Adding an index creates two tables

[andreimarinescu]

I'm trying to work with a table with four fields, similar to the example below:

class Logging
    include Dynamoid::Document
    table :name => :api_calls
    range :stamp, :integer
    field :appId, :integer
    field :function
    field :stamp, :integer
    field :version, :integer
    index :appId

When I use the table without the index directive, everything works ok (but running a query against appId uses scanning, even if I create the table with the index in AWS console beforehand).

When running it with the index directive, Dynamoid creates a table called index_api_calls_appIds and the insert fails with ArgumentError: a range key value is required for this table.

What am I doing wrong? I'm running Rails 3.2.13 and aws-sdk 1.11.1 (same problem with 1.8.6) and the latest official rubygem.

[stefanneculai]

Hi,

There are couple of issues with the indexes. I opened myself an issue for
this and there is even a pull request to fix it.

Also, please note that DynamoID does not support secondary indexes.

[stefanneculai]

PR #133 would fix it.

[andreimarinescu]

Thank you Stefan for answering so quickly. I'll pull the PR and build the gem to test it out. I guess I'll leave this open until the PR gets merged?

[stefanneculai]

Yes. Please let me know if the PR works for you. It would be good to get feedback from more people.

[andreimarinescu]

It seems that the issue is persisting for me. Apart from the fact that the table naming seems to be different, Dynamoid still creates a index_tablename_indexname table and then throws an error saying ArgumentError: a range key value is required for this table.

Here's the steps I took:

• I did a clone on your fork and git reset --hard d10acb4.
• gem build Dynamoid.gemspec and installed the gem.
• I used it's version as a dependency in Gemfile (I checked that bundler didn't download any versions of Dynamoid).

Any other ideas?

[stefanneculai]

I will give myself a try and come back with my thoughts.

[stefanneculai]

I looked a bit more over this and I think the issue is exactly the one from #131. As far as I see, you want to use appId as a secondary index. This is not implemented in DynamoID yet, because secondary indexes were introduced in Amazon API v2.0, but DynamoID is using the old client. Indexes that are right now in DynamoID work similar to way explained here: http://www.youtube.com/watch?v=meBjA68DeIU. That's why it creates a secondary table for it.

I started myself to work on a gem that uses the version 2.0 and introduces support for secondary indexes. If you'd interested, you can take a look over https://github.com/stefanneculai/dynamodb-rails. It is almost there, I've used it in a project and works just fine, but I haven't had time so far to add tests to it. If you are interested, I could send you an example of how to set up a model with indexes. Just let me know.

[andreimarinescu]

Thank you, Stefan! I'll be sure to have a look at dynamodb-rails, it's sound like just the thing we need. For the moment I reverted to writing a wrapper for AWS::DynamoDB::ClientV2 as what I'm trying to accomplish is quite straightforward.
Thanks again for your support!