Why was `transient-mark-mode' turned off for `delete-selection-mode'?

Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

I came across this item in NEWS:

** Delete Selection mode can now be used without Transient Mark mode.

"Can now be"?  That sounds positive, as if we are adding some new
possibility that is useful.  Are we?

Why would we do this - what is the use case?  Why should
`delete-selection-mode' no longer enable transient-mark-mode'?

Have we had requests from users who turn off `transient-mark-mode',
asking to be able to use `delete-selection-mode' without seeing
the selection?

I've searched a bit for a discussion about this, in emacs-devel and
in the bug list, but I didn't find anything related.  I searched around
the time of the change, which was apparently made on 2013-12-11, with
this commit:
http://bzr.savannah.gnu.org/lh/emacs/trunk/revision/115459

Stefan Monnier2013-12-11 14:42:34
Revision ID: monnier@iro.umontreal.ca-20131211144234-evm5nwt04uoggorp
* lisp/delsel.el (delete-selection-mode): Don't enable transient-mark-mode.
(delete-selection-helper): Make sure yank starts at the top of the
deleted region.
(minibuffer-keyboard-quit): Use region-active-p.

So far, this change doesn't sound like a very good idea to me, but I
would like to know more about it, in particular the rationale for it.  

Delete selection mode has always been about providing similar
delete/replace-the-selection behavior to what is prevalent outside
Emacs. Why would we now divorce this behavior from highlighting of
the selection?

RE: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

> So far, this change doesn't sound like a very good idea to me, but
> I would like to know more about it, in particular the rationale for
> it.
> 
> Delete selection mode has always been about providing similar
> delete/replace-the-selection behavior to what is prevalent outside
> Emacs.  Why would we now divorce this behavior from highlighting of
> the selection?

Any news on this?  Thx.

Re: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Stefan Monnier]

> Why should `delete-selection-mode' no longer enable
> transient-mark-mode'?

The question is the reverse: why should delete-selection-mode enable
transient-mark-mode?


        Stefan

RE: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

> > Why should `delete-selection-mode' no longer enable
> > transient-mark-mode'?
> 
> The question is the reverse: why should delete-selection-mode enable
> transient-mark-mode?

Why is that the question, for a proposed change to Emacs?
Why not justify the change with a good reason?  Is there none?
You are the first to say that it is not the status quo, but a
proposed change, that needs justifying. (Unless it is your change?)

Anyway, to follow your diversion:  Why should it enable d-s mode,
do you suppose?  Why do you suppose it always has?  Why do you
suppose we designed it to do that?

Why do you suppose that that is also the behavior *everywhere*
outside Emacs?  Highlight what's selected, so you can see what
you are deleting or replacing.  Sounds like a good idea, to me.
Not to mention that it is what everyone expects.  What's wrong
with it?

Again:

 Delete selection mode has always been about providing similar
 delete/replace-the-selection behavior to what is prevalent outside
 Emacs. Why would we now divorce this behavior from highlighting of
 the selection?

So how about answering the question, please: Why this change?

Have we had reports from users asking for it?  Do you see some
advantage to it?  (That's an open question; please don't just
answer yes.)  Tell us the advantages and disadvantages, as you
see them.

Re: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Stefan Monnier]

>> > Why should `delete-selection-mode' no longer enable
>> > transient-mark-mode'?
>> The question is the reverse: why should delete-selection-mode enable
>> transient-mark-mode?
> Why is that the question, for a proposed change to Emacs?

Because you're advocating for delete-selection-mode to enable
transient-mark-mode, so it seems quite relevant to ask why you want
delete-selection-mode to enable transient-mark-mode.

E.g. give us a concrete situation where that would be useful.


        Stefan

RE: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

> >> > Why should `delete-selection-mode' no longer enable
> >> > transient-mark-mode'?
> >>
> >> The question is the reverse: why should delete-selection-mode
> >> enable transient-mark-mode?
> >
> > Why is that the question, for a proposed change to Emacs?
> 
> Because you're advocating for delete-selection-mode to enable
> transient-mark-mode, so it seems quite relevant to ask why you want
> delete-selection-mode to enable transient-mark-mode.

You are the one advocating for a change in the longstanding behavior.

Apparently you can give no reason for it.  (?)

> E.g. give us a concrete situation where that would be useful.

I did, several times now: Be able to see clearly what it is
that you are typing over (replacing or deleting).

That's why this behavior was invented in the first place
(outside Emacs): highlighting the selection and letting you
type to replace what you see highlighted.

