The getChanges method returns all the changes that have happened, since
the specified syncToken and the current collection.
This function should return an array, such as the following:
[
'syncToken' => 'The current synctoken',
'added' => [
'new.txt',
],
'modified' => [
'modified.txt',
],
'deleted' => [
'foo.php.bak',
'old.txt'
],
'result_truncated' : true
];
The syncToken property should reflect the current syncToken of the
collection, as reported getSyncToken(). This is needed here too, to
ensure the operation is atomic.
If the syncToken is specified as null, this is an initial sync, and all
members should be reported.
If result is truncated due to server limitation or limit by client,
set result_truncated to true, otherwise set to false or do not add the key.
The modified property is an array of nodenames that have changed since
the last token.
The deleted property is an array with nodenames, that have been deleted
from collection.
The second argument is basically the 'depth' of the report. If it's 1,
you only have to report changes that happened only directly in immediate
descendants. If it's 2, it should also include changes from the nodes
below the child collections. (grandchildren)
The third (optional) argument allows a client to specify how many
results should be returned at most. If the limit is not specified, it
should be treated as infinite.
If the limit (infinite or not) is higher than you're willing to return,
the result should be truncated to fit the limit.
Note that even when the result is truncated, syncToken must be consistent
with the truncated result, not the result before truncation.
(See RFC6578 Section 3.6 for detail)
If the syncToken is expired (due to data cleanup) or unknown, you must
return null.
The limit is 'suggestive'. You are free to ignore it.
TODO: RFC6578 Setion 3.7 says that the server must fail when the server
cannot truncate according to the limit, so it may not be just suggestive.