Alternative scoll/scan implementation

[beowulfenator]

Hi!

I've implemented my own version of batch query for elasticsearch. It is consistent with other db classes. In other words, batch query processing is done with batch() and each() functions. Also, I've written some tests for this functionality.

I would love to see this commit merged into trunk, because then I shall be able to fix issues #48 and #19.

[cebe]

This looks awesome, thanks! Will make sure this goes into the next release.

[beowulfenator]

Should I fix updateAll and deleteAll in this branch as well?

[cebe]

not sure how this is related? you mean to use the scroll api to iterate over the models to update/delete?

[beowulfenator]

Exactly! The problem with updateAll/deleteAll is that it first needs to get all document ids, then perform the action. As far as I know, search API doesn't even let you get all results (you can set size to max_int or some really high value, but that's a bad solution). Even if it did, it would be terribly inefficient for anything over a few hundred records. So I suggest using scroll api in scan mode to get all ids first, and then carry on as usual.

[cebe]

Good idea, afair, setting size to max_int will result in ES to allocate that amount of memory regardsless of the number of results or so. May not be the case anymore but using scroll api sounds like a good idea.

[beowulfenator]

My last commit fixes issues #48 and #39.

[cebe]

parent method key() and current() will not access the private members in this class but the ones of the parent class.
I'd prefer to duplicate the BatchQueryResult from framework here, we may later extract a proper base class from these if it makes sense.

[cebe]

looks good so far except for the comments added. Also needs a CHANGELOG entry and @since 2.0.4 annotations.

[cebe]

👍