Hey, it took what, 20 years or so to get Emacs to turn on
transient-mark-mode by default, so users can see the region
which they act on.  Why is that?

The next step should be to turn on delete-selection-mode by
default, but that will likely take another 20 years or so
(10 have already passed).

Why do you want to go backwards from a useful and user-friendly
behavior that has established itself solidly everywhere over
the last 30 years?  Why take away the visual feedback showing
which text delete-selection-mode will replace by typing?

Go ahead: I've given a reason why users should see the region.
You tell us why they should not, please.

Re: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Stefan Monnier]

> You are the one advocating for a change in the longstanding behavior.

No, I did change the code, but that's done.  I didn't advocate for it,
nor do I feel any need to do it.

>> E.g. give us a concrete situation where that would be useful.

> I did, several times now: Be able to see clearly what it is
> that you are typing over (replacing or deleting).

> That's why this behavior was invented in the first place
> (outside Emacs): highlighting the selection and letting you
> type to replace what you see highlighted.

What does this have to do with a concrete situation?
Please: give me a concrete situation, starting from "emacs -Q".
I think if you go through the exercise, you'll understand that your
request is ill-founded.


        Stefan

RE: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

1. It seems that the description of this feature is misleading.

It apparently does not do what I feared from reading its
description: give you delete-selection (type-to-replace)
behavior without the region being active and highlighted.

That was what I understood from the NEWS description:

** Delete Selection mode can now be used without Transient
   Mark mode.

That would not have been an improvement.  Users of
type-to-replace behavior should see what they are replacing
or deleting.

The feature that was implemented is in fact a good one.
It is better described, IMO, like this (or similarly):

  You can now get Delete-Selection mode behavior (type to
  replace the active region) also when Transient Mark mode
  is activated temporarily.

That is, you cannot use delete-selection without the region
being active and highlighted.  You can use delete-selection
even when region activation and highlighting are temporary.

What is important is the behavior, regardless of what modes
happen to be turned on.  This feature brings d-s behavior
to all situations where the region is active.  Good.

I think it would have been good to raise this possibility
for discussion.  But I am glad that this is now available.
I hope the doc will make clear what this amounts to,
starting with the NEWS description.


2. There are now 4 possible behaviors that a user can
choose:

a. d-s mode on  and t-m mode on
b. d-s mode off and t-m mode on
c. d-s mode off and t-m mode off
d. d-s mode on  and t-m mode off
 
This is a good time to again consider the question of
what the default behavior should be.  It is currently (b).
I think it should be (a).

(a) is the behavior most new users are used to.  It is the
behavior I have preferred for Emacs, ever since it was first
available, 20 or so years ago.

(b) is the behavior we have now.

(c) is the behavior that Emacs had until a few years ago.
It was quite a battle to get Emacs to move from (c) to (b).

(d) might make sense for someone who generally likes the
behavior of (c) but occasionally wants the behavior of (a)
temporarily.  It is possible only now, because of the
new feature described in (1).

I think (a) is the best choice for the default behavior,
and (d) is arguably the second best choice.

It is true that some programmers develop code on platforms
that generally have the behavior of (c) (perhaps also with
some mouse-selection highlighting).

But even then many of them interact with programs (e.g.
web browsers) that have the behavior of (a).  And both
programmers on other platforms and non-programmers are
generally used to the behavior of (a).


3. Besides having the limitation of not being able to just
type to replace the region text, the current situation
suffers from treating mouse selection exceptionally.
A mouse selection has the deletion behavior of d-s mode,
but without its type-to-replace behavior.

At the time this mouse-selection feature was introduced
I argued that mouse-selection should be just another form
of region selection; it should not behave differently in
any way. I would still argue that.

Re: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[David Kastrup]

Drew Adams <drew.adams@oracle.com> writes:

> 2. There are now 4 possible behaviors that a user can
> choose:
>
> a. d-s mode on  and t-m mode on
> b. d-s mode off and t-m mode on
> c. d-s mode off and t-m mode off
> d. d-s mode on  and t-m mode off
>  
> This is a good time to again consider the question of
> what the default behavior should be.  It is currently (b).
> I think it should be (a).
>
> (a) is the behavior most new users are used to.  It is the
> behavior I have preferred for Emacs, ever since it was first
> available, 20 or so years ago.
>
> (b) is the behavior we have now.

I think it shouldn't.

