Previously the whole row was invalidated as a quick workaround for what
I would would be a complex task, but the solution is actually relatively
simple and small in code size, reusing much of the current invalidation
code. The cache data for the whole row is still removed, but the
important part of minimising the DOM interaction is in place.
Fix: Incorrect returns from sub-methods if a selector does not contain any results (DataTables/DataTables #431)
Update: `dt-api iterator()` has a fourth parameter to indicate that a return value is expected
The basic issue here was the `iterator()`'s automatic check to see if a
new API instance should be returned or not. If there are no results from
a selector the iterator will just use the original API instance to
return, but this isn't correct as we expect an empty result set in this
case. This is partly due to the fact that the top level methods (`row()`
for example) will return 2D arrays, while the child methods will flatten
the results, this is parhaps an error in the API, but too late to fix
now.
To address this, `iterator()` now has a parameter to tell it if a new
instance should be used or not. I debated about a new method
`iteratorNew()`, but I think this is just as clear (which isn't very
clear unfortunatly, but just needs to be made clear in the
documentation, which I've updated for this change).
I've added tests for the selectors with `deferRender` enabled. The test
suite is still in desperate need of update...
This is something that has been in DataTables since the version first
vresion and was contentious at that point so while this is a minor code
change, for me it is quite a big change! I feel that this is the correct
operation now, although it does result in rows being moved around where
they might not be needed to if partial pages could be used.
Improved last page calculation when page size is changed on last page.
e.g. Suppose there are total 12 records and page size is set to 5 records/page. Go to last page, which should be 3, change page size to 10. Page#2 should be selected and it should show 2 records on it.
* This reverts the last commit which documented the old behaviour.
Thinking again about it, if you want to modify the rendered data use
`$(cell).html()`.
* When `dt-init columns.data` / `dt-init columns.render` resolve to a
function, that function is executed. Previously this execution was in
the global scope, but in order to be able to use an instance as a data
source for the row (something I wanted as part of 1.10) we need to
execute the function with the instance's scope. A simple `call()` does
this.
* I could add a check to see if the `rowData` is a plain object or not,
but that would add significant overhead to the processing, and this
particular part of the code needs to be very fast. Additionally, I
think it is unlikely that any one was dependant on this being executed
in global scope, so the change is made.
* This fixes DataTables/DataTables #368.
