fauna / faunadb-ruby Goto Github PK

Now that Classes and Worlds are plain old objects we can drop some special cases.

It should be easy to bring Query methods into scope in order to eliminate the need to namespace them.

For instance, I'd like to be able to do this:

class Foo
  include Fauna::Query
  def createDB(dbName)
    create(Ref.new('databases'), quote(name: dbName))
  end
end

That's not possible without some hackery because the query methods are all module methods.

RestClient doesn't really buy us anything and it's underlying use of Net::HTTP is truly terrifying.

Right now the readme asks you to build it yourself.

in exponential form probably.

set.page vs set.eventPage doesn't make much sense. Can we get rid of the intermediate object and just act on the Set itself? Maybe put it in a specific mode?

Similar to field, now that we have generalized constraints.

However, perhaps we should rethink how the gem exposes access to deeply nested fields. It doesn't correspond to the regular Rails/relational way very well.

Also applies to references I suppose.

See https://faunadb.com/documentation#errors

If someone puts values like secret and observer into the client, it might be convenient to have them gettable later. They are accessible in python but not documented. Should we make it official?

• Show how to connect to a local faunadb
• Add a GET example
• Explain indexes a bit. Show how to create one before paginating.

I don't think the query interface is fully fleshed out.

It might make more sense to exclusively support string interpolation rather than expect reified nested objects. Whatever is easier for the user to understand.

Instead of manipulating them directly.

They read like english, not C.

Desired behavior:

append([1,2,3], [4,5,6]) => [1,2,3,4,5,6]
prepend([4,5,6], [1,2,3]) => [1,2,3,4,5,6]

When you are paginating over a set that contains arrays, the pagination helper flattens those arrays too, so it ends up emitting [a1, b1, a2, b2,....] as opposed to [[a1, b1], [a2, b2], ....]

completely unverified at this point. related to #37

We should be wrapping all network calls to Fauna in a retries block like this:

with_retries(:max_tries => 3, :rescue => [Timeout::Error, Errno::ECONNRESET, Fauna::Connection::NetworkError], :max_sleep_seconds => 10) do
    self.client.post params
end

https://github.com/ooyala/retries

This will give us more hope for fixing the concurrency issues under JRuby introduced by Typhoeus.

Degree of magic in the client does not inspire confidence. I think we can do better, in particular around the DDL manipulation and default fields in built-in records.

The first issue is in line 37 of 'fauna.rake'

class_name = ref[0..-8]

If you have a class ref like "classes/thisisatest", the the class_name will be "classes/this". Ought to be

class_name = [8..-1]

If the ref is long enough, when Dir.chdir(class_name) is called on line 40, an additional 'classes' directory also is created.

Also, the Fauna.connection.get(class_name) on line 43 ought to be changed to Fauna.connection.get(ref).

It's overly opinionated. Disable everything by default and only enable linters we find useful.

• Run bundle exec rake rdoc.
• Visit doc/index.html.
• Click on #add to go to doc/Fauna/Query.html#method-i-add.
• The page will just consist of headers: "Basic forms", "public instance methods", "Collections", "public instance methods", "Miscellaneous functions", "public instance methods", etc.

Since FaunaDB API access relies on keys to switch perspectives, it should be easy to temporarily do so with a client. For example:

client = Fauna::Client.new(secret: SERVER_KEY)

user_token = client.query do
  login ref('users/123'), password: 'sekrit'
end

user_client = client.with_secret(user_token[:secret])
user_client.query do
  paginate match(ref('indexes/items_by_user'), ref('users/self'))
end

When creating multiple new resources using Fauna::Resource.new(), the resulting Resource objects share the following attributes' instances between them:

• constraints
• data
• references

Since they are the same instance, edits to one of the attributes will result in the change being applied to all resources created with Fauna::Resource.new().

For example, in the following code:

settings = Fauna::Resource.new('settings', :ref => 'settings/12345678', :email => '[email protected]')
user = Fauna::Resource.new('users', :ref => 'users/87654321')

settings.data['test'] = true

The following statements will all return true:

user.data['test'] == true
user.constraints.object_id == settings.constraints.object_id
user.data.object_id == settings.data.object_id
user.references.object_id == settings.references.object_id

It's not included in the exception message, so it's inconvenient to get to when a raised exception is printed out.

Fauna::Connection::NotFound makes it seem like it's a connection issue, rather than a 404 response.

Fauna::NotFound or Fauna::Resource::NotFound would be better.

It looks like a new publisher key is generated on each boot of a rails app. I've added a publisher_key setting to fauna.yml that can be used in lieu of the email/password credentials. Changes here: https://github.com/wam/fauna-ruby/compare/allow-publisher-key

Makes sense? If so I can open a pull request.

Possible issue: When using a publisher key, no root connection is established in the rails helper. Not sure if it's intended to be used elsewhere.

We need to let you set the valid time on a finder, as well as add CRUD for manipulating individual versions of resources.

And document .page and its friends.

In 1.3.1, the gemspec has two hardcoded references to Evan's filesystem:

s.cert_chain = ["/Users/eweaver/fauna/fauna-ruby/fauna-ruby.pem"]
...
s.signing_key = "/Users/eweaver/cloudburst/configuration/gem_certificates/gem-private_key.pem"

https://github.com/fauna/fauna-ruby/blob/master/fauna.gemspec#L10
https://github.com/fauna/fauna-ruby/blob/master/fauna.gemspec#L22

Bundler is complaining about it:

Errno::ENOENT: No such file or directory - /Users/eweaver/cloudburst/configuration/gem_certificates/gem-private_key.pem
An error occurred while installing fauna (1.3.1), and Bundler cannot continue.
Make sure that `gem install fauna -v '1.3.1'` succeeds before bundling.

Is there a workaround?