[Geany-devel] "Ensure new line on file end" inserts new line when file is empty

[Eugene Arshinov]

On Sat, 16 Oct 2010 22:37:29 +1100%
Lex Trotman <elextr at gmail.com> wrote:

> On 16 October 2010 22:11, Eugene Arshinov <earshinov at gmail.com> wrote:
> > On Sat, 16 Oct 2010 22:04:33 +1100%
> > Lex Trotman <elextr at gmail.com> wrote:
> >
> >> On 16 October 2010 21:39, Eugene Arshinov <earshinov at gmail.com>
> >> wrote:
> >> > On Sat, 16 Oct 2010 21:06:12 +1100%
> >> > Lex Trotman <elextr at gmail.com> wrote:
> >> >
> >> >> On 16 October 2010 19:52, Eugene Arshinov <earshinov at gmail.com>
> >> >> wrote:
> >> >> > On Sat, 16 Oct 2010 17:49:27 +1100%
> >> >> > Lex Trotman <elextr at gmail.com> wrote:
> >> >> >
> >> >> >> Hi Eugene,
> >> >> >>
> >> >> >> I'm not sure it is a bug, the pref setting is there is to
> >> >> >> ensure that the file ends in a newline, so an empty file
> >> >> >> will become just a new line (see od below).  Thats working
> >> >> >> correctly AFAICT.
> >> >> >>
> >> >> >> lex at fred5 ~ $ od untitled
> >> >> >> 0000000 000012
> >> >> >> 0000001
> >> >> >>
> >> >> >> Cheers
> >> >> >> Lex
> >> >> >>
> >> >> >
> >> >> > Hi.
> >> >> >
> >> >> > Yes, of course it's logical that newline is always added, even
> >> >> > for empty files.  But it becomes that for non-empty files you
> >> >> > get one empty line at the end, for empty files -- two empty
> >> >> > lines.  It's not quite useful and (IMHO) bad.
> >> >>
> >> >> Well, if the file ends in a newline it will always appear as an
> >> >> empty line at the end even though there is nothing after the
> >> >> newline, the only way to avoid it is to not end in newline.  If
> >> >> you don't like the blank line, delete the newline ... and turn
> >> >> off the preference so Geany won't add it back again ;-)
> >> >
> >> > No, I like a new line, but not two of them :D
> >> >
> >> >>
> >> >> But there are some things that REQUIRE the file to end in
> >> >> newline, and IIUC that is what the "ensure newline at file end"
> >> >> preference is for, it has to add the newline.  So it isn't a
> >> >> bug if the preference is selected.
> >> >>
> >> >
> >> > We may consider `wc -l' as an example.  It (at least, the
> >> > version I have) accounts the last line only if it ends with \n.
> >> >  With the option turned on, if a user creates an empty file with
> >> > Geany and saves it, wc of course says it contains 1 line.  I
> >> > would prefer the file staying really empty and wc returning 0
> >> > lines.  Note that I have to keep the option turned on to keep wc
> >> > working properly in "usual" cases.
> >> >
> >>
> >> I'm a bit confused, why turn on the preference in that case?
> >
> > If I turn it off and write some text, for example
> >
> >  int main(int argc, char **argv){
> >    return 0;
> >  }
> >
> > , and do not manually insert a newline, wc -l will return 2 which (I
> > believe) would be incorrect as there are 3 lines of code here.
> 
> Ok, so wc counts newlines, not lines, I can't say I've looked that
> closely at it (or its documentation :-).  But this argues that if the
> option is set then Geany is correct in adding a newline, then it would
> count right. But that is just one tool.
> 
> I think we are, as we say here, getting our wires a bit crossed.
> 
> What I'm expressing badly is that if a file ends in a newline you can
> consider that there is nothing after that, or you can consider that it
> has an empty unterminated line after the newline. Similarly an empty
> file has no lines or it has one unterminated empty line.  Geany takes
> the empty unterminated line view, since a new "untitled" shows one
> line on screen.  But neither interpretation is "right", it depends on
> what you are using the file for and what the tools expect.
> 

This is absolutely correct.

> My argument is that, if someone (eg me) selects that I want my file to
> end in newline (which is what the preference says) then thats what it
> should do, irrespective of any interpretation or the length of the
> file.
>

I understand what you mean and agree that this point of view is
meaningful.

My argument for unterminated lines view is that
1) Geany already uses this view in one aspect, as you wrote (and I
agree)
2) and so, to my knowledge, do most text utilities like wc (they do not
process last line if it is unterminated)

From this point of view, it's logical that Geany should never insert
unterminated empty lines, either if file is empty or when it already
contains trailing newline.

By the way, there is another argument for your point of view and
current behaviour.  The code of editor_ensure_final_newline() is quite
old (it was inserted to utils.c in "Initial import" commit, rev 4) and
may be already considered a feature :-)

> >
> >> What I
> >> don't see is your patch checking the preference?
> >
> > The function is called only when the preference is turned on.  See
> > document_save_file() where editor_ensure_final_newline() is called.
> 
> 
> Oh Ok fine, I'm not at my development machine with all Geany source.

No problem.

Best regards,
Eugene.