develooper Front page | perl.perl6.language.datetime | Postings from September 2000

RFC 48 (v4) Replace localtime() and gmtime() with date() and utcdate()

Thread Next
From:
Perl6 RFC Librarian
Date:
September 25, 2000 22:49
Subject:
RFC 48 (v4) Replace localtime() and gmtime() with date() and utcdate()
Message ID:
20000926054938.21561.qmail@tmtowtdi.perl.org
This and other RFCs are available on the web at
  http://dev.perl.org/rfc/

=head1 TITLE 

Replace localtime() and gmtime() with date() and utcdate()

=head1 VERSION

  Maintainer: Nathan Wiger <nate@wiger.org>
  Date: 5 Aug 2000
  Last Modified: 25 Sep 2000
  Mailing List: perl6-language-datetime@perl.org
  Number: 48
  Version: 4
  Status: Frozen
  Requires: RFC 21, RFC 159

=head1 ABSTRACT

Currently, Perl uses the C library C<localtime()> and C<gmtime()>
functions for date access. However, because of many problems, these
should be replaced with two new functions, C<date()> and
C<utcdate()> that will be called as follows:

   $object  =  date;     # object with accessor functions (new)
   print "$object";      # scalar ctime date, same as current
   @array   =  date;     # list of date/time values (new order)
   %hash    =  date;     # hash of date/time values (new)

The new C<date()> function will return time adjusted for the local
timezone. The C<utcdate()> function will return time in UTC. In a scalar
context, an object is always returned. This object is then converted to
a string on-demand via B<RFC 159: True Polymorphic Objects>.

=head1 NOTES ON FREEZE

Everyone felt pretty good about this, I think. The only thing we'd all
probably like to see is one single C<date> function, but unfortunately
dealing with timezone specifications is an extraordinarily difficult
issue. 

=head1 DESCRIPTION

=head2 Overview

In the past, Perl has provided access to date and time information
through the C library C<localtime()> and C<gmtime()> functions.
Unfortunately, these functions have several "really bad things" about
them:

   1. some values are 0-indexed while others are 1-indexed
   2. they return massive lists as their only interface
   3. minute and second values aren't 0-padded
   4. the year has to have 1900 added to it to be correct.

While some of these are perhaps understandable, that last one just plain
doesn't make I<any> sense. :-)

With Perl 6 we have the chance to fix this. While some have suggested
merely changing what C<localtime()> and C<gmtime()> return, this is a
bad idea because:

   1. Many Perl scripts already use it
   2. Many C programmers are familiar with its stupidity

As such, we should replace these functions with versions that work the
way that we want them to. The new names will indicate that these are
"not the same old time functions", and that you'd better read the docs.

=head2 Proposed Syntax

The new syntax of C<date()> and C<utcdate> is proposed as:

   $|@|%return  =  date [ $time ], [ $format ];

If C<$time> is not specified, then the current return value from
C<time()> is used (just like C<localtime()>).

The optional C<$format> affects the format of the date returned. By
default, a ctime-formatted string is used, like C<localtime>. However,
any POSIX-compatible format can be passed in, in which case the date
returned in a scalar context is formatted accordingly.

I<Actually>, what it does is call the object function C<setformat()>
with the format, which in turn determines what calls to the C<date()>
object member function look like. Since the C<date()> member function
is what is output when the date object is stringified, this is how 
the scalar date appears to change. See below.

The C<$format> specifier is the same one used in C<POSIX strftime()>,
allowing complete date formatting with having to import the C<POSIX>
module.

Depending on the context within which C<date()> is called, many
different things are returned:

   # Get some various local date stuff
   $object  =  date;            # scalar ctime date, same as now
   $object  =  date time, '%m/%d/%Y';  # set date MM/DD/YYYY format
   $object  =  date $time;      # relative to time $time
   print "$object";             # scalar date string per format   

   @array   =  date;            # array of date/time values
   %hash    =  date;            # hash of values

   # Access UTC information
   $object  =  utcdate time, '%H:%M'; # return time in UTC

=head2 Return Values

The return values are dependent on the context within which C<date()> is
called. For all contexts, the following are the values and their ranges:

   $hour  =  0 .. 23  
   $min   =  00 .. 59   # 0-padded
   $sec   =  00 .. 59   # 0-padded
   $fsec  =  0 .. 1     # fractional seconds per hw clock

   $month =  1 .. 12    # hooray!
   $day   =  1 .. 31 
   $year  =  1 .. 9999  # 4-digit! 

   $wday  =  1 .. 7     # 1 == *Monday*, POSIX-compliant
   $yday  =  1 .. 366

   $isdst =  1 or undef;  # daylight savings?
   $isgmt =  1 or undef;  # is GMT?
   $tz    =  [depends on your system, this "eats people"]

For those who are concerned about $wday not being 0-indexed, please see
below under the object methods. Remember you can always use formats to
return what you want as well.

=head3 STRING Context

In a C<STRING> (not scalar) context, a single ctime-formatted date
string is returned, consistent with the current usage of C<localtime()>.
If the optional C<$format> is included, this format is used to change
the string accordingly.

In order to emulate a string, a call is made to the date object's
builtin C<STRING> function, which prints out a date string. This is per
RFC 159: True Polymorphic Objects.

=head3 LIST, ARRAY, and ARRAYREF Contexts

A list of date/time values is returned. The ordering and format of these
values has been radically changed to reflect what most of us probably
view as "ordinary":

  ($year, $month, $day, $hour, $min, $sec, $tz) = date;

You'll note that many elements, such as $isdst, $wday, and $yday have
been dropped from the list return context. This is to encourage the use
of more sane contexts, such as object and hash/hashrefs for complex
operations.

