DBIx::Class has_many join on substring of a column - perl

I have two class that are linked on
foreign.weirdkey => substr(self.key, 1, 9)
...and cannot for the life of me figure out how to construct the has_many call to indicate this.
The underlying database (a set of Oracle tables) defines no foreign keys, is fixed, and is outside of my control.
I've been through the docs and can't seem to find a syntax that will work within the confines of a manual has_many definition.
Any help would be much appreciated.

something like this should work:
__PACKAGE__->has_many( baubles => 'My::Schema::Result::Thing', sub {
my $args = shift;
return ({
"$args->{foreign_alias}.weirdkey" => \"substr($args->{self_alias}.key, 1, 9)",
},
$args->{self_rowobj} && {
"$args->{foreign_alias}.weirdkey" => substr($args->{self_rowobj}->key, 1, 9)
})
});
Note that I use perl's substr if you have the current row object, so then the join will collapse into merely a where clause and won't use the database for the string munging. Remove that or fiddle with it if you have problems. DBIC_TRACE should make it clear what goes on.
Documentation here: https://metacpan.org/pod/DBIx::Class::Relationship::Base#condition

Related

DbiX::Class / Creating tree of test data without persisting

I'm using DBIx::Class as or mapper for my Perl project. When it comes to generating test data I'm using DBIx::Class::ResultSet::new to create new entities in memory. In order to link entities with relationships I use set_from_related. This works absolutely flawless until I try to set the value(s) for a has_many relationship. Pseudo example:
# Table 'AUTHOR' has
# one-to-one (belongs_to) relationship named 'country' to table 'COUNTRY'
# one-to-many (has_many) relationship named 'books' to table 'BOOK'
my $s = Schema::getSchema();
my $author = $s->resultset('Author')->new({ name => 'Jon Doe', year_of_birth => 1982 });
my $country = $s->resultset('Country')->new({ name => 'Germany', iso_3166_code => 'DE' });
my $book = $s->resultset('Book')->new({ title => 'A star far away', publishing_year => 2002 });
# Now let's make 'em known to each other
$author->set_from_related('country', $country);
$author->set_from_related('books', $book);
# At this point
# $author->country is defined
# $author->books->first is undef <<<---- Problem
I cannot find a suitable method in the DBIx::Class::Relationship::Base documentation. The one closest to what I need is add_to_$rel but this method creates (persists) the entities. This is not an option for me as some of the entites used in my project don't belong to me (no write permission).
Does anyone have an idea how to add entities in memory for a has_many relationship ?
That's not possible as newly created result objects don't have their primary key column(s) populated until they are persisted to the database.
What you possibly want is to use multi-create and run that inside a transaction.

Runtime error: MongoDB::DatabaseError: bad hint

I am trying to call the hint method on a MongoDB::Cursor object. However, it throwing an exception when it's trying to execute the query. See the code sample below:
sub some_method_which_returns_cursor {
my $cursor = $collection->find($filter);
if ($hint) {
$cursor->hint({‘some_index’ => 1}); #failing here.
}
if ($sort) {
$cursor->sort($sort);
}
return $cursor;
}
Any thoughts as to what's going on and how I can fix this?
Harish asked me via email and I'll repeat my answer here for posterity:
The hint method takes a string when given an index name, or an array reference when given keys/order pairs:
$cursor->hint("some_index"); # by name
$cursor->hint([field1 => 1, field2 => -1]); # by keys
It also takes a hash reference, but don't use that because modern Perls randomize key order when serializing, so your hint may not match an index.

Traversing hashes in Perl based on config

I need to be able to pluck specific values from data that I receive from different 3rd parties. The data can structured differently depending on the 3rd party. For example:
my $first =
{
email => "joe\#example.com",
firstname => "Joe",
lastname => "Regular",
};
my $second =
{
user => {
e-mail => "joe\#example.com",
firstName => "Joe",
lastName => "Regular",
}
};
I know what the data structure will be for each 3rd party, so I can define that as config. What I want to end up with is
my $email = _magic($first_config,$first);
my $other_email = _magic($second_config,$second);
Any ideas much appreciated.
Build a look-up table. And you can use a dispatch table, hash with values being code references, so that when a party-identification is used as the key the code for that party executes
my %get_value = ( first => \&fetch_first, second => \&fetch_second );
my $party = 'first'; # input via command-line options, STDIN ...
my $email = $get_value{$party}->();
where \&fetch_first is a reference to the subroutine fetch_first. You can also enter it directly, first => sub { ... }, suitable for simple code. See perlreftut, perlref, and perlsub.
There are many ways to carry data in your program, and so to implement the lookup itself.
Here is an illustration, built in steps. It uses the (confirmed) fact that the data is in valid Perl data structures, and for simplicity it specifies the data right in each sub.
sub fetch_first {
my $data = {
email => '...',
firstName => '...',
};
return $data->{email};
}
This only delivers the email address, but we can do better.
Once you dereference a code reference you can also pass arguments
my $first_name = $get_value{$party}->('firstName');
where the subs are now written to use this input to return the required field
sub fetch_first {
my ($query) = #_;
my $data = {
email => '...',
firstName => '...',
};
return $data->{$query};
}
A big weakness of the above is that the calling code must use valid names of keys, so it needs to know the details of implementation of what it is using.
This can be improved, for example by choosing an interface for the call which is then translated in the subs into key names (or via yet another look-up structure). Then you make calls such as
my $email = $get_value{$party}->('email'); # or: 'first', 'last'
and somewhere you have association first => 'firstName' (etc) which subs can look up.
The flexibility is greatly helped by data being set up in a consistent way. The whole thing can also be quite maintainable if the code is organized thoughtfully.
If this grows more complex the solution is to write a class. Then you can build a very nice system.

