[WIP] Store an optional value on doc deletions.

[hackergrrl]

No description provided.

[gmaclennan]

I feel pretty good about this approach. Any blockers on moving ahead with this?

[gmaclennan]

I don't quite understand this change, can you explain the implications / reasons?

[hackergrrl]

Now that DELETEs can have a v (value) as well as PUTs, checking for k (key) is a more reliable way of differentiating between PUTs and DELETEs. I can wrap this in a helper function to be more explicit -- not a bad idea.

[gmaclennan]

Given this is is performance-critical code I think better use a comment rather than function, which would have a slight performance overhead.

One thing though, a delete could have lat and lon defined, that wouldn't affect here right?

[hackergrrl]

Good call re comment vs function.

One thing though, a delete could have lat and lon defined, that wouldn't affect here right?

Right. Deletions have a d key, and insertions have a k key. Since this new code checks for the presence of a k, even deletions with values would be ignored.

[hackergrrl]

This is ready for another look-over. Here's the gist:

1. Writing deletions: starting at osm.del(), the deletion values are threaded through into _getDocumentDeletionBatchOp(), which generates a new batch op structure to be passed into osm.batch(), which in turn get fed into kv.batch().
2. Exposing values on deletions: look in osm.get() and addDocFromNode() in _collectNodeAndReferers().