[PATCH] strftime: support %q to output the quarter of year

[PATCH] strftime: support %q to output the quarter of year

[Pádraig Brady]

This is already supported by gnulib.

* manual/time.texi: Document %q outputs quarter 1..4.
* time/strftime_l.c: Implement %q.
---
 manual/time.texi  | 6 ++++++
 time/strftime_l.c | 4 ++++
 2 files changed, 10 insertions(+)

diff --git a/manual/time.texi b/manual/time.texi
index 8a5f94e..f4c5f87 100644
--- a/manual/time.texi
+++ b/manual/time.texi
@@ -1510,6 +1510,12 @@ most locales @samp{AM}/@samp{PM} format is not supported, in such cases
 
 This format is a GNU extension.
 
+@item %q
+Quarter of the year (@samp{1}@dots{}@samp{4}),
+with January starting the first quarter.
+
+This format is a GNU extension.
+
 @item %r
 The complete calendar time using the AM/PM format of the current locale.
 
diff --git a/time/strftime_l.c b/time/strftime_l.c
index 869e0b9..13db490 100644
--- a/time/strftime_l.c
+++ b/time/strftime_l.c
@@ -1108,6 +1108,10 @@ __strftime_internal (s, maxsize, format, tp, tzset_called ut_argument
 	  goto underlying_strftime;
 #endif
 
+        case L_('q'):           /* GNU extension.  */
+          DO_SIGNED_NUMBER (1, tp->tm_mon < -3, tp->tm_mon / 3 + 1U);
+          break;
+
 	case L_('R'):
 	  subfmt = L_("%H:%M");
 	  goto subformat;
-- 
2.5.5

Re: [PATCH] strftime: support %q to output the quarter of year

[Florian Weimer]

* Pádraig Brady:

> This is already supported by gnulib.

This needs a test case (at least a test case update).

> +        case L_('q'):           /* GNU extension.  */
> +          DO_SIGNED_NUMBER (1, tp->tm_mon < -3, tp->tm_mon / 3 + 1U);
> +          break;

I don't think we have DO_SIGNED_NUMBER in glibc.

Re: [PATCH] strftime: support %q to output the quarter of year

[Adhemerval Zanella]

On 03/11/2016 17:58, Florian Weimer wrote:
> * Pádraig Brady:
> 
>> This is already supported by gnulib.
> 
> This needs a test case (at least a test case update).
> 
>> +        case L_('q'):           /* GNU extension.  */
>> +          DO_SIGNED_NUMBER (1, tp->tm_mon < -3, tp->tm_mon / 3 + 1U);
>> +          break;
> 
> I don't think we have DO_SIGNED_NUMBER in glibc.
> 

Indeed and you need to indicate how/where exactly you have tested
it.

Re: [PATCH] strftime: support %q to output the quarter of year

[Joseph Myers]

[-- Attachment #1: Type: text/plain, Size: 229 bytes --]

On Thu, 3 Nov 2016, Florian Weimer wrote:

> * Pádraig Brady:
> 
> > This is already supported by gnulib.
> 
> This needs a test case (at least a test case update).

And a NEWS entry.

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: [PATCH] strftime: support %q to output the quarter of year

[Pádraig Brady]

[-- Attachment #1: Type: text/plain, Size: 442 bytes --]

Updated patch attached.

Tested using:

 sudo alternatives --set ld /usr/bin/ld.bfd
 mkdir build.tmp
 cd build.tmp
 ../configure --quiet --disable-sanity-checks
 make PARALLELMFLAGS=-j$(nproc)
 make -r PARALLELMFLAGS=-j$(nproc) -C $PWD/../time objdir=$(pwd) check
 ./testrun.sh time/tst-strftime && echo ok

Note even though the code is trivial here,
%q is useful as from the shell you need to:
$(( ($(date +%-m)-1)/3+1 ))

thanks,
Pádraig.

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: glibc-strftime-%q.patch --]
[-- Type: text/x-patch; name="glibc-strftime-%q.patch", Size: 2977 bytes --]

From 137cc31d7870e3ecda6c67cda47e29ace2c5acf6 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?P=C3=A1draig=20Brady?= <P@draigBrady.com>
Date: Thu, 3 Nov 2016 19:36:59 +0000
Subject: [PATCH] strftime: support %q to output the quarter of year

This is already supported by gnulib.

