samedhi / gaend Goto Github PK

View Code? Open in Web Editor NEW

1.0 2.0 0.0 90 KB

Convert GAE Models into endpoints

License: MIT License

Python 89.54% HTML 10.46%

google-app-engine restful elasticsearch bigquery taskqueue

gaend's Introduction

gaend - Google App Engine Endpoints

Synopsis

This project converts a google.appengine.ext.ndb.Model into a HTTP endpoints. It provides validation, routing, documentation, and CRUD server endpoints. Its purpose is to act as a "DSL" for Google App Engine, allowing you to quickly write CRUD servers with just a Model.

Full Code Example

# app.yaml
runtime: python27
api_version: 1
threadsafe: true

handlers:
- url: /.*
  script: gaend.main.APP

# main.py
from google.appengine.ext import ndb
from gaend.models import GaendFullMixin

class Alien(ndb.Model, GaendFullMixin):
    proboscis_count = ndb.IntegerProperty(default=1)
    kind = ndb.StringProperty(required=True, choices=["Grey", "Martian", "Hutt"])
    name = ndb.StringProperty(lambda s: len(s) >= 1 and len(s) <= 20)
    full_name = ndb.ComputedProperty(lambda self: self.name + " the " + self.kind)

> gcloud deploy app.yaml
...
Deployed service [modeling-agency] to [https://aliens-are-real.appspot.com]
...
> curl -H "Content-Type: application/json" \
  -d '{"kind": "Martian", "name": "Bob"}' \
  https://aliens-are-real.appspot.com/alien

{"key": "fsdafsdaf",
 "kind": "Martian",
 "proboscis_count": 1,
 "name": "Bob",
 "full_name": "Bob the Martian"}

Motivation

Sometimes your data is actually pretty simple.

You want to Create, Read, Update, and Destroy entities using REST endpoints.
You want to do validation every time an entity is written.
You want to have up to date documentation of your endpoints.

Sometimes, the ndb.Model class (Alien in the example) contains enough information to do all of the above bullet points. Why spend time writing boilerplate code that can be inferred? Are we programmers or are we programmers? Automate all the things.

Installation

> pip install gaend

Documentation

Gaend Read the Docs

Test

brew install entr

find ./gaend ./tests -name "*.py" -print | entr -drc python runner.py $GOOGLE_CLOUD_SDK

gaend's People

Contributors

Stargazers

Watchers

gaend's Issues

Anchor Tags don't work

You should write each kind into own elasticsearch index

https://www.elastic.co/guide/en/elasticsearch/guide/current/mapping.html

I should be able to specify key

Each model should have a function that lets me specify the key generated if I so desire.

If too much content on one table row of docstring, content is off of visible page

If you provide 2000 characters worth of choices, you will not be able to see all of them. Needs the rows to increase in height if there is too much content.

For each ndb.Model in `/documentation`, provide example using curl of all restful operations.

I would like a way to limit you to only having classes of one lowercase form

So class Abc and class abc are not permitted. The second one would throw up with an assertion error. This is desirable because I need the lowercase names for the actual endpoints.

Function can be specified that allows you to specify es info

[( INDEX, DOCTYPE, ID, DOC)]

Task queue is based on callback hook

Basically, so every put using a gaend model always results in that model being added to the taskqueue (not just using the endpoints). The only question is if this is also done "within" the transaction? Otherwise, probably not that useful.

Lets override _get_kind() and put in duplicate checks?

override _get_kind() so that every class's kind is the lower() value of its class name. Will that help with detecting duplicates classes like class Hello and class HELLO that would end up with the same restful resource name?

Audit Logs

Probably should also be another project. I would like the ability to keep every previous version of of an entity in the datastore. Out of index, not directly visible, but I can see what the previous value of each entity was.

`/docs` cannot handle computed properties

Make it 'latitude' and 'longitude'

Google App Engine Soft Scan

GAE has no ability to do a scan of the database (this would require a lock). I have a solution that is probabilistically very likely to be able to scan every record. I think perhaps I should include this in another project, or maybe as a module under gaend?

How do we deal with nil/null/undefined in maps?

I think the rule should be that passing in nil always results in nil being written to the datastore. If you want to use the default value then don't include the key.

Or maybe it is the intersection of nil, default, and required? Should it automatically write the default in the event that it is also required and nil is passed in as a named value?

Thinking about it a bit more, I kind of feel that nil just shouldn't be allowed at all as a value to a property. nil should be indicated by the absence of a property, not by the fact that there is property that has been written with nil.

I should be able to specify the elasticsearch schema if desired.

Would be nice to be able to specify in the Model the elasticsearch schema. If no schema is provided elasticsearch will just default to its automatic mappings. However, if I need to it would be nice to actually specify them so that the ES indexes can be rebuilt from scratch at any time.