[Gsoc Table Header] Deal with Broken Table

Hi Simon,

Now I have merged most code from Table Header branch and start to do
more debug to make all things stable.

One thing need to check with you about some functions for "Deal with
Broken Tables":

I find that there are some important functions are deleted by you on revision:
http://www.abisource.com/viewvc/abiword/trunk/src/text/fmt/xp/fp_TableContainer.cpp?r1=32578&r2=32579

such as:
* UT_sint32 fp_CellContainer::tweakBrokenTable(fp_TableContainer * pBroke)
* UT_sint32 fp_TableContainer::getBrokenTop(void)
* void fp_TableContainer::adjustBrokenTables(void)
* UT_sint32 fp_TableContainer::tweakBrokenTable(fp_TableContainer * pBroke)

Can you share me the reason that you delete the code and the logic,
thanks a lot,

Thanks,
Joe

On Fri, Apr 12, 2013 at 9:35 AM, joe joe <joegsoc2013@gmail.com> wrote:
> Thanks Martin & Simon.
>
> I am looking into the points that you guys point out. Will proposal a
> solution one or two days, thanks
>
> Joe
>
> On Thu, Apr 11, 2013 at 10:41 AM, Simon Larochelle
> <larochelle.simon.1@gmail.com> wrote:
>>
>> Well bug 8707 is not really fixed. There is a big mess hidden behind
>> these two bugs. The basic issue is that we select a table by defining
>> two endpoints: the anchor and the current cursor position. Both of
>> these points need to be valid cursor positions in the piece table
>> (that is positions in the piece table where text could be inserted).
>> Now to select an object (a table, a cell, a frame ...), we should have
>> as endpoints the beginning and the end of the object (for example the
>> fragment PTX_SectionTable and the fragment PTX_EndTable). This does
>> not work because these endpoints are not valid cursor positions. So we
>> use instead the nearest valid cursor positions, which causes a series
>> of little bugs (for example if the table is at the beginning of the
>> header, the endpoint at the beginning of the table is actually before
>> the start of the header in the main section. That's the origin of bug
>> 8984).
>>
>> A fix would be to use our multiple selection framework to select
>> objects (the selection endpoints and the cursor are different in this
>> selection mode, so we could use any point in the piece table as
>> selection endpoints). However, the multiple selection framework is
>> only partially implemented (that's the cause of the bugs seen when
>> applying text formatting commands on a column). Fully implementing
>> this feature will require revising every function in FV_View. So
>> fixing these bugs is a big project.
>>
>>
>> Simon
>>
>> On Wed, Apr 10, 2013 at 10:16 PM, Martin Edmund Sevior
>> <martines@unimelb.edu.au> wrote:
>>>
>>>
>>> Sorry, Instead of:
>>>
>>> "Bug 8984 is still present. I believe it occurs because the special
>>> case code used to handle bugs at the beginning of a document doesn't
>>> work for headers or footers."
>>>
>>> I should have said:
>>>
>>> "Bug 8984 is still present. I believe it occurs because the special
>>> case code used to handle tables at the beginning of a document doesn't
>>> work for headers or footers."
>>>
>>> Cheers
>>>
>>> Martin
>>>
Received on Fri Jul 19 18:48:42 2013