Dave Rolsky wrote:
>First, I don't want to make these new modules requirements. What I'd
>prefer to do is just load them if they exist and use them. No one (that I
>recall) has asked for Posix or binary file support, so making them
>dependencies seems like overkill.
OK. But it seems a pity that a default install won't be able to parse
all $TZ settings that are valid for libc. Actually there have been
two DT::TZ users who wanted their System V timezone strings handled
(see bottom of this message).
>Given that, I don't want to change the handling of these names. This would
>be a backwards-incompatible change as it is, since in the Olson database a
>name like EST5EDT uses the US rules, meaning it includes many historical
>changes.
Fair enough. Using US rules is a valid interpretation of "EST5EDT"
as a System V string anyway.
>This should probably use File::Spec->file_name_is_absolute() to figure
>this out. None of the currently valid names are absolute file names on any
>system (I think).
That would be a bad idea: it would mean that unknown filename syntaxes
constrain what system-independent syntaxes we can use in the future.
It's theoretically totally unworkable. Anyway, among what we do know,
on the Mac any string that doesn't start with ":" is a valid absolute
filename.
I found a better way: URI::file can translate between filename syntaxes.
We can specify that the $TZ setting is a Unix-style filename, and it gets
translated as appropriate. On Windows you can do "/c:/etc/localtime"
and the file C:\etc\localtime is used, and "/etc/localtime" translates
to the not-quite-absolute \etc\localtime.
>It seems like a leading colon for other things (like "EST5EDT") would be
>an error. Of course, removing it isn't a big deal.
I addressed this issue in my survey of $TZ parsing. glibc and HP-UX
both accept System V strings both with and without leading ":". glibc,
Olson, and Solaris all accept filenames (including Olson names) both
with and without ":". Basically, the consensus $TZ syntax is the same
both with and without ":". Consistently ignoring the leading ":" is
where the trend is heading.
>1. Make SystemV & Tzfile entirely optional.
>2. Use the Olson versions of existing SystemV names (follows from 1)
>3. Use File::Spec->file_name_is_absolute to determine if a name could be a
>path to a tz file.
Patch attached does 1 and 2, and does filename translation instead of 3.
I extended 2 to the non-SysV single-part Olson names that my previous
patch implemented as links to SysV strings.
There were also two SysV-style strings that were never in Olson, but which
DT::TZ had as links to Olson zones. These are "AKST9AKDT" and "JST-9".
They were added to satisfy particular DT::TZ users who were using these
$TZ settings. As they're not in Olson, I reckon these should be treated
as any ordinary SysV string, so this patch has no grandfathering for them.
Those users will have to install DT::TZ::SystemV. Grandfathering for
them is easily added if you really want to.
In other respects it's equivalent to the previous patch. As before
you'll need to rerun tools/parse_olson, to pick up the changes to
TimeZoneCatalog.pm.
-zefram
diff -ur dttz-0.6101/lib/DateTime/TimeZone/Local.pm
dttz-mod1/lib/DateTime/TimeZone/Local.pm
--- dttz-0.6101/lib/DateTime/TimeZone/Local.pm 2007-02-18 15:54:12.000000000
+0000
+++ dttz-mod1/lib/DateTime/TimeZone/Local.pm 2007-02-19 19:30:30.018396544
+0000
@@ -92,7 +92,7 @@
return 0 unless defined $_[0];
return 0 if $_[0] eq 'local';
