Re: [Nano-devel] undoing replacings in a marked region goes wrong

[Mark Majeres]

On 3/29/15, Benno Schulenberg <address@hidden> wrote:
>
> On Sat, Mar 28, 2015, at 22:24, Mark Majeres wrote:
>> I looked into this a little.  It's like there are two completely
>> different scenarios, which do not co-exist.  The case when the entire
>> file is being looked at, and the case when only a selection is.  A
>> selection is first partitioned out, so the operation is working only
>> on a portion of the file.  The undo's being captured are relative to
>> this partitioned file.
>
> Yeah, it's as if you're working in a new, separate buffer.  Bummer.
>
>> I think the approach would work best if the
>> selection is not partioned, but just recognized as the start and
>> ending locations of the do_replace_loop().  In summary,
>> do_replace_loop() needs to be reworked so that it will function the
>> same for selections and non-selections.
>
> Oof...  But... would that mean that, when replacing things in a selection,
> the surrounding text does not disappear from the screen?  But if you then
> want to highlight the thing to be replaced, it means the highlighting of
> the selected region will have to be temporarily turned off.  Which would
> make me panic: "It's going to replace in all of the file!"  The replacement
> prompt ("Replace this instance?") will have to clearly say that it is about
> a selection: "Replace this instance in the selected region?".
>
> How much work do you think it would be to make the change, from
> partitioning to not partitioning?
>


Please try this patch.

--Mark

replace.nopartions.patch
Description: Text Data