Reset transient mark mode to previous value when stopping expansion

[mallt]

This PR retains the value of transient-mark-mode before expanding, and resets transient-mark-mode to this value when stopping expansion.

When the value of transient-mark-mode contains 'only, the exchange-point-and-mark function will cons new 'only symbols to the transient-mark-mode value. When stopping expansion, this will leave the transient-mark-mode variable with several 'only symbols, causing the regular set-mark command (C-SPC) to be broken. When the value of transient-mark-mode is reset to the value before expanding, the set-mark command behaves correctly again.

This fixes #220 when shift-select-mode is active and when stopping the expansion with C-g or 0.

When copying the region after expanding however (with M-w f.ex.), the value of transient-mark-mode is not reset and the regular set-mark command is broken again. I guess interfering in all these cases would lead too far, but this PR already fixes the basic case of stopping expansion with C-g or 0.

@magnars Please feel tree to post your comments on this PR. I'm not a user of shift-select-mode myself, but since it seems to be enabled by default, I think it's important expand-region already handles the basic case.

Thanks!

[coveralls]

Coverage increased (+0.08%) to 85.133% when pulling 8c3492e on mallt:reset-transient-mark-mode into d125220 on magnars:master.

[magnars]

Thanks!

[milkypostman]

here is the problem with this change: having shift-select-mode enabled should not assume I want that behavior with expand-region.

Old behavior was that regardless of shift-select-mode, once I select something with expand-region, I can move the cursor and the transient mark stays the same.

With the new behavior I have to either hold shift, or I need disable shift-select-mode, which I had to end up doing.

[mallt]

Ok I see, I'll have a look at it!

[mallt]

To me it seems most intuitive for users of shift-select-mode to be able to further extend a region with shift-translated point motion keys after using expand-region.

If we would ignore shift-select-mode we would expect from the users of this mode to use the regular point motion keys to modify a region marked with expand-region. Wouldn't that be counterintuitive?

[milkypostman]

But transient mark works even with shift-select-mode on. So the question is whether expand region is simulating a shifted selection or a transient mark selection and it has always, as far as I can tell, in the past represented a transient mark based solution. My ask is simply that if this change is made it should be configurable. Because one day I upgraded and behavior changed. Further, it makes more sense to me that the default is a transient mark based selection because nothing about the expand region binding relates to the shift key.

…

On Sun, Oct 15, 2017 at 3:28 PM Tijs Mallaerts ***@***.***> wrote: To me it seems most intuitive for users of shift-select-mode to be able to further extend a region with shift-translated point motion keys after using expand-region. If we would ignore shift-select-mode we would expect from the users of this mode to use the regular point motion keys to modify a region marked with expand-region. Wouldn't that be counterintuitive? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#225 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AACUskj8A8G3KmNnQNzlPuX0Rde98uSmks5ssmrcgaJpZM4MQJO2> .

[mallt]

The support for shift-select-mode was added with this commit based on issue #190.

I'm not a shift-select-mode user myself, but the commit seemed to introduce the expected behavior. I agree with your remarks however, it might make sense to add an extra configuration parameter to expand-region to indicate if shift-select-mode should be respected or not. I think the main problem is however that shift-select-mode is set to true by default although a lot of emacs users expect it to be disabled.