- return $_[0] =~ m{^[\w/\-\+]+$};
+ return 1;
}
diff -ur dttz-0.6101/lib/DateTime/TimeZone/OlsonDB.pm
dttz-mod1/lib/DateTime/TimeZone/OlsonDB.pm
--- dttz-0.6101/lib/DateTime/TimeZone/OlsonDB.pm 2007-02-18
15:54:11.000000000 +0000
+++ dttz-mod1/lib/DateTime/TimeZone/OlsonDB.pm 2007-02-19 00:02:04.000000000
+0000
@@ -111,8 +111,6 @@
$name = shift @items;
}
- return if $name =~ /[WCME]ET/ && ! $self->{backwards_compat};
-
@obs{ qw( gmtoff rules format until ) } = @items;
if ( $obs{rules} =~ /\d\d?:\d\d/ )
diff -ur dttz-0.6101/lib/DateTime/TimeZone.pm dttz-mod1/lib/DateTime/TimeZone.pm
--- dttz-0.6101/lib/DateTime/TimeZone.pm 2007-02-18 15:54:12.000000000
+0000
+++ dttz-mod1/lib/DateTime/TimeZone.pm 2007-02-19 20:19:13.383139216 +0000
@@ -6,10 +6,6 @@
$VERSION = '0.6101';
use DateTime::TimeZoneCatalog;
-use DateTime::TimeZone::Floating;
-use DateTime::TimeZone::Local;
-use DateTime::TimeZone::OffsetOnly;
-use DateTime::TimeZone::UTC;
use Params::Validate qw( validate validate_pos SCALAR ARRAYREF BOOLEAN );
use constant INFINITY => 100 ** 1000 ;
@@ -24,7 +20,8 @@
use constant IS_DST => 5;
use constant SHORT_NAME => 6;
-my %SpecialName = map { $_ => 1 } qw( EST MST HST EST5EDT CST6CDT MST7MDT
PST8PDT );
+my %grandfather_olson = map { $_ => 1 }
+ qw( CST6CDT EST EST5EDT HST MST MST7MDT PST8PDT );
sub new
{
@@ -33,66 +30,85 @@
{ name => { type => SCALAR } },
);
- if ( exists $DateTime::TimeZone::LINKS{ $p{name} } )
+ if ( $p{name} eq 'local' )
{
- $p{name} = $DateTime::TimeZone::LINKS{ $p{name} };
+ require DateTime::TimeZone::Local;
+ return DateTime::TimeZone::Local->TimeZone;
}
- elsif ( exists $DateTime::TimeZone::LINKS{ uc $p{name} } )
+
+ my $name = $p{name};
+ $name =~ s/\A://;
+
+ if ( exists $DateTime::TimeZone::LINKS{ $name } )
{
- $p{name} = $DateTime::TimeZone::LINKS{ uc $p{name} };
+ $name = $DateTime::TimeZone::LINKS{ $name };
}
-
- unless ( $p{name} =~ m,/,
- || $SpecialName{ $p{name} }
- )
+ elsif ( exists $DateTime::TimeZone::LINKS{ uc $name } )
{
- if ( $p{name} eq 'floating' )
- {
- return DateTime::TimeZone::Floating->new;
- }
-
- if ( $p{name} eq 'local' )
- {
- return DateTime::TimeZone::Local->TimeZone();
- }
-
- if ( $p{name} eq 'UTC' || $p{name} eq 'Z' )
- {
- return DateTime::TimeZone::UTC->new;
- }
-
- return DateTime::TimeZone::OffsetOnly->new( offset => $p{name} );
+ $name = $DateTime::TimeZone::LINKS{ uc $name };
}
- my $subclass = $p{name};
- $subclass =~ s/-/_/g;
- $subclass =~ s{/}{::}g;
- my $real_class = "DateTime::TimeZone::$subclass";
-
- die "The timezone '$p{name}' in an invalid name.\n"
- unless $real_class =~ /^\w+(::\w+)*$/;
-
- unless ( $real_class->can('instance') )
+ if ( $name eq 'UTC' || $name eq 'Z' )
{
- eval "require $real_class";
+ require DateTime::TimeZone::UTC;
+ return DateTime::TimeZone::UTC->new;
+ }
+ elsif ( $name =~ m#\A[-\w]+(?:/[-\w]+)+\z# || $grandfather_olson{$name} )
+ {
+ my $subclass = $name;
+ $subclass =~ s/-/_/g;
+ $subclass =~ s{/}{::}g;
+ my $real_class = "DateTime::TimeZone::$subclass";
- if ($@)
+ unless ( $real_class->can('instance') )
{
- my $regex = join '.', split /::/, $real_class;
- $regex .= '\\.pm';
+ eval "require $real_class";
- if ( $@ =~ /^Can't locate $regex/i )
+ if ($@)
{
- die "The timezone '$p{name}' could not be loaded, or is an
invalid name.\n";
- }
- else
- {
- die $@;
+ my $regex = join '.', split /::/, $real_class;
+ $regex .= '\\.pm';
+
+ if ( $@ =~ /^Can't locate $regex/i )
+ {
+ die "The timezone '$name' could not be loaded, or is an
