Re: Strptime issues
[EMAIL PROTECTED] wrote:
>
> Sounds good .. and it's as simple as using
> DateTime->today as the base, which may well be a
> good default behaviour for ->to_datetime when no
> base is in the object or none is passed in the call.
So this is what we've got so far:
0.0102 2003-11-06
- to_datetime() uses "today" as a base date, if none was given.
The following methods are affected by this change:
epoch, hires_epoch, is_dst, utc_rd_values, utc_rd_as_seconds,
sprintf("%s")
- can_be_datetime()
- has_date(), has_time()
- defined_fields() method, contributed by Rick Measham
- to_datetime() sets fields in a known order,
instead of hash-order. Based on a patch by Rick Measham
- Included has() method, contributed by Rick Measham
0.0101 2003-11-03
- runs in Perl 5.00503.
Dave Rolsky & Flavio S. Glock, creating a cvs conflict :)
- there is no set_locale() in DateTime.
Modified from a patch by Rick Measham
- fixed an infinite loop in _format_nanoseconds.
Patch by Rick Measham
This is a sample of the new API:
$dti = $dti->to_datetime;
if ( $dti->can_be_datetime )
{
$dti = $dti->to_datetime
}
else
{
print "only @{[ $dti->defined_fields ]} are set"
}
if ( $dti->can_be_datetime && $dti->has_date ) ...
if ( $dti->has_date && $dti->has_time ) ...
if ( $dti->has_year ) ...
if ( $dti->has( qw( year month day hour ) ) ) ...
- Flavio S. Glock
Re: Re: Strptime issues
> Flavio S. Glock <[EMAIL PROTECTED]> wrote:
>
> How about to extend the 'to_datetime' method,
instead of creating a
> new
> one:
>
>if ($dti->can_be_datetime) {
> $dti = $dti->to_datetime
>}
>
Sounds good .. and it's as simple as using
DateTime->today as the base, which may well be a
good default behaviour for ->to_datetime when no
base is in the object or none is passed in the call.
Cheers!
Rick
Re: Strptime issues
Rick Measham wrote:
>
>if ($dti->can_be_datetime) {
> $dti->become_datetime
>}
implemented as:
if ($dti->can_be_datetime) {
$dti = $dti->become_datetime
}
because changing the class of '$dti' is a bit hard.
How about to extend the 'to_datetime' method, instead of creating a new
one:
if ($dti->can_be_datetime) {
$dti = $dti->to_datetime
}
- Flavio S. Glock
Re: Re: Strptime issues
> Flavio S. Glock <[EMAIL PROTECTED]> wrote:
> > sub has {
>
> I implemented this in CVS with tests, but I believe it
> should be 2 separate methods instead. What do
you think?
I'm cool with two different methods ... it really is two
different things. Maybe there's also a need for
has_any('hour', 'minute') which returns true if it has
either of hours or minutes.
> > > * can_be_datetime / become_datetime
> > sub can_be_datetime {
> I think any dti that has_year can become a
datetime.
> For example, DT::I->new( year=>2003
)->become_datetime->datetime
> 2003-01-01T00:00:00
Hmmm ... I think this is OK, but I'd like some way then
to control what is required.
$dti->become_datetime if
$dti->can_be_datetime('minutes')
requires:
-MM-DD HH:MM
-MM-DD HH:MM:SS
-MM-DD HH:MM:SS.N
and not
-MM-DD HH:MM:xx.N
That is, it must have years to minutes and no holes.
Re: Strptime issues
[EMAIL PROTECTED] wrote:
>
> > Flavio S. Glock <[EMAIL PROTECTED]> wrote:
> >
> > * keys or defined_fields
> > returns the list of names of "defined" fields
>
> sub has {
># called with parameters you get true or false
># called with no params and you get a list of fields
I implemented this in CVS with tests, but I believe it
should be 2 separate methods instead. What do you think?
> > * can_be_datetime / become_datetime
>
> sub can_be_datetime {
># -MM-DD
># -MM-DD HH:MM
># -MM-DD HH:MM:SS
># -MM-DD HH:MM:SS.N
I think any dti that has_year can become a datetime.
For example, DT::I->new( year=>2003 )->become_datetime->datetime
2003-01-01T00:00:00
Or perhaps, allow any dti without 'holes':
-MM
-MM-DD
-MM-DD HH
-MM-DD HH:MM
-MM-DD HH:MM:SS
-MM-DD HH:MM:SS.N
this excludes, for example:
--DD
- Flavio S. Glock
Re: Re: Re: Strptime issues
> [EMAIL PROTECTED]
<[EMAIL PROTECTED]> wrote:
> sub has_time { $_[0]->has{'hour', 'minute'} }
Maybe should be:
sub has_time {
return 1 if ($_[0]->has('hour', 'minute') and not
$_[0]->has('nanosecond'));
return 1 if ($_[0]->has('hour', 'minute', 'second'))
return 0
}
Which only returns true if we have HH:MM or
HH:MM:SS or HH:MM:SS.N+. The older version
allowed us to get true for HH:MM:xx.N+ which is not
really a time.
