- The issue was effectively duplicate code - for 1.9- compatiblity when
I introduced 1.10 I had the set function run specifically if a DOM
node was given. But later in the 1.10 series when it became possible
to have the data written to an object rather than an array, that code
would itself call the setter, thus leading to two calls when there is
a DOM node present for the row.
- The first is simply to remove the code that was running the setter for
the DOM specific case.
- This fixes DataTables/DataTables #638 and thread 25846. Thanks to
@Jamaur and @strang91 for following up on this.
- The issue was caused by the check for the `bScrollOversize` option
which looks for the container being the same width as its inner
content area. Obviously that happens to be true if there is no
scrollbar. Adding a check for the clientWidth not being 100 as well
resolves this.
- I've also combined the scrollbar width detection with the other
browser detect functions. The two functions were performing very
similar operations and so easily combined
- The browser detect DOM creation will only run once, rather than for
every table initialisation, optimising multi-table loading
- Fixes DataTables/DataTables #633 - Thanks to @nddery and @epitaphmike
for the input
- If d or objectRead were truthful then the row's data object would be
replaced with an empty object. This was one shortcut in code too many!
- Relates to thread 29530 with thanks to idleog.
- This appears to be caused by setting the prototype chain as an object
directly. For example:
var test = function () {};
test.prototype = {};
$.isPlainObject( new test() );
will return `true` with jQuery 2.1.4 (and `false` for 1.11.3).
If an item is added to the prototype object then it will return
`false`, although for some reason that wasn't the case with
DataTables' API prototype.
More investigation required, but for full compatiblity with exisiting
jQuery releases, the correct thing to do here is to build on the
exisiting prototype, which we can do with `$.extend`.
- This is an interesting one, and I suspect a bit of an edge case in the
browsers - what is happening is that the window is going from having a
scrollbar to not having a scrollbar when the scrolling viewport is
collapsed. This causes the browser to show the scrollbars in the
viewport for an instant, even although they aren't actually required.
If the height of the viewport is 53px and you up it to 80 so there are
no scrollbars and reduce it back to 53 the scrollbars are gone! The
browser must be stopping a race condition once the scrollbars are
shown.
- The fix is to use `max-height` for the scrolling container if the
scroll collapse is set and `height` otherwise. The browser native
methods makes the code _much_ easier to work with, more performant and
easier to understand.
column width for these tables. However, that was in error - it is in
fact required still as the smart width calculations cause the browser to
crash (APPCRASH). There is no indication as to why this is, not is it as
simple as just disabling one or two pieces of code to stop the crash.
This leads me to suspect that it is the number of operations being
performed - particularly to do with the cloning of the elements. If this
is the case, then its just never going to work. Also IE6/7 are such a
small market now (thank goodness) this is just a feature that won't be
available in those browsers.
If anyone is interested in working on fixing it, or paying me to do so,
I'm happy to hear from you :-)
Fix: The auto width calculation table, when scrolling is enabled, being inserted into the scrolling container could cause the viewport to scroll, even when the main table doesn't actually need scrolling. This causes the column width calculations to be incorrect by the scrollbar's width. The fix is to have the calculation table `position:absolute` in the scrolling container and height:1px so it has no effect on the vewport's scrolling.
- These two fixes address DataTables/DataTables #612. Kudos to @Delgan
for reporting these issues and supplying test cases - thanks!
Fix: Table could overscroll if `dt-init data` is used to populate the table
This is a fairly significant change to how DataTables operates for the
scrolling width calculations and primarily is due to a simplification
with the use of `clientWidth` to get the scrolling container width.
Rather than getting the offsetWidth and then removing the scrollbar
width if required the `clientWidth` will give us the value directly.
With the scrollbar width already accounted for the scrolling draw code
can be updated to not need to account for it itself.
Finally, when using the `data` property to add data to the table we need
to recalculate the column widths after the first draw, just like with
Ajax loaded data, otherwise the data added won't be taken into account
and we can get an error in the scrolling.
when body's position is changed (left: 375px) _fnBrowserDetect incorrecly calculates offsets. thus header is a bit misaligned. commit changes test object's position to fixed. in this case it's position will
be relative to the viewport not to the body.
When the table was being destroyed and the remove parameter was passed
in as `true` to remove the table from the DOM there was potential for
bound events to not be unbound from the table elements. This was caused
by the use of the `detach` jQuery method which will not remove events.
Furthermore, rows would not be added to the table before this method (or
`remove()`) was called. As such we need to reorder the code slightly -
attach the rows back into the DOM and then remove them (or detach if the
table is left in the DOM) - otherwise rows which were hidden by paging
would not have their events unbound.
Many thanks for KnowledgeNet ( http://kn-it.com/ ) for their support in
finding and allowing this bug to be addressed!
Being able to get and keep a reference to each row is quite important.
You can't use the DataTables row index for this as it can potentially
change, but we can use IDs, based on a unique value for each row in the
data - `dt-init rowId`.
I've made the decision to not require escaping of data in the ID. This
means that selectors can be ID selectors with a cobontation of other
things (#div.class for example) as that really doesn't make much sense
in this contact - you only have rows that you can select from. Not
requiring escaping makes both the client and library code much easier.
This is documented.
The implemention is to store a reference to each row's data object based
on its id allowing for a trivial lookup. The object has its own index
in the `aoData` array, so there is no `indexOf` required. The only
downside is that the index must be updated when a row is deleted. This
is done with a trivial for loop.
* Issue was that an `error` event was being triggered before the
settings object had been configured. There is no option to trigger an
error event in this case.
* Fixes DataTables/DataTables #572
On line 605, column.sTitle is set to the TH's innerHTML.
On line 1816, column.sTitle is compared with cell.html().
If you have child elements with event handlers on them inside the TH, these are not the same in IE8 (see https://gist.github.com/simonbrent/933e552739477f4be3ab#file-test-html)
As a result, the check on 1816 fails, the contents of the cell is replaced, and the event handlers are removed.
Comparing instead with cell[0].innerHTML fixes this issue.