* manual/time.texi: Document %q outputs quarter 1..4.
* time/strftime_l.c: Implement %q.
* time/tst-strftime.c: Add a test case.
* NEWS: Mention the new feature.
---
 NEWS                |  2 ++
 manual/time.texi    |  6 ++++++
 time/strftime_l.c   |  4 ++++
 time/tst-strftime.c | 27 +++++++++++++++++++++++++++
 4 files changed, 39 insertions(+)

diff --git a/NEWS b/NEWS
index 65184b1..9b9fa33 100644
--- a/NEWS
+++ b/NEWS
@@ -119,6 +119,8 @@ Version 2.25
   variable for a particular architecture in the GCC source file
   'gcc/config.gcc'.
 
+* strftime now supports the %q directive to output the quarter of the year.
+
 Security related changes:
 
   On ARM EABI (32-bit), generating a backtrace for execution contexts which
diff --git a/manual/time.texi b/manual/time.texi
index 6a899b7..b6466d1 100644
--- a/manual/time.texi
+++ b/manual/time.texi
@@ -1510,6 +1510,12 @@ most locales @samp{AM}/@samp{PM} format is not supported, in such cases
 
 This format is a GNU extension.
 
+@item %q
+Quarter of the year (@samp{1}@dots{}@samp{4}),
+with January starting the first quarter.
+
+This format is a GNU extension.
+
 @item %r
 The complete calendar time using the AM/PM format of the current locale.
 
diff --git a/time/strftime_l.c b/time/strftime_l.c
index 1205035..25746a0 100644
--- a/time/strftime_l.c
+++ b/time/strftime_l.c
@@ -1085,6 +1085,10 @@ __strftime_internal (CHAR_T *s, size_t maxsize, const CHAR_T *format,
 	  goto underlying_strftime;
 #endif
 
+        case L_('q'):           /* GNU extension.  */
+          DO_NUMBER (1, tp->tm_mon / 3 + 1);
+          break;
+
 	case L_('R'):
 	  subfmt = L_("%H:%M");
 	  goto subformat;
diff --git a/time/tst-strftime.c b/time/tst-strftime.c
index af3ff72..d087266 100644
--- a/time/tst-strftime.c
+++ b/time/tst-strftime.c
@@ -154,6 +154,33 @@ do_test (void)
 	}
     }
 
+  /* Check %q.  */
+  for (size_t mon = 1; mon <= 12; mon++)
+    {
+      char out[2];
+      char exp[2] = {0,};
+      struct tm qtm = { .tm_mon = mon - 1 };
+
+#pragma GCC diagnostic push
+#pragma GCC diagnostic ignored "-Wformat"
+      size_t r = strftime (out, sizeof (out), "%q", &qtm);
+#pragma GCC diagnostic pop
+      if (r == 0)
+        {
+          puts ("strftime(\"%q\") failed");
+          result = 1;
+          break;
+        }
+
+      exp[0] = mon < 4 ? '1' : mon < 7 ? '2' : mon < 10 ? '3' : '4';
+      if (strcmp (out, exp) != 0)
+        {
+          printf ("strftime %%q: expected \"%s\", got \"%s\"\n", exp, out);
+          result = 1;
+          break;
+        }
+    }
+
   return result + do_bz18985 ();
 }
 
-- 
2.5.5

Re: [PATCH] strftime: support %q to output the quarter of year

[Paul Eggert]

Pádraig Brady wrote:

> +          DO_NUMBER (1, tp->tm_mon / 3 + 1);

I couldn't resist puzzling this one out. We can assume 0 <= tp->tm_mon < 12, so 
this should be a tad more efficient:

    DO_NUMBER (1, ((tp->tm_mon * 11) >> 5) + 1);

With gcc 6.2.0 -O2 x86-64, the former uses imull with 6 insns, the latter just 
leal with 4 insns. Perhaps GCC should be smart enough to figure this out on its 
own, but what would be the fun of that?

