Delete old, unused datastores.

[Stebalien]

Having to maintain old, unused, untested code is a pain (they won't work anyways because most of them haven't really been tested).

[magik6k]

List of what is actually used and how (chocked by searching for usages on latest gx version used by go-ipfs):

• callback
  • Used in coalesce datastore tests (which is not used)
  • Used in tiered datastore tests (which is not used)
• coalesce
  • doesn't appear to be used by anything
• elastigo
  • ...
  • not used
• fs
  • not used
  • we have flatfs
  • it's a tiny, simple implementation of persistent datastore which can be used by anyone who wants to use this package
  • it may be buggy
  • we have flatfs
• keytransform
  • used by namespace
• mount
  • not used
• namespace
  • Used by mount, syncmount, in go-ipfs for blackstore, filestore
• panic
  • not used
  • ...
  • I mean, it may be useful to debug some weird issue, though I'd much rather just use conditional breakpoints.
• sync
  • used all over the place in go-ipfs
• syncmount
  • used in go-ipfs
• tiered
  • not used
  • /might/ be useful, though it doesn't implement batching so I'd say we should drop it, and if anyone wants it back, VCS will still have it.
• timecache
  • Same as above

Datastores in basic_ds.go:

• MapDatastore:
  • used in many tests everywhere
• NullDatastore:
  • Used in some tests
• LogDatastore:
  • Usable in go-ipfs (though mdns needs to be disabled due to a bug)

So I'd say we should remove callback, coalesce, elastigo, fs, mount, tiered and timecache, @Stebalien LGTY?

[magik6k]

And probably just rename syncmount to mount

[Stebalien]

LGTM. @whyrusleeping? You should probably sign off on this since we're deleting code.

[whyrusleeping]

@magik6k @Stebalien lets make a tag on the commit the code exists in before we delete it, and note it in an issue somewhere. After that, SGTM

[kevina]

Some of those might be useful as examples, such as fs and may be worth keeping around. Maybe we can move those to the examples directory.

[Stebalien]

I agree that some of them might be useful (e.g., tiered, timecache) but I'd actually like to get rid of the fs one, even if we keep the others. It's insecure/broken by default (it uses the key names as the file names without modifying them).

[kevina]

The idea was to move it to the examples directory as use it as documentation, not to keep around as a useful datastore. However, I am not set on the idea, any other trivial datastore could serve the same purpose.

[Stebalien]

Yeah, I see the value as documentation... I guess we could just put a big warning saying "don't actually do it this way".

[kevina]

I would also move it under examples/fs.