This ordering follows the predictable pattern of being in increasing
granularity and so should be easy to remember. Just think of a
computer-esque decending timestamp:

   2000/11/4 12:03:09

Note that the month, day, and year are not 0-padded. If you want a
padded date, perhaps as a file backup suffix, you'll have to say:

   $backup_suffix = date time, '%Y%m%d%H%M%S';

Which would return something like "20001104120309" (actually, it would
return an object with the appropriate default format so that, when
stringified, it would yield the above result).

These same values can be retrieved when referred to as an arrayref,
so these two:

    $month = (date)[1];
    $month = date->[1];

Have the same effect. Now, of course, if you want specific values you
should be using...

=head3 HASH and HASHREF Contexts

A hash of values is returned. These are named the same as the variables
above. So,

   %date = date;

Would return a hash with these values (assuming the same date shown
above):

   $date{year}  = 2000;
   $date{month} = 11; 
   $date{day}   = 4;   
   $date{hour}  = 12;
   $date{min}   = 03; 
   $date{sec}   = 09;
   $date{fsec}  = 0.4291;  # up to hw resolution
   $date{wday}  = 2; 
   $date{yday}  = 331;
   $date{tz}    = 'PST';   # or maybe 'US/Pacific', or ...
   $date{isdst} = undef;

These same values will be available via a HASHREF, so again these:

   $month = ${(date)}{month};
   $month = date->{month};

Will work the same (at least, I think that first one's right per RFC
21)...

=head3 SCALAR Context

In an C<SCALAR> context, an object is returned. This is somewhat adapted
from C<Time::Object> and Larry's own comments, but with several key
differences. The ones noted with * are incompatible with C<Time::Object>

   $t = date;            # create date object

   "$t"                  # calls $t->STRING, same as $t->date
   $t++                  # calls $t->NUMBER, same as $t->time

   $t->year
   $t->month             # *
   $t->day               # *
   $t->hour
   $t->min
   $t->sec
   $t->fsec              # fractional seconds
   $t->fsecres           # fsec resolution (num of digits for hw)
   $t->wday              # *
   $t->yday
   $t->isdst
   $t->tz
   $t->tzoffset          # timezone offset in seconds ($t->tzsec?)
   $t->time              # number of seconds since the epoch, so it
                         # contains what time() was when you called
                         # date() 

These functions are object-specific:

   $t->date              # Tue Feb 29 01:23:45 2000 (depending on 
                         # default format set by $t->setformat)

   $t->date(FORMAT)      # Return a temporarily-modified date,
                         # without affecting the default format.

   $t->setformat(FORMAT) # Set default format which determines what
                         # calls to $t->date (stringified $t) look like
                         # This is called on object creation

These functions are separate for an important reason:

   print $t->date('%H:%M');    # prints out hours and minutes
   print "$t";                 # without touching default format

   $oldfmt = $t->setformat('%Y');  # sets POSIX fmt string
   print "$t";                 # which affects the default format

The C<set> prefix was explicitly added to make it obvious this sets
something (rather than just formatting the date like strftime).

Many functions were deemed of dubious quality/excessive bloat and were
all removed from the RFC. Note that you can get to any extra data by
calling $t->date with the correct format. So,

   print $t->date('%w');

Will give you the weekday numbered starting with 0 == Sunday.

=head2 Date Arithmetic

You'll note that simple date arithmetic is available via polymorphic
overloading in numeric contexts. For example:

   $t1 = date;
   $t2 = date;
   $diff = $t2 - $t1;   # calls ->NUMBER (->time) on both

The result will simply be the difference in seconds between the two, in
simple scalar form (i.e., "15"). The result is designed to be the same
as if you had created $t1 and $t2 based on the output of time(), without
requiring you to put extra calls in your code. To simply increment the
date by one second each loop, do:

   $d = date;
   while (stuff()) {
      $d++;
   }

This gives the basis from which more complex date arithmetic can be
derived, which should be in a module.

=head1 IMPLEMENTATION

Slash and burn C<localtime()> and C<gmtime()>. Actually, we should
probably move them to C<Time::Local> or some other place and keep them
available, since some people like pain.

Nonetheless, C<localtime()> and C<gmtime()> SHOULD BE REMOVED FROM CORE
and placed in a module. Perl 6 should support ONE date/time interface
natively.

Extensive date calculations, including reverse date calculations, should
be left to an external module.

I am currently working on a Perl 5 prototype and will announce it to the
list when it is completed.

=head1 ACKNOWLEDGEMENTS

Jonathan Scott Duff, Tim Jenness, Johan Vromans, Buddha Buck, and lots
of other members of perl6-language for great feedback.

Russ Allbery and John Tobey for suggesting parts to slice-and-dice.

=head1 REFERENCES

Larry Wall's post of 8 Jan 2000 on deprecating localtime:
http://www.xray.mpe.mpg.de/mailing-lists/perl5-porters/2000-01/msg00241.html

RFC 159: True Polymorphic Objects

RFC 21: Replace wantarray() with a generic want() function

RFC 37: Positional Return Lists Considered Harmful

Matt Sergeant's great Time::Object CPAN module

http://www.jach.hawaii.edu/JACpublic/stardocs/sun67.htx/node217.html

http://tycho.usno.navy.mil/systime.html


Thread Next


nntp.perl.org: Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at ask@perl.org | Group listing | About