Re: [PATCH] strftime: support %q to output the quarter of year

[Rafal Luzynski]

Hi,

Shouldn't strptime() family also support the same?

Regards,

Rafal

Re: [PATCH] strftime: support %q to output the quarter of year

[Szabolcs Nagy]

On 03/11/16 19:46, Pádraig Brady wrote:
> This is already supported by gnulib.

why?

Re: [PATCH] strftime: support %q to output the quarter of year

[Pádraig Brady]

On 04/11/16 11:41, Szabolcs Nagy wrote:
> On 03/11/16 19:46, Pádraig Brady wrote:
>> This is already supported by gnulib.

> why?

It's a marginal benefit, but as
noted in the cover of my updated patch:

"Note even though the code is trivial here,
%q is useful from the shell as there you need to:
$(( ($(date +%-m)-1)/3+1 ))".

I'll add that to the actual commit.

Re: [PATCH] strftime: support %q to output the quarter of year

[Szabolcs Nagy]

On 04/11/16 12:00, Pádraig Brady wrote:
> On 04/11/16 11:41, Szabolcs Nagy wrote:
>> On 03/11/16 19:46, Pádraig Brady wrote:
>>> This is already supported by gnulib.
> 
>> why?
> 
> It's a marginal benefit, but as
> noted in the cover of my updated patch:
> 
> "Note even though the code is trivial here,
> %q is useful from the shell as there you need to:
> $(( ($(date +%-m)-1)/3+1 ))".
> 
> I'll add that to the actual commit.

this can conflict with future standard, so
there need to be a strong reason for adding
such extensions to portability libraries
such as gnulib or to c runtimes.

how does gnulib plan to deal with the conflict
once posix adds %q with different meaning?

Re: [PATCH] strftime: support %q to output the quarter of year

[Pádraig Brady]

On 04/11/16 12:07, Szabolcs Nagy wrote:
> On 04/11/16 12:00, Pádraig Brady wrote:
>> On 04/11/16 11:41, Szabolcs Nagy wrote:
>>> On 03/11/16 19:46, Pádraig Brady wrote:
>>>> This is already supported by gnulib.
>>
>>> why?
>>
>> It's a marginal benefit, but as
>> noted in the cover of my updated patch:
>>
>> "Note even though the code is trivial here,
>> %q is useful from the shell as there you need to:
>> $(( ($(date +%-m)-1)/3+1 ))".
>>
>> I'll add that to the actual commit.
> 
> this can conflict with future standard, so
> there need to be a strong reason for adding
> such extensions to portability libraries
> such as gnulib or to c runtimes.
> 
> how does gnulib plan to deal with the conflict
> once posix adds %q with different meaning?

Perl's date lib also uses %q for quarter.
So between that and gnulib (and glibc?)
POSIX would be unlikely to choose %q for something else.
In any case I intend to propose it to the POSIX folks.

thanks,
Pádraig

Re: [PATCH] strftime: support %q to output the quarter of year

[Szabolcs Nagy]

On 04/11/16 12:23, Pádraig Brady wrote:
> On 04/11/16 12:07, Szabolcs Nagy wrote:
>> On 04/11/16 12:00, Pádraig Brady wrote:
>>> On 04/11/16 11:41, Szabolcs Nagy wrote:
>>>> On 03/11/16 19:46, Pádraig Brady wrote:
>>>>> This is already supported by gnulib.
>>>
>>>> why?
>>>
>>> It's a marginal benefit, but as
>>> noted in the cover of my updated patch:
>>>
>>> "Note even though the code is trivial here,
>>> %q is useful from the shell as there you need to:
>>> $(( ($(date +%-m)-1)/3+1 ))".
>>>
>>> I'll add that to the actual commit.
>>
>> this can conflict with future standard, so
>> there need to be a strong reason for adding
>> such extensions to portability libraries
>> such as gnulib or to c runtimes.
>>
>> how does gnulib plan to deal with the conflict
>> once posix adds %q with different meaning?
> 
> Perl's date lib also uses %q for quarter.

i don't know what "Perl's date lib" is,
the perl DateTime module does not seem to support %q.
https://metacpan.org/pod/DateTime#strftime-Patterns