invalid name.\n";
+ }
+ else
+ {
+ die $@;
+ }
}
}
+ return $real_class->instance( name => $name, is_olson => 1 );
+ }
+ elsif ( $name =~ /\A(?:[A-Za-z]{3,}|\<[-+0-9A-Za-z]{3,}\>)[-+]?\d/ )
+ {
+ require DateTime::TimeZone::SystemV;
+ return DateTime::TimeZone::SystemV->new( $name );
+ }
+ elsif ( $name =~ m#\A/# )
+ {
+ require URI::file;
+ my $filename = URI::file->new($name, "Unix")->file;
+ die "The timezone name '$name' couldn't be translated to a filename.\n"
+ unless defined $filename;
+ require DateTime::TimeZone::Tzfile;
+ return DateTime::TimeZone::Tzfile->new( $filename );
+ }
+ elsif ( $name =~ /\A[-+]?(?:\d\d?:\d\d(?::\d\d)?|\d{4}(?:\d\d)?)\z/ )
+ {
+ require DateTime::TimeZone::OffsetOnly;
+ return DateTime::TimeZone::OffsetOnly->new( offset => $name );
+ }
+ elsif ( $name eq 'floating' )
+ {
+ require DateTime::TimeZone::Floating;
+ return DateTime::TimeZone::Floating->new;
+ }
+ else
+ {
+ die "The timezone name '$name' was not understood.\n";
}
-
- return $real_class->instance( name => $p{name}, is_olson => 1 );
}
sub _init
@@ -521,29 +537,83 @@
=head2 DateTime::TimeZone->new( name => $tz_name )
-Given a valid time zone name, this method returns a new time zone
-blessed into the appropriate subclass. Subclasses are named for the
-given time zone, so that the time zone "America/Chicago" is the
+Parses the given timezone name, and returns a timezone object representing
+the appropriate zone. The following types of name are understood:
+
+=over
+
+=item *
+
+If the name parameter is "local", then the module attempts to determine
+the local time zone for the system. This is described in more detail
+below.
+
+=item *
+
+If the name begins with a ":" character, that character is dropped and the
+rest of the name is parsed without it. This is for compatibility with
+parsing of the B<TZ> environment variable on many systems. The special
+meaning of "local" does not apply to ":local".
+
+=item *
+
+If the name is an alias established by L<DateTime::TimeZone::Alias>,
+or a "link" in the Olson database, or one of a small number of built-in
+aliases, then the target of the alias is parsed instead. An alias name
+that is all uppercase will be recognised in any case, but any other alias
+name must be given exactly. The target of an alias is not itself subject
+to ":" removal or alias interpretation, and "local" is not a valid target.
+
+=item *
+
+If the name is "UTC" or "Z", then a C<DateTime::TimeZone::UTC>
+object is returned.
+
+=item *
+
+If the name looks like a multipart name in the Olson database, or is
+one of a small number of grandfathered single-part names, an appropriate
+subclass is used. For example, the time zone "America/Chicago" is the
DateTime::TimeZone::America::Chicago class.
-If the name given is a "link" name in the Olson database, the object
-created may have a different name. For example, there is a link from
-the old "EST5EDT" name to "America/New_York".
+=item *
+
+If the name looks like a System V style timezone string,
+including one of the POSIX extended form, it is parsed as such
+by C<DateTime::TimeZone::SystemV>. Exception: the System V style
+timezone strings "CST6CDT", "EST5EDT", "MST7MDT", and "PST8PDT" (none
+of which specify their DST rules) are treated as Olson database names,
+for historical reasons. Note: the System V timezone facility is only
+available if the module C<DateTime::TimeZone::SystemV> is installed;
+a default installation of C<DateTime::TimeZone> does not include that
+modules.
