Re: printf, left-justification ignored in 5.0.8

Re: printf, left-justification ignored in 5.0.8

[Oliver Kiddle]

Roman Neuhauser wrote:
> 
> 5.0.8:
> 
> % printf "x:%-20s:y\n" fubar
> x:               fubar:y

That'll be my fault for not checking thoroughly enough. The - flag is
fine for numeric values but for strings it seems there is separate code
to allow for UTF-8 which looks in the flags array. The patch should
fix this, adds a test and makes the maximum specification length test
include the ' flag.

Oliver

diff --git a/Src/builtin.c b/Src/builtin.c
index 9358e8b..8bfc419 100644
--- a/Src/builtin.c
+++ b/Src/builtin.c
@@ -4461,7 +4461,7 @@ bin_print(char *name, char **args, Options ops, int func)
 			lleft -= chars;
 			ptr += chars;
 		    }
-		    if (width > 0 && flags[2]) width = -width;
+		    if (width > 0 && flags[3]) width = -width;
 		    if (width > 0 && lchars < width)
 		    	count += fprintf(fout, "%*c", width - lchars, ' ');
 		    count += fwrite(b, 1, lbytes, fout);
diff --git a/Test/B03print.ztst b/Test/B03print.ztst
index 48574c2..9360416 100644
--- a/Test/B03print.ztst
+++ b/Test/B03print.ztst
@@ -169,11 +169,15 @@
 0:%n count zeroed on format reuse
 >1
 
-# this may fill spec string with '%0+- #*.*lld\0' - 13 characters
- printf '%1$0+- #-08.5dx\n' 123
+# this may fill spec string with '%0'+- #*.*lld\0' - 14 characters
+ printf '%1$0'"'+- #-08.5dx\n" 123
 0:maximal length format specification
 >+00123  x
 
+ printf "x:%-20s:y\n" fubar
+0:left-justification of string
+>x:fubar               :y
+
  printf '%*smorning\n' -5 good
 0:negative width specified
 >good morning

Re: printf, left-justification ignored in 5.0.8

[Stephane Chazelas]

2015-06-07 22:39:24 +0200, Oliver Kiddle:
> Roman Neuhauser wrote:
> > 
> > 5.0.8:
> > 
> > % printf "x:%-20s:y\n" fubar
> > x:               fubar:y
> 
> That'll be my fault for not checking thoroughly enough. The - flag is
> fine for numeric values but for strings it seems there is separate code
> to allow for UTF-8 which looks in the flags array. The patch should
> fix this, adds a test and makes the maximum specification length test
> include the ' flag.
[...]

BTW, there was a discussion lately on the Austin group mailing
list confirming that the %20s should count bytes, not
characters.

That is in a UTF-8 locale

printf '|%3s|\n' e é €

should print:

|  e|
| é|
|€|

not:

|  e|
|  é|
|  €|

I find the zsh behaviour more useful though, especially
considering that in zsh we can get the POSIX behaviour with:

$ LC_ALL=C printf '|%3s|\n' e é €
|  e|
| é|
|€|

But note that to get proper alignment, rather than number of
characters, the width of the glyphs should be considered
instead anyway:

$ printf '|%3b|\n' e '\ue9' 'e\u301' '\uff45'
|  e|
|  é|
| é|
|  ｅ|

ksh93 has %3Ls for that:

$ printf '|%3Ls|\n' e $'\ue9' $'e\u301' $'\uff45'
|  e|
|  é|
|  é|
| ｅ|

(possibly uses wcswidth()).

-- 
Stephane

Re: printf, left-justification ignored in 5.0.8

[Vincent Lefevre]

On 2015-06-07 22:15:29 +0100, Stephane Chazelas wrote:
> BTW, there was a discussion lately on the Austin group mailing
> list confirming that the %20s should count bytes, not
> characters.
> 
> That is in a UTF-8 locale
> 
> printf '|%3s|\n' e é €
> 
> should print:
> 
> |  e|
> | é|
> |€|
> 
> not:
> 
> |  e|
> |  é|
> |  €|
> 
> I find the zsh behaviour more useful though,

Well, it depends on the context. As I've said in

  http://www.zsh.org/mla/workers/2012/msg00151.html

  Yes, the number is the size in bytes, not in characters. I think
  that the intent is to deal with internal structures (e.g. with
  file formats where some fields have a fixed or limited size, and
  the same syntax can be used in C to avoid buffer overflows).

I don't know if this is the real reason.

> especially considering that in zsh we can get the POSIX behaviour
> with:
> 
> $ LC_ALL=C printf '|%3s|\n' e é €
> |  e|
> | é|
> |€|

However the change of locale may affect other format specifiers,
like %f, and error messages.

> But note that to get proper alignment, rather than number of
> characters, the width of the glyphs should be considered
> instead anyway:
> 
> $ printf '|%3b|\n' e '\ue9' 'e\u301' '\uff45'
> |  e|
> |  é|
> | é|
> |  ｅ|
> 
> ksh93 has %3Ls for that:
> 
> $ printf '|%3Ls|\n' e $'\ue9' $'e\u301' $'\uff45'
> |  e|
> |  é|
> |  é|
> | ｅ|
> 
> (possibly uses wcswidth()).

This could be useful in zsh, and its support would be needed for
"emulate ksh".

-- 
Vincent Lefèvre <vincent@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)

Re: printf, left-justification ignored in 5.0.8

[Stephane Chazelas]

2015-06-12 13:30:50 +0200, Vincent Lefevre:
> On 2015-06-07 22:15:29 +0100, Stephane Chazelas wrote:
> > BTW, there was a discussion lately on the Austin group mailing
> > list confirming that the %20s should count bytes, not
> > characters.
> > 
> > That is in a UTF-8 locale
> > 
> > printf '|%3s|\n' e é €
> > 
> > should print:
> > 
> > |  e|
> > | é|
> > |€|
> > 
> > not:
> > 
> > |  e|
> > |  é|
> > |  €|
> > 
> > I find the zsh behaviour more useful though,
> 
> Well, it depends on the context. As I've said in
> 
>   http://www.zsh.org/mla/workers/2012/msg00151.html
> 
>   Yes, the number is the size in bytes, not in characters. I think
>   that the intent is to deal with internal structures (e.g. with
>   file formats where some fields have a fixed or limited size, and
>   the same syntax can be used in C to avoid buffer overflows).
> 
> I don't know if this is the real reason.

I beleive the real reason was to align on C printf (not wprintf
for instance) (makes little sense to me)

> > especially considering that in zsh we can get the POSIX behaviour
> > with:
> > 
> > $ LC_ALL=C printf '|%3s|\n' e é €
> > |  e|
> > | é|
> > |€|
> 
> However the change of locale may affect other format specifiers,
> like %f, and error messages.

Well, that's a general problem with LC_* handling. Here however
it can be worked around by using several invocations of printf.

[...]
> > ksh93 has %3Ls for that:
> > 
> > $ printf '|%3Ls|\n' e $'\ue9' $'e\u301' $'\uff45'
> > |  e|
> > |  é|
> > |  é|
> > | ｅ|
> > 
> > (possibly uses wcswidth()).
> 
> This could be useful in zsh, and its support would be needed for
> "emulate ksh".
[...]

While I agree that could be useful feature, note that zsh is not
(and will never be I hope) a superset of ksh93. The emulation
mode changes some behaviour that othewise differ from ksh. It
covers mostly ksh88, and zsh has a few features from ksh93 but
that's about it.

-- 
Stephane