> I think (a) is the best choice for the default behavior,
> and (d) is arguably the second best choice.

I find myself trying to figure out a good way to unhighlight a selection
half of the time when I am trying to edit a link in a browser or
something that pops up highlighted for some reason.  It is total
nuisance behavior without a generally available way to turn the
unasked-for selection off.  The other half of the time I inadvertantly
delete material.  Sometimes it can be recovered by C-z (depending on the
application).  Sometimes not.

Your only argument so far has been "others do it".  But that's not, in
itself, a good reason to let Emacs adopt default behavior that gets in
the way of productive work.  "Now how do I undo that damage" _is_ a
distraction from whatever you have been doing.  I do get this in Emacs
when accidentally typing C-w (rather than C-e or something).  But typing
C-w accidentally does not occur as frequently as, well, typing anything
at all.

> But even then many of them interact with programs (e.g.
> web browsers) that have the behavior of (a).

In my case, that interaction involves a lot of swearing.

> 3. Besides having the limitation of not being able to just
> type to replace the region text, the current situation
> suffers from treating mouse selection exceptionally.
> A mouse selection has the deletion behavior of d-s mode,
> but without its type-to-replace behavior.

Are you sure about that?  Like, really sure?

-- 
David Kastrup

RE: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

> > 2. There are now 4 possible behaviors that a user can choose:
> > a. d-s mode on  and t-m mode on
> > b. d-s mode off and t-m mode on
> > c. d-s mode off and t-m mode off
> > d. d-s mode on  and t-m mode off
> >
> > This is a good time to again consider the question of
> > what the default behavior should be.  It is currently (b).
> > I think it should be (a).
> >
> > (a) is the behavior most new users are used to.  It is the
> > behavior I have preferred for Emacs, ever since it was first
> > available, 20 or so years ago.
> > (b) is the behavior we have now.
> 
> I think it shouldn't.

OK.

> > I think (a) is the best choice for the default behavior,
> > and (d) is arguably the second best choice.
> 
> I find myself trying to figure out a good way to unhighlight a
> selection half of the time when I am trying to edit a link in a
> browser or something that pops up highlighted for some reason.
> It is total nuisance behavior without a generally available way
> to turn the unasked-for selection off.  The other half of the time
> I inadvertantly delete material.  Sometimes it can be recovered
> by C-z (depending on the application).  Sometimes not.

Clearly, d-s mode is not for you.  Fortunately, anyone can choose
the behavior (a-d) that s?he prefers.

But I wonder if your surprise and annoyance with such behavior
outside Emacs might have something to do with your not using d-s
mode inside Emacs.

I use d-s mode, and I am not surprised outside Emacs by a highlighted
selection or its deletion if I hit a deletion key or its replacement
if I type text.  And that's a point in favor of turning on d-s mode
by default: it is quite common UI outside Emacs.

> Your only argument so far has been "others do it".  But that's not,
> in itself, a good reason to let Emacs adopt default behavior that
> gets in the way of productive work.  

I agree 100% that how things work outside Emacs should not be the
*only* reason for how they should work inside Emacs.  I have opposed
turning on CUA mode by default, for instance.

Just as you find d-s mode off more natural and less conducive to
mistakes, so I find the opposite.  People are different, and they
have different habits and use cases.  Each of us probably thinks
that what seems natural for us personally makes for a good default
setting, at least in this case.

But the same thing was true for the discussion about t-m mode.
Plenty of people argued strongly about the annoying distraction of
highlighting; the naturalness of an invisible region and the absence
of any concept of "active" region; and so on.  It was a long
discussion, but we finally turned `transient-mark-mode' on.

I admit that I don't have a particular, killer argument in favor of
d-s by default.  Like you in the opposite direction, I can argue that
its particulars are helpful and say why, etc.  But in this case I'm
not sure that will get us far.  Everyone is likely already familiar
with the differences, which can be described either postively or
negatively without adding much light. The same was true for t-m mode.

A "user poll" might be useful in this case.

> "Now how do I undo that damage" _is_ a distraction from whatever
> you have been doing.  I do get this in Emacs when accidentally typing
> C-w (rather than C-e or something).  But typing C-w accidentally does
> not occur as frequently as, well, typing anything at all.

Which is a strong argument against having type-to-replace without
highlighting the text to be replaced/deleted.  (Which was what I
mistakenly thought the new d-s NEWS entry was announcing.)