> So between that and gnulib (and glibc?)
> POSIX would be unlikely to choose %q for something else.
> In any case I intend to propose it to the POSIX folks.
> 
> thanks,
> Pádraig
> 

Re: [PATCH] strftime: support %q to output the quarter of year

[Pádraig Brady]

On 04/11/16 12:41, Szabolcs Nagy wrote:
> On 04/11/16 12:23, Pádraig Brady wrote:
>> On 04/11/16 12:07, Szabolcs Nagy wrote:
>>> On 04/11/16 12:00, Pádraig Brady wrote:
>>>> On 04/11/16 11:41, Szabolcs Nagy wrote:
>>>>> On 03/11/16 19:46, Pádraig Brady wrote:
>>>>>> This is already supported by gnulib.
>>>>
>>>>> why?
>>>>
>>>> It's a marginal benefit, but as
>>>> noted in the cover of my updated patch:
>>>>
>>>> "Note even though the code is trivial here,
>>>> %q is useful from the shell as there you need to:
>>>> $(( ($(date +%-m)-1)/3+1 ))".
>>>>
>>>> I'll add that to the actual commit.
>>>
>>> this can conflict with future standard, so
>>> there need to be a strong reason for adding
>>> such extensions to portability libraries
>>> such as gnulib or to c runtimes.
>>>
>>> how does gnulib plan to deal with the conflict
>>> once posix adds %q with different meaning?
>>
>> Perl's date lib also uses %q for quarter.
> 
> i don't know what "Perl's date lib" is,
> the perl DateTime module does not seem to support %q.
> https://metacpan.org/pod/DateTime#strftime-Patterns

Sorry Perl's Date::Format supports %q
https://metacpan.org/pod/Date::Format#CONVERSION-SPECIFICATION

Re: [PATCH] strftime: support %q to output the quarter of year

[keld]

ISO 30112 will poropose %q in the new revision.
I think it is difficult for POSIX then to do something else.

Have you got standardese for the %q spec, which I can use in 30112?

Best regards
keld

On Fri, Nov 04, 2016 at 12:23:28PM +0000, Pádraig Brady wrote:
> On 04/11/16 12:07, Szabolcs Nagy wrote:
> > On 04/11/16 12:00, Pádraig Brady wrote:
> >> On 04/11/16 11:41, Szabolcs Nagy wrote:
> >>> On 03/11/16 19:46, Pádraig Brady wrote:
> >>>> This is already supported by gnulib.
> >>
> >>> why?
> >>
> >> It's a marginal benefit, but as
> >> noted in the cover of my updated patch:
> >>
> >> "Note even though the code is trivial here,
> >> %q is useful from the shell as there you need to:
> >> $(( ($(date +%-m)-1)/3+1 ))".
> >>
> >> I'll add that to the actual commit.
> > 
> > this can conflict with future standard, so
> > there need to be a strong reason for adding
> > such extensions to portability libraries
> > such as gnulib or to c runtimes.
> > 
> > how does gnulib plan to deal with the conflict
> > once posix adds %q with different meaning?
> 
> Perl's date lib also uses %q for quarter.
> So between that and gnulib (and glibc?)
> POSIX would be unlikely to choose %q for something else.
> In any case I intend to propose it to the POSIX folks.
> 
> thanks,
> Pádraig

Re: [PATCH] strftime: support %q to output the quarter of year

[Szabolcs Nagy]

On 04/11/16 13:12, keld@keldix.com wrote:
> ISO 30112 will poropose %q in the new revision.
> I think it is difficult for POSIX then to do something else.

this should be the other way around: propose the interface
to posix, then add locale specifications later, otherwise
the two standards can diverge.

[PATCH v3] strftime,strptime: have %q represent the quarter of year

[Pádraig Brady]

[-- Attachment #1: Type: text/plain, Size: 289 bytes --]

On 04/11/16 11:33, Rafal Luzynski wrote:
> Hi,
> 
> Shouldn't strptime() family also support the same?

I was debating that, but you're right,
strptime() should support that for consistency.

