Which is faster: narrow-to-region or delete-region?

Which is faster: narrow-to-region or delete-region?

[Arthur Miller]

I have a DEFUN in the middle of a temporary buffer; which is faster to use to
eliminate the rest of the code from the calculations for font lock:

(narrow-to-region defun-begin defun-end)

or

(delete-region (point-min) defun-begin) +
(delete-region defun-end (point-max))

I appreciate if I can get A or B answer so I don't have to benchmark, if it even
matters for the speed?

I have tried both and notice no differences. Looking at the code it seems that
narrow-to-region is faster, but I am not familiar with all the C calls the
respective function does, so hard for me to tell. The feel say narrow-to-region.

Re: Which is faster: narrow-to-region or delete-region?

[Eli Zaretskii]

> From: Arthur Miller <arthur.miller@live.com>
> Date: Tue, 21 Sep 2021 22:10:09 +0200
> 
> 
> I have a DEFUN in the middle of a temporary buffer; which is faster to use to
> eliminate the rest of the code from the calculations for font lock:
> 
> (narrow-to-region defun-begin defun-end)
> 
> or
> 
> (delete-region (point-min) defun-begin) +
> (delete-region defun-end (point-max))

The narrow-to-region method must be much faster, since it only sets a
bunch of variables and does little else.  Moreover, it cannot trigger
GC, whereas delete-region method very well can.

> I have tried both and notice no differences.

Maybe in a very small buffer.

Re: Which is faster: narrow-to-region or delete-region?

[Arthur Miller]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Arthur Miller <arthur.miller@live.com>
>> Date: Tue, 21 Sep 2021 22:10:09 +0200
>> 
>> 
>> I have a DEFUN in the middle of a temporary buffer; which is faster to use to
>> eliminate the rest of the code from the calculations for font lock:
>> 
>> (narrow-to-region defun-begin defun-end)
>> 
>> or
>> 
>> (delete-region (point-min) defun-begin) +
>> (delete-region defun-end (point-max))
>
> The narrow-to-region method must be much faster, since it only sets a
> bunch of variables and does little else.

Thought so when saw the code, but wasn't sure.

> bunch of variables and does little else.  Moreover, it cannot trigger
> GC, whereas delete-region method very well can.

That too, wasn't even thinking about that. I am not so familiar how GC works
with code in C.

>> I have tried both and notice no differences.
>
> Maybe in a very small buffer.
I was just tested in help buffers with that patch I made. But it is interactive
so it is hard to see the difference; I haven't benchmarked.

Thank you. That was educative. And helpful.

Re: Which is faster: narrow-to-region or delete-region?

[Eli Zaretskii]

> From: Arthur Miller <arthur.miller@live.com>
> Cc: emacs-devel@gnu.org
> Date: Wed, 22 Sep 2021 09:00:54 +0200
> 
> > bunch of variables and does little else.  Moreover, it cannot trigger
> > GC, whereas delete-region method very well can.
> 
> That too, wasn't even thinking about that. I am not so familiar how GC works
> with code in C.

GC can be triggered in C code by any of the following:

  . calling Lisp
  . calling some primitives, like funcall, directly from C
  . modifying text of some buffer
  . calling a memory-allocation function such as xmalloc

Re: Which is faster: narrow-to-region or delete-region?

[Arthur Miller]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Arthur Miller <arthur.miller@live.com>
>> Cc: emacs-devel@gnu.org
>> Date: Wed, 22 Sep 2021 09:00:54 +0200
>> 
>> > bunch of variables and does little else.  Moreover, it cannot trigger
>> > GC, whereas delete-region method very well can.
>> 
>> That too, wasn't even thinking about that. I am not so familiar how GC works
>> with code in C.
>
> GC can be triggered in C code by any of the following:
>
>   . calling Lisp
>   . calling some primitives, like funcall, directly from C
>   . modifying text of some buffer
>   . calling a memory-allocation function such as xmalloc

Allright, thank you. Noted in my notebook! :).