Same caveat:
Above code passes all tests on the perl installed in
my head. Havn't tried any of it with the much fussier
software version.
perl -v
This is perl, v5.8.0 built for ricks-brain-1.0
Re: Re: Strptime issues
[webmail logged me out, so sorry if you get ++ copies]
> Flavio S. Glock <[EMAIL PROTECTED]> wrote:
> So this is the current list of proposed
DT::Incomplete methods that
> are
> "waiting for votes":
>
> * epoch
> $epoch = $dti->epoch if $dti->can_be_datetime;
Sounds good
sub epoch {
my $self = shift;
return undef unless $self->can_be_datetime;
return $self->clone->become_datetime->epoch;
}
> * has_date / has_time
Sounds good, see below and:
sub has_time { $_[0]->has{'hour', 'minute'} }
sub has_date { $_[0]->has{'year', 'month', 'day'} }
> * keys or defined_fields
> returns the list of names of "defined" fields
sub has {
# called with parameters you get true or false
# called with no params and you get a list of fields
my $self = shift;
if (@_) {
foreach (@_) {
return 0 unless exists $self->{has}{$_}
}
return 1
}
return keys %{$self->{has}}
}
Which gives:
print "Got hour" if $dti->has('hour');
print "Got time" if $dti->has('hour', 'minute')
and
print "Object has " . join(', ', @{$dti->has} );
> * join( $dti )
> join the "defined" keys of two DT::Incomplete
objects
'join' feels like 'append'. 'union' feels better, but relates
to sets. Maybe 'merge'? Also consider 'overlay'.
dti1 2003-xx-21Txx:18:xx.12345
dti2 2003-09-22T15:17:xx.x
merge: 2003-09-21T15:18:xx.12345
overlay: 2003-09-22T15:17:xx.12345
Note that $dti1->overlay($dti2) is the same as
$dti2->merge($dti1);
sub merge { # not 'submerge'!!!
my $self = shift;
my $dti = shift;
die "Argument to merge must be a DateTime or
DateTime::Incomplete" unless
ref($dti)=~/^DateTime(::Incomplete)?$/;
# There must be a better way to do this:
$self->{has}{year} ||= $dti->year;
$self->{has}{month} ||= $dti->month;
$self->{has}{day} ||= $dti->day;
$self->{has}{hour} ||= $dti->hour;
$self->{has}{minute} ||= $dti->minute;
$self->{has}{second} ||= $dti->second;
$self->{has}{nanosecond} ||= $dti->nanosecond;
$self->{has}{time_zone} ||= $dti->time_zone;
$self->{has}{locale} ||= $dti->locale;
return $self
}
sub overlay { $_[0] = $_[1]->clone->merge( $_[0] ) }
### PROBABLY A VERY UN-KOSHER WAY TO DO
IT! ###
> * is_incomplete / is_complete
sub is_complete { $_[0]->has('year', 'month', 'day',
'hour', 'minute', 'second', 'nanosecond') }
sub is_incomplete { ($_[0]->is_complete) ? 0 : 1 }
> * can_be_datetime / become_datetime
sub can_be_datetime {
my $self = shift;
# -MM-DD HH:MM:SS
# -MM-DD HH:MM:SS.N
return 1 if $self->has('year', 'month', 'day', 'hour',
'minute', 'second');
# -MM-DD HH:MM
# -MM-DD HH:MM::SS, but is already caught
return 1 if $self->has('year', 'month', 'day', 'hour',
'minute') and not $self->has('nanosecond');
# -MM-DD
return 1 if $self->has('year', 'month', 'day')
and not ($self->has('hour')
or $self->has('minute')
or $self->has('second')
or $self->has('nanosecond')
);
return 0;
}
sub become_datetime {
my $self = shift;
return undef unless $self->can_be_datetime;
my $dt = DateTime->new(
year => $self->year,
month => $self->month,
day => $self->day,
);
$dt->set_time_zone( $self->time_zone )
if $self->has('time_zone');
$dt->set( locale => $self->locale )
if $self->has('locale');
$dt->set( hour => $self->hour, minute => $minute )
if $self->has('hour', 'minute');
$dt->set( second => $self->second )
if $self->has('second');
$dt->set( nanosecond => $self->nanosecond )
if $self->has('nanosecond');
$self = $dt; DOES THAT DO THE TRICK, OR
IS IT MORE COMPLEX?
return $self;
}
Above code passes all tests on the perl installed in
my head. Havn't tried any of it with the much fussier
software version.
perl -v
This is perl, v5.8.0 built for ricks-brain-1.0
Re: Strptime issues
Rick Measham wrote:
>
> If so, maybe a method inside Incomplete would be good:
>if ($dti->can_be_datetime) {
> $dti->become_datetime
>}
I've put this in the TODO.
So this is the current list of proposed DT::Incomplete methods that are
"waiting for votes":
* epoch
$epoch = $dti->epoch if $dti->can_be_datetime;
* has_date / has_time
* keys or defined_fields
returns the list of names of "defined" fields
* join( $dti )
join the "defined" keys of two DT::Incomplete objects
* is_incomplete / is_complete
* can_be_datetime / become_datetime
- Flavio S. Glock
Re: Strptime issues
Rick Measham wrote:
>
> I'm adapting Strptime to return DateTime::Incomplete objects when it
> gets an incomplete datetime. The old behaviour was to return the lowest
> possible value.