+
+=item *
+
+If the name begins with a "/" character, it is interpreted as a
+Unix-style absolute filename. It is translated into the local filename
+syntax, and used as the name of a L<tzfile(5)> file. The file is
+parsed by C<DateTime::TimeZone::Tzfile>. Note: this facility is only
+available if the modules C<URI::file> (for the filename translation) and
+C<DateTime::TimeZone::Tzfile> (for the parsing) are installed; a default
+installation of C<DateTime::TimeZone> does not include those modules.
-There are also several special values that can be given as names.
+=item *
-If the "name" parameter is "floating", then a
+If the name looks like a fixed offset in hours and minutes
+(and optional seconds), it is converted to a number, and a
+C<DateTime::TimeZone::OffsetOnly> object is returned.
+
+=item *
+
+If the name is "floating", then a
C<DateTime::TimeZone::Floating> object is returned. A floating time
zone does have I<any> offset, and is always the same time. This is
useful for calendaring applications, which may need to specify that a
given event happens at the same I<local> time, regardless of where it
occurs. See RFC 2445 for more details.
-If the "name" parameter is "UTC", then a C<DateTime::TimeZone::UTC>
-object is returned.
-
-If the "name" is an offset string, it is converted to a number, and a
-C<DateTime::TimeZone::OffsetOnly> object is returned.
+=back
=head3 The "local" time zone
diff -ur dttz-0.6101/tools/parse_olson dttz-mod1/tools/parse_olson
--- dttz-0.6101/tools/parse_olson 2007-02-18 15:54:12.000000000 +0000
+++ dttz-mod1/tools/parse_olson 2007-02-19 21:20:36.000420408 +0000
@@ -20,6 +20,9 @@
my $INFINITY = 100 ** 100 ** 100;
+my %grandfather_olson_zones = map { $_ => 1 }
+ qw( CST6CDT EST EST5EDT HST MST MST7MDT PST8PDT );
+
my %opts;
GetOptions( 'dir:s' => \$opts{dir},
'clean' => \$opts{clean},
@@ -151,6 +154,7 @@
foreach my $zone_name ( sort $odb->zone_names )
{
+ next unless $zone_name =~ m{/} || $grandfather_olson_zones{$zone_name};
if ( $opts{name} )
{
next unless $zone_name eq $opts{name};
@@ -377,23 +381,6 @@
sub clean_links
{
- # override some links and add others
- %links =
- ( %links,
- 'Etc/GMT' => 'UTC',
- 'Etc/GMT+0' => 'UTC',
- 'Etc/Universal' => 'UTC',
- 'Etc/UCT' => 'UTC',
- 'Etc/UTC' => 'UTC',
- 'Etc/Zulu' => 'UTC',
- 'GMT0' => 'UTC',
- 'GMT' => 'UTC',
- 'AKST9AKDT' => 'America/Anchorage',
- 'JST-9' => 'Asia/Tokyo',
- );
-
- delete $links{UTC};
-
# Some links resolve to other links - chase them down until they point
# to a real zone.
while ( my @k = grep { $links{ $links{$_} } } keys %links )
@@ -403,6 +390,24 @@
$links{$k} = $links{ $links{$k} };
}
}
+
+ # Anything linking to Etc/UTC, Etc/UCT, or Etc/GMT we want to link to
+ # our UTC.
+ foreach ( keys %links )
+ {
+ $links{$_} = 'UTC' if $links{$_} =~ m#\AEtc/(?:U(?:TC|CT)|GMT)\z#;
+ }
+
+ # The UTC-equivalent zones themselves need to be links to our UTC.
+ # We also support some other Etc/ links, even though we don't provide
+ # the full set of Etc/ names.
+ foreach (qw(Etc/UTC Etc/UCT Etc/GMT Etc/GMT+0 Etc/Universal Etc/Zulu GMT))
+ {
+ $links{$_} = 'UTC'
+ }
+
+ # UTC itself must not be a link: for us it's the canonical name of a zone.
+ delete $links{UTC};
}
sub make_catalog_pm
