"""

from bors.app.strategy import IStrategy


class Print(IStrategy):
    """Print Strategy implementation"""
    def bind(self, context):
        """
        Bind the strategy to the middleware pipeline,
        returning the context
        """
        print(f"""PrintStrategy: {context}""")

        # just a pass-through
        return context

The important things to note here: * We're inheriting from IStrategy. * We're implementing a bind method. * The bind method receives, potentially augments, and then returns the context.

API Integration

Request Schema

Because our API is simple, we're going to use this as-is.

from bors.generics.request import RequestSchema

Response Schema

Our API sends us data in the following format:

{
    "data": ...,
    "status": "OK"
}

For this, we'll need to supplement a bit, removing the root fields and returning the data value:

from marshmallow import fields
from bors.generics.request import ResponseSchema


class MyAPIResponseSchema(ResponseSchema):
    """Schema defining how the API will respond"""
    status = fields.Str()
    def get_result(self, data):
        """Return the actual result data"""
        return data.get("data", "")

    class Meta:
        """Add 'data' field"""
        strict = True
        additional = ("data",)

API Class

from bors.api.requestor import Req


class MyAPI(LoggerMixin):
    name = "my_api"
    def __init__(self, context):
        self.create_logger()

        self.request_schema = RequestSchema
        self.result_schema = MyAPIResponseSchema
        self.context = context

        self.req = Req("http://some.api.endpoint/v1", payload, self.log)

        # We don't need to deal directly with requests, so we pass them through
        self.call = self.req.call

    def shutdown(self):
        """Perform last-minute stuff"""
        pass

Here we use the built-in Req class to issue requests to the API, we assign the request_schema and result_schema to classes in our object, and we set the name, context, and call attributes. The results passed through on the API are referencable from within the middleware context under the key my_api.

Pulling it all together

from bors.app.builder import AppBuilder
from bors.app.strategy import Strategy


def main():
    strat = Strategy(Print())
    app = AppBuilder([MyAPI], strat)
    app.run()

if __name__ == "__main__":
    main()

Here, we set as many strategies and API's as we want, then create and run the app.

Architecture

  +------------+
+-+ MIDDLEWARE +------> out
| +------------+
|                       API/WEB
| +------------+
+-+ PREPROCESS +<------ in
  +------------+

At its most basic level, a bors integrator engages with an integration library (API) passing incoming data through a prepocessor to generate and validate incoming objects, then passes that data through middlewares. Outgoing interactions are initiated from within a middleware and passed directly to an API, allowing easily for request/response type behavior in addition to observe and react.

Ingesting Data

      ^
      |
+-----+------+
| MIDDLEWARE |
+-----+------+
      ^
+-----+------+
| PREPROCESS |
+-----+------+
      ^
      |
      +
     API/
     WEB

Ingested data provokes calls along the pipeline.

Outgoing Data

     API/
     WEB
      ^
      |
+-----+------+
| MIDDLEWARE |
+------------+

Enacted events stimulate API or web actions.

Preprocessing

Preprocessing is nothing more than an object-ization of the incoming data. This provides two benefits: 1. Data can be generalized across API interfaces. 2. Data structure can be validated and enforced.

Middlewares

Middlewares allow for a data processing pipeline to pass data through.

  +-+  +-+  +-+
  |M|  |M|  |M|
  |I|  |I|  |I|
  |D|  |D|  |D|
  |D|  |D|  |D|
->+L+->+L+->+L+->
  |E|  |E|  |E|
  |W|  |W|  |W|
  |A|  |A|  |A|
  |R|  |R|  |R|
  |E|  |E|  |E|
  +-+  +-+  +-+

With this model, we gain a lot of flexibility in the behavior of our integration. Middleware is up to the developer to create, and can be any of the following:

• Data post-processing, filtering, aggregation, or augmentation
• External integrations and interfaces
• Stimulate an API/web transaction from external actors or time-based criteria
• Hooks and callbacks