>
> eg. 'November 2003' used to return 2003-11-01T00:00:00, but will now
> return 2003-11-xxTxx:xx:xx
>
> I'm currently checking to see if a datetime is possible and only
> returning incomplete if not. Is that a good idea?
How about it to be an option in the Strptime constructor?
my $Strp = new DateTime::Format::Strptime(
pattern => '%T',
locale => 'en_AU',
time_zone => 'Melbourne/Australia',
incomplete => 'never' | 'always' | 'maybe'
);
> If so, maybe a method inside Incomplete would be good:
>if ($dti->can_be_datetime) {
> $dti->become_datetime
>}
> or else an incomplete could automatically convert itself to a DateTime
> once it had enough information?!?!?
>
> I figure an Incomplete can become a DateTime if:
>
> We have a Year, Month and Day, but no time (becomes 00:00:00) -or-
> We have a Year, Month, Day, Hour and Minute -or-
> We have a Year, Month, Day, Hour, Minute and Second -or-
> We have a Year, Month, Day, Hour, Minute, Second and Nanosecond
That's ok to me. Do you already have code to do this?
> 3. Keep the current (shonky) behaviour
That's not too bad, except that DT::Incomplete is a bit slower than
DateTime.
- Flavio S. Glock
Re: Strptime issues
Rick Measham wrote:
>
> I'm adapting Strptime to return DateTime::Incomplete objects when it
> gets an incomplete datetime. The old behaviour was to return the lowest
> possible value.
>
> eg. 'November 2003' used to return 2003-11-01T00:00:00, but will now
> return 2003-11-xxTxx:xx:xx
>
> I'm currently checking to see if a datetime is possible and only
> returning incomplete if not. Is that a good idea?
>
> If so, maybe a method inside Incomplete would be good:
>if ($dti->can_be_datetime) {
> $dti->become_datetime
>}
> or else an incomplete could automatically convert itself to a DateTime
> once it had enough information?!?!?
I think automatic conversion is not a good thing. I'm thinking about the
other alternatives...
The patches were applied - thanks.
- Flavio S. Glock
Re: Re: Strptime issues
> David Hood <[EMAIL PROTECTED]> wrote: > Perhaps you should return only the information that is given, in an > iso 8601 > compliant format, so for November 2003 you could simply return > 2003-11. The Nah, that's not going to happen. The entire point of the module is to get a DateTime object. So it's either going to be a full DateTime or a DateTime::Incomplete. Also: What's the ISO format for "11pm November"?
Re: Strptime issues
Perhaps you should return only the information that is given, in an iso 8601
compliant format, so for November 2003 you could simply return 2003-11. The
iso standard is probably useful, especially for things like day of month for
an implied year (--MM-DD) etc. Perhaps format::iso could then be used to
create only an incomplete datetime if there isn't enough information to
create a full datetime object.
Cheers,
David
- Original Message -
From: "Rick Measham" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Monday, November 03, 2003 5:09 PM
Subject: Strptime issues
> I'm adapting Strptime to return DateTime::Incomplete objects when it
> gets an incomplete datetime. The old behaviour was to return the lowest
> possible value.
>
> eg. 'November 2003' used to return 2003-11-01T00:00:00, but will now
> return 2003-11-xxTxx:xx:xx
>
> I'm currently checking to see if a datetime is possible and only
> returning incomplete if not. Is that a good idea?
>
> If so, maybe a method inside Incomplete would be good:
>if ($dti->can_be_datetime) {
> $dti->become_datetime
>}
> or else an incomplete could automatically convert itself to a DateTime
> once it had enough information?!?!?
>
> I figure an Incomplete can become a DateTime if:
>
> We have a Year, Month and Day, but no time (becomes 00:00:00) -or-
> We have a Year, Month, Day, Hour and Minute -or-
> We have a Year, Month, Day, Hour, Minute and Second -or-
> We have a Year, Month, Day, Hour, Minute, Second and Nanosecond
>
> But maybe that's just me.
>
>
> My other option with Strptime are:
>
> 1. Always return an Incomplete object, but I doubt that's what people
> would want/expect.
>
> 2. Always return an Incomplete unless we are given a base, in which case
> the base fills in the blanks. This would mean if you have a full
> datetime specifier in some format you still need to pass a DateTime into
> the strptime object in order to receive a DateTime in return:
>
> $strptime = new DateTime::Format::Strptime( base => DateTime->today,
> pattern=$pattern );
> $strptime->parse_datetime("Nov 3, 2003");
> # Returns DateTime: 2003-11-03T00:00:00
>
> $strptime = new DateTime::Format::Strptime( pattern=$pattern );
> $strptime->parse_datetime("Nov 3, 2003");
> # Returns DateTime::Incomplete: 2003-11-03Txx:xx:xx
>
> 3. Keep the current (shonky) behaviour
>
>
> Cheers!
> Rick
>
>
>
>
>
>
>
>