Updated patch is attached, which also adjusts
to Paul Eggert's optimization.

thanks,
Pádraig

[-- Attachment #2: glibc-strptime-%q.patch --]
[-- Type: text/x-patch, Size: 5387 bytes --]

From ad0ed0247612bf7d85afe1af0c8a61e804db274c Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?P=C3=A1draig=20Brady?= <P@draigBrady.com>
Date: Thu, 3 Nov 2016 19:36:59 +0000
Subject: [PATCH] strftime,strptime: have %q represent the quarter of year

This is already supported by gnulib, and for example
simplifies quarter determination at the shell, which
currently needs to use "$(( ($(date +%-m)-1)/3+1 ))".

* manual/time.texi: Document %q for str[fp]time.
* time/strftime_l.c: Implement %q.
* time/strptime_l.c: Likewise.
* time/tst-strftime.c: Add a test case.
* time/tst-strptime3.c: Likewise.
* NEWS: Mention the new feature.
---
 NEWS                 |  3 +++
 manual/time.texi     | 12 ++++++++++++
 time/strftime_l.c    |  7 +++++++
 time/strptime_l.c    | 18 ++++++++++++++++++
 time/tst-strftime.c  | 25 +++++++++++++++++++++++++
 time/tst-strptime3.c | 20 ++++++++++++++++++++
 6 files changed, 85 insertions(+)

diff --git a/NEWS b/NEWS
index 65184b1..e08f9ed 100644
--- a/NEWS
+++ b/NEWS
@@ -119,6 +119,9 @@ Version 2.25
   variable for a particular architecture in the GCC source file
   'gcc/config.gcc'.
 
+* strftime and strptime now support the %q directive to represent
+  the quarter of the year, with January starting the first quarter.
+
 Security related changes:
 
   On ARM EABI (32-bit), generating a backtrace for execution contexts which
diff --git a/manual/time.texi b/manual/time.texi
index 6a899b7..5ba0558 100644
--- a/manual/time.texi
+++ b/manual/time.texi
@@ -1510,6 +1510,12 @@ most locales @samp{AM}/@samp{PM} format is not supported, in such cases
 
 This format is a GNU extension.
 
+@item %q
+Quarter of the year (@samp{1}@dots{}@samp{4}),
+with January starting the first quarter.
+
+This format is a GNU extension.
+
 @item %r
 The complete calendar time using the AM/PM format of the current locale.
 
@@ -1926,6 +1932,12 @@ all and therefore the conversion fails.
 
 @code{%P} is a GNU extension following a GNU extension to @code{strftime}.
 
+@item %q
+Quarter of the year (@samp{1}@dots{}@samp{4}),
+with January starting the first quarter.
+
+This format is a GNU extension.
+
 @item %r
 The complete time using the AM/PM format of the current locale.
 
diff --git a/time/strftime_l.c b/time/strftime_l.c
index 1205035..09d9373 100644
--- a/time/strftime_l.c
+++ b/time/strftime_l.c
@@ -1085,6 +1085,13 @@ __strftime_internal (CHAR_T *s, size_t maxsize, const CHAR_T *format,
 	  goto underlying_strftime;
 #endif
 