I don't see accidentally deleting a highlighted region as a problem,
but if you say it happens to you (don't you see the highlighting?)
then, again, I would guess that maybe the problem is at least partly
that you are between two chairs: you are used to Emacs without d-s,
and then you encounter type-to-replace outside Emacs.

[FWIW, I bind `kill-region' & company (e.g. `clipboard-kill-region')
to a command that makes me `y-or-n-p' the killing if the region is
greater than a given size - 2000 chars by default.  Dunno whether
that would be useful for Emacs generally.  But it helps me,
especially since even though the region is highlighted, and the
highlighting is always visible (since point is on screen), the
extent of a large region is not so obvious.  I also show the
region size in the mode line in chars and lines, using face`region':
"2962 ch, 62 l".]

Similarly, I don't see things "popping up highlighted for some
reason" (either outside Emacs or with d-s mode) and having a need
to unhighlight them.  But if I did, it would not be a big deal to
unhighlight them (C-g in Emacs).

Well, OK, maybe I do see that sometimes.  E.g., if I click the URL
in the address field of a browser it typically becomes selected
and highlighted, so that I can quickly replace it.  What do I do
if I want to keep 95% of it?  I click mouse-1 where I want to
change it, or I use a cursor key, and the highlighting disappears.
Poof!  No swearing.  But I am used to the behavior.

Now of course, if for some reason you don't notice that the text
is selected - you don't notice the highlighting, then perhaps
you will be in for some swearing.  But as you say, `C-z' (undo)
usually reverts any deletion or replacement.  Again, a problem
for you and some others, no doubt, but not for yet others.

> > But even then many of them interact with programs (e.g.
> > web browsers) that have the behavior of (a).
> 
> In my case, that interaction involves a lot of swearing.

See above.  My guess is that that is a consequence mainly of
not being used to the behavior, due to not using the same
behavior within Emacs.  Precisely the argument I was making.

Some people are very used to the type-to-replace behavior
outside Emacs, and they no doubt engage in similar swearing
when they step inside Emacs.  Others, like you, though not
used to it, might benefit from the behavior being the same
(or similar) inside and outside Emacs - less swearing.

> > 3. Besides having the limitation of not being able to just
> > type to replace the region text, the current situation
> > suffers from treating mouse selection exceptionally.
> > A mouse selection has the deletion behavior of d-s mode,
> > but without its type-to-replace behavior.
> 
> Are you sure about that?  Like, really sure?

Well, let's see (I'm on MS Windows; YMMV):

1. emacs -Q (any Emacs version you like).

2. In *scratch*, select some text with the mouse.  Take
   your pick how you do this: drag mouse-1 or double-click it,
   or click mouse-1 then mouse-3 elsewhere, etc.

3. Type some text.

For me, the typed text is inserted at point, without any of the
selected (and highlighted) text being replaced.

If I hit a deletion key (e.g. Backspace or Delete) then the
selected text is deleted.  And that's the behavior I described.
Do you see something different?

Re: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[David Kastrup]

Drew Adams <drew.adams@oracle.com> writes:

>> I find myself trying to figure out a good way to unhighlight a
>> selection half of the time when I am trying to edit a link in a
>> browser or something that pops up highlighted for some reason.
>> It is total nuisance behavior without a generally available way
>> to turn the unasked-for selection off.  The other half of the time
>> I inadvertantly delete material.  Sometimes it can be recovered
>> by C-z (depending on the application).  Sometimes not.
>
> Clearly, d-s mode is not for you.  Fortunately, anyone can choose
> the behavior (a-d) that s?he prefers.
>
> But I wonder if your surprise and annoyance with such behavior
> outside Emacs might have something to do with your not using d-s
> mode inside Emacs.

My suprise and annoyance with getting punched in the face in a bar might
also have something to do with me not being used to being punched in the
face at home.

That does not make it a better idea to let me get punched in the face at
home.

Basically you continue arguing that one can get used to getting annoyed.

> I use d-s mode, and I am not surprised outside Emacs by a highlighted
> selection or its deletion if I hit a deletion key or its replacement
> if I type text.  And that's a point in favor of turning on d-s mode
> by default: it is quite common UI outside Emacs.
>
>> Your only argument so far has been "others do it".  But that's not,
>> in itself, a good reason to let Emacs adopt default behavior that
>> gets in the way of productive work.  
>
> I agree 100% that how things work outside Emacs should not be the
> *only* reason for how they should work inside Emacs.  I have opposed
> turning on CUA mode by default, for instance.
>
> Just as you find d-s mode off more natural and less conducive to
> mistakes, so I find the opposite.

