Re: printf issue with thousands separator

[Dr . Jürgen Sauermann]

    Hi Christian,
      
      thanks for the info. The problem is, though, that calling
      setlocale()
      in GNU APL is problematic since I cannot possibly know beforehand
      what the locale of the user will be. And I cannot know all
      peculiarities
      of all possible locales. In our case the thousands separator comes
      after every 3 digits, but rumour has it that, for example, China
      has
       a different scheme where the grouping size varies along the
      number.
      
      For these and other reasons (serious portability concerns on
      different
      platforms) I removed support for locales in GNU APL 1.2 (around
      2014).
      
      Since then GNU APL relies on the C standard which says that the
      default
      locale for all C programs is "C" and, as a consequence, GNU APL
      runs
      under the C locale.
      
      Unfortunately the C locale has no thousands separator:
      
      #include <stdio.h>
          #include <locale.h>
          
          int
          main(int argc, char *argv[])
          {
             setlocale (LC_NUMERIC, "C.UTF-8");
          
             printf("%'f\n", 555555.66);
          }
          
        which prints:
      
      555555.660000
      
      I believe that using the C locale is arguably better than e.g.
      reading
      the environment variables related to locales because it makes the
      behaviour of GNU APL more predictable across different platforms.
      
      I find it somewhat inconsistent if a programmer can explicitly
      request
      the output of the thousands separators with the ' flag and that
      glibc
      may then not follow suit because of the locale.
      
      Best Regards,
      Jürgen

On 7/6/22 1:33 AM, Christian Robert wrote:

> Jürgen,
>       
>       
>       #include <stdio.h> 
>       #include <locale.h> 
>       
>       int 
>       main(int argc, char *argv[]) 
>       { 
>          setlocale (LC_NUMERIC, "en_US.utf8"); 
>       
>          printf("%'f\n", 555555.66); 
>       } 
>       
>       print: 
>       
>       555,555.660000 
>       
>       
>       my 2 cents... 
>       
>       Xtian. 
>       
>       
>       On 2022-07-05 12:40, Dr. Jürgen Sauermann wrote: 
> > Hi Martin, 
> >         
> >         I believe I understood your problem now. As far as *⎕FIO* is
> >         concerned it should be fixed 
> >         in *SVN 1579*: 
> >         
> >         *      "%'.2f" ⎕FIO.sprintf 1234567.89** 
> >         **1,234,567.89** 
> >         * 
> >         
> >         There could be a misunderstanding as to how 'Format by example'
> >         works (and that misunderstanding 
> >         might also be caused by my testcases). What the APL2 calls a
> >         left or right "decorator" is not intended 
> >         as an arbitrary string in the printf() fashion, but something
> >         that is directly attached (and possibly 
> >         suppressed) to a number. In most cases the left decorator is
> >         either the minus sign or the overbar and 
> >         the right decorator is not used. The rare exceptions where the
> >         right decorator is used are: 
> >         
> >         1. negative numbers printed in parentheses instead of a minus
> >         sign. Sometimes used in bookkeeping. 
> >         2. output of negative numbers in a different color (typically
> >         red) using VT100 escape sequences as 
> >         left and right decorators. 
> >         
> >         In the examples that you apparently tried there were 2 left
> >         decorators (one being the $ 
> >         and the other being the minus sign) and Format by example simply
> >         does not allow that. 
> >         
> >         The problem with ⎕FIO was actually caused by glibc which
> >         apparently ignores  the ' flag even 
> >         if the locale tells otherwise, On my machine (and in plain C): 
> >         
> >         *int 
> >         main(int argc, char *argv[]) 
> >         { 
> >             printf("%'f\n", 555555.66); 
> >         } 
> >         * 
> >         prints: 
> >         
> >         *555555.660000* 
> >         
> >         I did already pass the ' flag to *sprintf() *in glibc but glibc
> >         then simply ignored it. 
> >         
> >         Best Regards, 
> >         Jürgen 
> >         
> >         
> >         
> >         On 7/5/22 10:08 AM, Martin Michel wrote: 
> > > Hi Jürgen, 
> > >           
> > >           thanks for your quick reply. 
> > >           
> > > > I will look into adding support for
> > > >             the thousands separator (but not supporting the 
> > > >             locales nonsense). That is, the thousands separator will
> > > >             always be comma and not 
> > > >             e.g. full-stop like in some countries and comma in others. 
> > > >           
> > >           That would be great and I fully agree, locales support would
> > >           be too much 
> > >           of a requirement. 
> > >           
> > > > BTW format by example should have done
> > > >             the job (see the APL2 language 
> > > >             reference page 140): 
> > > >             
> > > >             
> > > >                    "5,555.50" ⍕ 1234.56 
> > > >             1,234.56 
> > > >           
> > >           Well, I tried this but it only works for the simpler cases. I
> > >           could 
> > >           manage to format figures with the currency symbol on the right
> > >           side but 
> > >           the format by examples falls short if I want it on the left,
> > >           combined 
> > >           with negative numbers and also rounding (e.g. no decimal point
> > >           in 
> > >           example spec). 
> > >           
> > >           So this works: 
> > >           
> > >                  ' -1,555,555.40 $' ⍕ 123456.789 ¯987654.12 
> > >                 123,456.79 $   -987,654.12 $ 
> > >           
> > >           But I have not found anything which would give me these
> > >           results: 
> > >           
> > >                $ +123,457   $ -987,654 
> > >           
> > >           As far as I have understood IBM APL2 Language Reference this
> > >           is not 
> > >           possible with format by example. I would be happy if you can
> > >           convince me 
> > >           otherwise, then I would indeed not need ⎕FIO. 
> > >           
> > >           Kind regards, 
> > >           Martin 
> > >           
> > >           
> > >           
> > >           
> > >           
> > >         
> >         
> >       
>       
>