+	case L_('q'):           /* Quarter of year.  GNU extension.  */
+	  if (modifier == L_('E'))
+	    goto bad_format;
+
+	  DO_NUMBER (1, ((tp->tm_mon * 11) >> 5) + 1);
+	  break;
+
 	case L_('R'):
 	  subfmt = L_("%H:%M");
 	  goto subformat;
diff --git a/time/strptime_l.c b/time/strptime_l.c
index 3a56947..f957e40 100644
--- a/time/strptime_l.c
+++ b/time/strptime_l.c
@@ -593,6 +593,15 @@ __strptime_internal (const char *rp, const char *fmt, struct tm *tmp,
 	  else
 	    s.is_pm = 0;
 	  break;
+	case 'q':
+	  /* Match quarter of year.  GNU extension.  */
+	  get_number (1, 4, 1);
+	  tm->tm_mon = (val - 1) * 3;
+	  tm->tm_mday = 1;
+	  s.have_mon = 1;
+	  s.have_mday = 1;
+	  s.want_xday = 1;
+	  break;
 	case 'r':
 #ifdef _NL_CURRENT
 	  if (s.decided != raw)
@@ -1050,6 +1059,15 @@ __strptime_internal (const char *rp, const char *fmt, struct tm *tmp,
 	      get_alt_number (0, 59, 2);
 	      tm->tm_min = val;
 	      break;
+	    case 'q':
+	      /* Match quarter using alternate numeric symbols.  */
+	      get_alt_number (1, 4, 1);
+	      tm->tm_mon = (val - 1) * 3;
+	      tm->tm_mday = 1;
+	      s.have_mon = 1;
+	      s.have_mday = 1;
+	      s.want_xday = 1;
+	      break;
 	    case 'S':
 	      /* Match seconds using alternate numeric symbols.  */
 	      get_alt_number (0, 61, 2);
diff --git a/time/tst-strftime.c b/time/tst-strftime.c
index af3ff72..62ee6c0 100644
--- a/time/tst-strftime.c
+++ b/time/tst-strftime.c
@@ -154,6 +154,31 @@ do_test (void)
 	}
     }
 
+  /* Check %q.  */
+  for (size_t mon = 1; mon <= 12; mon++)
+    {
+      char out[2];
+      char exp[2] = {0,};
+      struct tm qtm = { .tm_mon = mon - 1 };
+      char fmt[3] = {'%','q','\0'};
+
+      size_t r = strftime (out, sizeof (out), fmt, &qtm);
+      if (r == 0)
+        {
+          puts ("strftime(\"%q\") failed");
+          result = 1;
+          break;
+        }
+
+      exp[0] = mon < 4 ? '1' : mon < 7 ? '2' : mon < 10 ? '3' : '4';
+      if (strcmp (out, exp) != 0)
+        {
+          printf ("strftime %%q: expected \"%s\", got \"%s\"\n", exp, out);
+          result = 1;
+          break;
+        }
+    }
+
   return result + do_bz18985 ();
 }
 
diff --git a/time/tst-strptime3.c b/time/tst-strptime3.c
index d53f51e..b08f6f4 100644
--- a/time/tst-strptime3.c
+++ b/time/tst-strptime3.c
@@ -48,6 +48,26 @@ do_test (void)
       result = 1;
     }
 
+  memset (&tm, 0xaa, sizeof (tm));
+  for (size_t q = 1; q <= 4; q++)
+    {
+      char in[2] = {0,};
+      in[0] = q + '0';
+      int exp_mon = q == 1 ? 0 : q == 2 ? 3 : q == 3 ? 6 : 9;
+      char fmt[3] = {'%','q','\0'};
+
+      if (strptime (in, fmt, &tm) == NULL)
+        {
+          puts ("strptime failed");
+          result = 1;
+        }
+      if (tm.tm_mon != exp_mon || tm.tm_mday != 1)
+        {
+          puts ("unexpected tm content");
+          result = 1;
+        }
+    }
+
   return result;
 }
 
-- 
2.5.5

Re: [PATCH] strftime: support %q to output the quarter of year

[Adhemerval Zanella]

On 04/11/2016 12:12, Szabolcs Nagy wrote:
> On 04/11/16 13:12, keld@keldix.com wrote:
>> ISO 30112 will poropose %q in the new revision.
>> I think it is difficult for POSIX then to do something else.
> 
> this should be the other way around: propose the interface
> to posix, then add locale specifications later, otherwise
> the two standards can diverge.

I tend to agree with this rationale.

Re: [PATCH] strftime: support %q to output the quarter of year

[keld]

On Fri, Nov 04, 2016 at 02:12:28PM +0000, Szabolcs Nagy wrote:
> On 04/11/16 13:12, keld@keldix.com wrote:
> > ISO 30112 will poropose %q in the new revision.
> > I think it is difficult for POSIX then to do something else.
> 
> this should be the other way around: propose the interface
> to posix, then add locale specifications later, otherwise
> the two standards can diverge.

Oh, well, 30112 has taken the lead  on many i18n issues, and this is also the agreement with the posix folks.

Best regards
keld