And then you continue to explain how to best deal with the mistakes it
causes you to make, and how to anticipate them best.

You don't actually point out anything useful about that behavior.

>> > 3. Besides having the limitation of not being able to just
>> > type to replace the region text, the current situation
>> > suffers from treating mouse selection exceptionally.
>> > A mouse selection has the deletion behavior of d-s mode,
>> > but without its type-to-replace behavior.
>> 
>> Are you sure about that?  Like, really sure?
>
> Well, let's see (I'm on MS Windows; YMMV):
>
> 1. emacs -Q (any Emacs version you like).
>
> 2. In *scratch*, select some text with the mouse.  Take
>    your pick how you do this: drag mouse-1 or double-click it,
>    or click mouse-1 then mouse-3 elsewhere, etc.
>
> 3. Type some text.
>
> For me, the typed text is inserted at point, without any of the
> selected (and highlighted) text being replaced.
>
> If I hit a deletion key (e.g. Backspace or Delete) then the
> selected text is deleted.  And that's the behavior I described.
> Do you see something different?

That's not the entirety of your claim.  The entirety of your claim is
"the current situation suffers from treating mouse selection
exceptionally."

Try marking a region in emacs -Q with _any_ of the mouse, or with
shift-cursor, or with C-SPC and some movement.  Then type DEL.  You'll
see that _any_ way of marking a highlighted region will cause DEL to
delete that region.

Your claim of the mouse selection being treated specially is bogus.

-- 
David Kastrup

RE: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[Drew Adams]

> Try marking a region in emacs -Q with _any_ of the mouse, or with
> shift-cursor, or with C-SPC and some movement.  Then type DEL.
> You'll see that _any_ way of marking a highlighted region will
> cause DEL to delete that region.

Yes, you are right...since Emacs 24.  The mouse is no longer
exceptional in this respect.  I was wrong about that.

Are you glad that deletion-key-deletes-highlighted-selection
behavior was extended beyond the mouse in this way?  Are you
now OK with this behavior in Emacs, in spite of swearing
about it when you encounter it outside Emacs?  Yes?  Good.

So I guess it is just the type-to-replace part of d-s mode
that you still don't like.

Funny how "one can get used to getting annoyed"... and even
come to appreciate or prefer (?!) formerly annoying behavior...

From arguments against t-m mode (annoying highlighting!),
to arguments against deletion keys deleting the highlighted
selection, to arguments against self-insert keys replacing
the selection,...

I guess you are now used to 2 out of 3 (and actually prefer
their behavior?).  In another 20 years or so we can perhaps
finally go for the 3rd as well, and end up with what the
rest of the world will have been using comfortably for the
past 40 years.

Re: Why was `transient-mark-mode' turned off for `delete-selection-mode'?

[David Kastrup]

Drew Adams <drew.adams@oracle.com> writes:

>> Try marking a region in emacs -Q with _any_ of the mouse, or with
>> shift-cursor, or with C-SPC and some movement.  Then type DEL.
>> You'll see that _any_ way of marking a highlighted region will
>> cause DEL to delete that region.
>
> Yes, you are right...since Emacs 24.  The mouse is no longer
> exceptional in this respect.  I was wrong about that.
>
> Are you glad that deletion-key-deletes-highlighted-selection
> behavior was extended beyond the mouse in this way?  Are you
> now OK with this behavior in Emacs, in spite of swearing
> about it when you encounter it outside Emacs?  Yes?  Good.

Can you please stop making up fantasies about me?  First, this behavior
is _only_ happening for typing DEL, _not_ for typing an arbitrary
character, so you are again manipulatively changing the topic.  Second,
even then it gets in the way but since it is for a _single_ key only
(rather than for _any_ key), it occurs rarely enough that I have not
bothered to actively override it yet.  I probably should since I use
double right click for deleting with a mouse and C-w for deleting with
the keyboard anyway.

> So I guess it is just the type-to-replace part of d-s mode
> that you still don't like.
>
> Funny how "one can get used to getting annoyed"... and even
> come to appreciate or prefer (?!) formerly annoying behavior...

I don't think I will bother replying to you any more.  Your way of
twisting words and topics and doling out insinuations is disgusting.

If you cannot argue for your goals in an upright manner respecting the
integrity and opinion of others, you are not going to convince anybody.
You may hope to annoy them into submission.  I consider that despicable.

-- 
David Kastrup