# PODNAME: DBIx::Class::Helper::ResultSet::DateMethods1::Announce

# ABSTRACT: Original Announcement of DBIx::Class::Helper::ResultSet::DateMethods1

__END__

=pod

=head1 NAME

DBIx::Class::Helper::ResultSet::DateMethods1::Announce - Original Announcement of DBIx::Class::Helper::ResultSet::DateMethods1

=head1 Content

L<A little over three years ago|https://github.com/frioux/dbic-withdates/commit/5e6893d4cb873eca75118061c104bed9b891dee0>
I got inspired while on vacation to Crested Butte, CO and started a branch
in DBIC called merely, "date-ops."  The idea was to allow users to call
various date functions, portably, directly in DBIC.  With some help from
some people who use other databases, I got it working with SQL Server,
SQLite, PostgreSQL, MySQL, and Oracle.

Unfortunately after we finished it (
L<about six months after I started|https://github.com/frioux/dbic-withdates/commit/fcb5b33a58426ee9d4b87fd84144d05d7a1cf9e9>
) it merely languished.  There were some technical issues we never got
around to ironing out, mostly because it wasn't clear to us what the
cost of not taking care of them would be.

Fast forward a few more months and I was working on a greenfield project
at work.  I wanted to do some date math in the database, so far I did all
of my development against SQLite but deployed to SQL Server, and it looked
like the date ops were my solution.  I decided that given that I was the
primary author of them, I could live with deploying them to production.
I did exactly that and had pretty much no problems.  Well, no problems
until I had to upgrade DBIC.  Every time I needed to upgrade DBIC I had
to merge/rebase the branch.  It turned out to be much more work than I
bargained for, and I ended up just never updating DBIC.

At some point (
L<just under a month ago|https://github.com/frioux/DBIx-Class-Helpers/commit/5fefda2e5dafb0b78e9ee5a687fb698899d8d2ff>
) I decided that I
needed to upgrade DBIC and that maintaining these date ops was no longer
tenable.  Armed with three more years of experience than I had when I
started I embarked on converting the date ops to date methods, that would
work as Helpers.  In addition to not being core, so I could release at
my own pace, I could also version the API, so if I end up making some
critical mistakes or needing to break the API for some features in the
future, I can merely release C<::DateMethods2>.  So without further ado:

=head1 Announcing C<DBIx::Class::Helper::ResultSet::DateMethods1>

Do you store dates in your database?  Do you ever want to manipulate them
efficiently?  Well here's your solution!

First, how do you search in a more comprehensible way?

 $rs->dt_on_or_before(
   { -ident => '.when_created' },
   DateTime->now->subtract(days => 7),
 );

C<dt_on_or_before> (as well as C<dt_before>, C<dt_on_or_after>, or C<dt_after>)
merely aliases C<< <= >>, C<< < >>, C<< >= >>, and C<< > >>, respectively.
Instead of trying to think about the numerical meaning of a date on a timeline,
just use these named methods.  In addition to the nicer name, they can take
L<DateTime> object (which are automatically converted to UTC), and autoprepend
L<DBIx::Class::ResultSet/current_source_alias> when passed an C<-ident> that
starts with a C<.>.  You can pass any of a value, a column (via C<-ident>), a
subquery, literal sql, or a C<DateTime> object to either parameter slots of
these methods.

Second, how do I really leverage this module to do stuff with dates in my
database?

Here's a query I originally wrote with date ops.  Basically it groups some
columns by some "date parts" like year, month, day, etc.  You can use it to make
nice reports of things like how many things have been done per month, or maybe
find out if the system is more busy in the summer:

 $rs->search(undef, {
    columns => {
       count => '*',
       year  => $rs->dt_SQL_pluck({ -ident => '.start' }, 'year'),
       month => $rs->dt_SQL_pluck({ -ident => '.start' }, 'month'),
    },
    group_by => [
      $rs->dt_SQL_pluck({ -ident => '.start' }, 'year'),
      $rs->dt_SQL_pluck({ -ident => '.start' }, 'month'),
    ],
 )->hri->all

I use that exact query (though I give the user a UI for which dateparts
to include) in my system, and it works on SQL Server and SQLite, and
it's fast.  Awesome.

Or how about a query to discover how many issues were resolved before the next
full day after their creation?  Check it out:

 # note that 'day', 1 should also work
 $rs->dt_before(
   { -ident => '.resolution' },
   $rs->dt_SQL_add({ -ident => '.creation' }, 'hour', 24),
 )->all

Both of the above queries work on all of the supported datebases!

Third, some little helpers to extend the above.

On top of those things, I also throw in a couple other handy methods.  One,
C<utc> converts a DateTime object to a string, in the UTC timezone.  Hopefully
you shouldn't need it directly, but I've already ended up using it in places
where our code forced me to return a simple hash to get merged into a search
query, instead of letting me call methods on an RS.

Another lagniappe is C<utc_now> which returns some literal sql that resolves to
the current date and time in UTC on your database.  You can pass it in to search
just like you would datetime.  So if your server and your database have in sync
clocks, these would do the same thing:

 $rs->dt_on_or_before(
   { -ident => '.when_created' },
   DateTime->now->subtract(days => 7),
 );

 $rs->dt_on_or_before(
   { -ident => '.when_created' },
   $rs->dt_SQL_add($rs->utc->now, 'day', -7),
 );

(NOTE: many people seem to hold suspect the idea that the clock is correct on a
given server.  If you can't trust the clock of a server, you probably can't
trust the server.  Use NTP.)

And that's it.  I hope you can use and enjoy these helpers!

=head1 SEE ALSO

L<original blog post|https://blog.afoolishmanifesto.com/posts/announcing-helper-resultset-datemethods1/>

=head1 AUTHOR

Arthur Axel "fREW" Schmidt <frioux+cpan@gmail.com>

=head1 COPYRIGHT AND LICENSE

This software is copyright (c) 2020 by Arthur Axel "fREW" Schmidt.

This is free software; you can redistribute it and/or modify it under
the same terms as the Perl 5 programming language system itself.

=cut