On 25.01.2013, at 11:52, Chris Muller <asqueaker at gmail.com> wrote:

>>>> One change in behavior is that the package snapshot is already taken when pressing the save button, not after accepting the version name dialog.
>>> 
>>> Hmmmm....  Ok, so this is to support the new Ignore option.  I don't
>>> think I like this -- because a regular part of _reviewing_ is being
>>> able to make a tweak fix!  If you can't remove a leftover "self halt"
>>> or make any other fixes then what's the point of reviewing it in this
>>> fashion?  I'll have to close the dialog, make the change, and then
>>> re-click the Save button -- which is quite a few more gestures than
>>> what we have today.
>> 
>> Try it: if you revert a change from the save dialog, it will also be ignored. This modifies the snapshot taken earlier before saving. But it will not see changes you make outside the save dialog, yes. In my workflow, when I made changes, I always used to cancel the save dialog and bring it up again. So for me it's the same.
> 
> I sometimes like to leave Save dialogs open for a _long_ time;
> sometimes days, so I can accumulate the notes as I develop them.
> 
> Even when I *don't* do that, I can still accumulate a lot of changes
> between versions and so there is a lot of reviewing to make sure
> everything is ok and to document the why of everything I did.  But
> because I've accumulated a lot of changes, the process of saving can
> take more than just a few minutes, during which often I will get
> interrupted by something external.
> 
> When I come back, the Save dialog is staring at me but there is this
> "doubt" about whether I had changed anything in the image.  Or, maybe
> just during review I see something that needs a spelling correction,
> or add a comment, but not reverted.  Did I do that BEFORE this save
> dialog or was that the LAST save dialog?  I'm forced to access my
> short-term memory to make that judgement and so there is always this
> lingering doubt about whether what I'm saving is congruent with the
> state of my image.  I just can't take that.
> 
> I do like what you're trying to do here in terms of streamlining the
> Review & Save use-case.  How about, upon clicking Save, it simply
> re-snapshots the package, then enumerates the definitions and
> re-ignore the ones which the user said to (I)gnore the first time and
> re-ignore them just before saving?

I have to think about that. 

The advantage of what I proposed is that in the changes list you see exactly what will be committed, even if it does not match exactly what's in your image.

But I see that this would not mash with your working style.

Maybe the best of both worlds would be if any change done while the save dialog is open would be added to the changes list automatically? Not sure how to implement that though, MC is thoroughly snapshot-based, and taking a snapshot is not cheap.

Easier to implement might be to warn about intermediate changes. Upon accepting the save dialog I snapshot the package anyway, to mark the working copy dirty or clean. If this snapshot differs from the earlier snapshot, it could bring up a warning. Or even better, it could update the changes list with the current one. How does that sound?

- Bert -