Chain-ify remove

[timhuff]

@ApeChimp
I'm going to need Table.remove tomorrow so I'll probably have a go at getting it to resemble Table.find tonight. My thoughts, for your consideration:

I don't want any mishaps with remove. As a result, I thought that all remove calls could take one of the following forms:

Hash Tables
users.remove().one().hash(id)
users.remove().all()

Hash/Range Tables
posts.remove().one().hash(user_id).range(timestamp)
posts.remove().some().hash(user_id)
posts.remove().all()

Any deviation from this (such as providing a range key on a some call) will result in the function throwing an error.

This function is going to be a bit more involved than find because some and all will have to use a query or scan and a batchWrite operation, which can be a bit of a pain. Luckily, I already have these functions in When-DynamoDB, so I should be able to just port them over pretty quickly.

[mike-zorn]

What's the use case for remove().all() and remove().some()? Normally when I delete a bunch of stuff, I'm doing it as some sort of maintenance thing which isn't a use case I'd like to optimize our api for.

[timhuff]

Yea, my main reason for wanting them would be setting up/tearing down tests. I could also see it coming in handy for a "delete my account" type of functionality.

[mike-zorn]

I'd just create/destroy the database for setup/teardown (normally a lot less work 😉). I would think delete my account would be able to just use remove().one().hash('foo').

[timhuff]

Creating/destroying the database always seemed a bit extreme but it'd probably be a lot cheaper (depending on the size of the table). But yea, aside from testing, let's say that the user has a collection of posts associated with their account and you want to make it such that deleting the user results in the deletion of all the posts. You would need to use remove().some().