Perl Catalyst: Resultset and Relations

I have two tables in my database and one of the tables is associated with my Accounts table.
So in my Schema Result for Account.pm I added the following line.
__PACKAGE__->has_many('subjects', 'MyApp::DBIC::Schema::Subject', {'foreight.account_id' => 'self.account_id'});
Then in my controller I make a search like this.
$c->stash->{search_results} = $c->model('DB::Account')->search(
{ -or => [
firstname => {like => '%'.$search_term.'%'},
'subjects.subject_title' => {like => '%'.$search_term.'%'},
]
},
{
join => 'subjects',
rows => '3',
},
{
order_by => 'first name ASC',
page => 1,
rows => 10,
}
);
It does not output any errors, but I can't figure out how to output the results on my view file. Is this a correct method of making relations between two tables?
My goal: provided a search_term, search two tables and output the result in view file. My SQL would look something like this:
SELECT FROM Accounts,Subjects WHERE Accounts.firstname=$search_term OR Subjects.subject_title=$search_term LEFT JOIN Subjects ON Accounts.account_id=Subject.account_id
And would want to output the result in view file, as I stated above.
I am fairly new to Perl and some of the documentations don't make that much sense to me, still. So any help and tips are appreciated.
The join looks OK to me, but it would make sense to try a simplified version without the join to check that everything else is OK.
The behaviour of DBIx::Class::ResultSet::search differs depending on the context in which it's called. If it's called in list context then it executes the database query and returns an array of MyApp::DBIC::Schema::Account objects. For example:
my #accounts = $c->model('DB::Account')->search();
In your case you're calling search in scalar context, which means that rather than returning an array it will return a DBIx::Class::ResultSet object (or a subclass thereof), and crucially it won't actually execute a db query. For that to happen you need to call the all method on your resultset. So, assuming you're using the default template toolkit view you probably want something like this:
[% FOREACH search_result IN search_results.all %]
[% search_result.first_name %]
[% END %]
This 'lazy' behaviour of DBIx::Class is actually very useful, and in my opinion somewhat undersold in the documentation. It means you can keep a resultset in a variable and keep executing different search calls on it without actually hitting the DB, it can allow much nicer code in cases where you want to conditionally build up a complex query. See the DBIx::Class::Resultset documentation for further details.
You have error in your query:
Try:
$c->stash->{search_results} = $c->model('DB::Account')->search(
{ -or => [
firstname => {like => '%'.$search_term.'%'},
'subjects.subject_title' => {like => '%'.$search_term.'%'},
]
},
{
join => 'subjects',
order_by => 'firstname ASC',
page => 1,
rows => 10,
}
);

MongoID and multikey query problem

I have a model Item with an indexed field named _key, that is array of strings (keywords for search).
Now I need to do autocompletion for this model (through JSON) in another form, and the problem is that instead of exact search by all words input by user, I need to do exact search by all but one last word. So I made this scopes in this model:
scope :find_by_keywords, lambda { |keys| where(:_keys.all => keys) }
scope :for_autocomplete, lambda { |keys| where(:_keys.all => keys[0..-2], :_keys => /^#{keys[-1]}/i ) }
the first scope for exact search works well, but I have problems with second scope for autocomplete. MongoID optimises (or something like) this query, so it becomes
db_development['items'].find({:_keys=>/^qwer/i}, {})
i.e. it allways misses the first condition. It's not surprising, because it needs different criterias on field for different conditions.
So I have tried many-many options. Different combinations of .all and .in, separate to different 'wheres', 'all_in' method, 'find(:conditions => ...)' and so on. Could you please suggest, how I can do this job?
I've found such solution:
scope :for_autocomplete, lambda { |keys| where(:_keys.all => keys[0..-2]+ [ /^#{keys[-1]}/ ] ) }
Seems to be working.