print -C and terminators

print -C and terminators

[Bart Schaefer]

On Sat, Jan 30, 2021 at 7:00 AM Beverly Pope <countryone77@gmail.com> wrote:
>
> On Jan 30, 2021, at 4:50 AM, Roman Perepelitsa <roman.perepelitsa@gmail.com> wrote:
>
> >  print -rC1 --
> >
> > This command doesn't print anything.
>
> % print -rC1 -- no_such_file(N)
>
> zsh 5.8 (Ubuntu): prints nothing
> zsh 5.7.1 (Mac): prints return

Hm; I get no output, in both cases.  Similarly on Centos (zsh 5.0).

This behavior seems like a bug to me.  Neither -n nor -N were
specified, so "print" should always output at least a newline.
However, it's been around for quite some time and obviously some
people know about it, so I suppose it should just be documented.  The
doc does make a very oblique reference to it under the -N option:
          print -rNC1 -- "$list[@]" is a canonical way to print an
          arbitrary list as null-delimited records.
The implication is that -C prevents the terminating NUL byte from
being printed when $list is empty.

ubuntu% print -rN -- | od -c
0000000  \0
0000001
ubuntu% print -rN -- 1 2 3 | od -c
0000000   1  \0   2  \0   3  \0
0000006
ubuntu% print -rNC1 -- | od -c
0000000
ubuntu% print -rNC1 -- 1 2 3 | od -c
0000000   1  \0   2  \0   3  \0
0000006

The related behavior of -n is a bit weird:

ubuntu% print -rnN -- | od -c
0000000
ubuntu% print -rnN -- 1 2 3 | od -c
0000000   1  \0   2  \0   3
0000005

This, however, is more questionable:

ubuntu% print -raNC1 -- 1 2 3 | od -c
0000000   1      \0   2      \0   3      \0
0000011

This explains the trailing spaces on the test case for globbing that I
posted a few days ago.    I sort of get why -a needs to do that when
the column count is >1, but it shouldn't be inserting trailing spaces
when the column count is 1.

Re: print -C and terminators

[Beverly Pope]

[-- Attachment #1: Type: text/plain, Size: 683 bytes --]



> On Jan 30, 2021, at 12:51 PM, Bart Schaefer <schaefer@brasslantern.com> wrote:
> 
> On Sat, Jan 30, 2021 at 7:00 AM Beverly Pope <countryone77@gmail.com <mailto:countryone77@gmail.com>> wrote:
>> 
>> On Jan 30, 2021, at 4:50 AM, Roman Perepelitsa <roman.perepelitsa@gmail.com <mailto:roman.perepelitsa@gmail.com>> wrote:
>> 
>>> print -rC1 --
>>> 
>>> This command doesn't print anything.
>> 
>> % print -rC1 -- no_such_file(N)
>> 
>> zsh 5.8 (Ubuntu): prints nothing
>> zsh 5.7.1 (Mac): prints return
> 

Sorry!  When I copied/pasted the command, I accidentally copied the return at the end of the line.  I tried it again on Mac and it didn’t print anything/

[-- Attachment #2: Type: text/html, Size: 3490 bytes --]

Re: print -C and terminators

[Roman Perepelitsa]

On Sat, Jan 30, 2021 at 7:52 PM Bart Schaefer <schaefer@brasslantern.com> wrote:
>
> On Sat, Jan 30, 2021 at 7:00 AM Beverly Pope <countryone77@gmail.com> wrote:
> >
> > On Jan 30, 2021, at 4:50 AM, Roman Perepelitsa <roman.perepelitsa@gmail.com> wrote:
> >
> > >  print -rC1 --
> > >
> > > This command doesn't print anything.
> >
> > % print -rC1 -- no_such_file(N)
>
> This behavior seems like a bug to me.  Neither -n nor -N were
> specified, so "print" should always output at least a newline.
> However, it's been around for quite some time and obviously some
> people know about it, so I suppose it should just be documented.

I learned about it from the documentation for print:

  -l  Print the arguments separated by newlines instead of spa-
      ces.  Note: if the list of arguments is empty,  print  -l
      will  still  output  one  empty  line.  To print a possi-
      bly-empty list of arguments one per line, use print  -C1,
      as in `print -rC1 -- "$list[@]"'.

I don't think I ever wrote or saw code where -l is correct and -C1 is wrong.

Roman.

Re: print -C and terminators

[Bart Schaefer]

On Sat, Jan 30, 2021 at 2:47 PM Roman Perepelitsa
<roman.perepelitsa@gmail.com> wrote:
>
> I learned about it from the documentation for print:

The paragraph about the "-l" option seems like an odd place for the
only mention of this, but the zsh doc is like that a lot.

> I don't think I ever wrote or saw code where -l is correct and -C1 is wrong.

With the exception of those doc examples, I don't think I ever saw or
wrote code where the argument to -C was less than two (unless it was a
variable).  Oh, well.

Re: print -C and terminators

[Stephane Chazelas]

2021-01-30 17:08:21 -0800, Bart Schaefer:
> On Sat, Jan 30, 2021 at 2:47 PM Roman Perepelitsa
> <roman.perepelitsa@gmail.com> wrote:
> >
> > I learned about it from the documentation for print:
> 
> The paragraph about the "-l" option seems like an odd place for the
> only mention of this, but the zsh doc is like that a lot.
> 
> > I don't think I ever wrote or saw code where -l is correct and -C1 is wrong.
> 
> With the exception of those doc examples, I don't think I ever saw or
> wrote code where the argument to -C was less than two (unless it was a
> variable).  Oh, well.
[...]

I have always been annoyed at that behaviour of print -rl -
outputting an empty line when not being given any argument (and
print -f '%s\n' -- has the same problem).

Until I saw someone on unix.stackexchange.com use print -rC1 --
which addresses the problem. I was the one suggesting it be more
explicitly documented at
https://www.zsh.org/mla/workers/2019/msg01202.html
Since then, I've replaced all my (broken) usages of print -rl
with print -rC1.

I agree it wouldn't harm to also clarify in the -C documentation
that nothing is output if no argument is given.

This:

~$ print -raC2 a b c | sed -n l
a  b  $
c  $

(
~$ print -rC2 a b c | sed -n l
a  c$
b$
)

could also be improved.

-- 
Stephane

Re: print -C and terminators

[Bart Schaefer]

On Sun, Jan 31, 2021 at 1:36 AM Stephane Chazelas <stephane@chazelas.org> wrote:
>
> ~$ print -raC2 a b c | sed -n l
> a  b  $
> c  $
>
> could also be improved.

Is that just this?

diff --git a/Src/builtin.c b/Src/builtin.c
index 09eb3728c..35a0fb2db 100644
--- a/Src/builtin.c
+++ b/Src/builtin.c
@@ -4822,7 +4822,7 @@ bin_print(char *name, char **args, Options ops, int func)
                {
                    fwrite(args[n], len[n], 1, fout);
                    l = widths[n];
-                   if (n < argc)
+                   if (n < argc && ic < nc - 1)
                        for (; l < sc; l++)
                            fputc(' ', fout);
                }