package Catmandu::Store::DBI; use Catmandu::Sane; use Catmandu::Util qw(require_package); use DBI; use Catmandu::Store::DBI::Bag; use Moo; use MooX::Aliases; use Catmandu::Error; use namespace::clean; our $VERSION = "0.13"; with 'Catmandu::Store'; with 'Catmandu::Transactional'; has data_source => ( is => 'ro', required => 1, alias => 'dsn', trigger => sub { my $ds = $_[0]->{data_source}; $ds = $ds =~ /^DBI:/i ? $ds : "DBI:$ds"; $_[0]->{data_source} = $ds; }, ); has username => (is => 'ro', default => sub {''}, alias => 'user'); has password => (is => 'ro', default => sub {''}, alias => 'pass'); has default_order => (is => 'ro', default => sub {'ID'}); has handler => (is => 'lazy'); has _in_transaction => (is => 'rw', writer => '_set_in_transaction',); has _dbh => (is => 'lazy', builder => '_build_dbh', writer => '_set_dbh',); # DEPRECATED methods. Were only invented to tackle of problem of reconnection sub timeout { warn "method timeout has been replaced by auto reconnect"; } sub has_timeout { warn "method has_timeout has been replaced by auto reconnect"; 0; } sub reconnect_after_timeout { warn "method reconnect_after_timeout has been replaced by auto reconnect"; } sub handler_namespace { 'Catmandu::Store::DBI::Handler'; } sub _build_handler { my ($self) = @_; my $driver = $self->dbh->{Driver}{Name} // ''; my $ns = $self->handler_namespace; my $pkg; if ($driver =~ /pg/i) { $pkg = 'Pg'; } elsif ($driver =~ /sqlite/i) { $pkg = 'SQLite'; } elsif ($driver =~ /mysql/i) { $pkg = 'MySQL'; } else { Catmandu::NotImplemented->throw( 'Only Pg, SQLite and MySQL are supported.'); } require_package($pkg, $ns)->new; } sub _build_dbh { my ($self) = @_; my $opts = { AutoCommit => 1, RaiseError => 1, mysql_auto_reconnect => 1, mysql_enable_utf8 => 1, pg_utf8_strings => 1, sqlite_use_immediate_transaction => 1, sqlite_unicode => 1, }; my $dbh = DBI->connect($self->data_source, $self->username, $self->password, $opts,); $dbh; } sub dbh { my $self = $_[0]; my $dbh = $self->_dbh; # reconnect when dbh is not set (should never happen) return $self->reconnect unless defined $dbh; # check validity of dbh # for performance reasons only check every second if ( defined( $self->{last_ping_t} ) ) { return $dbh if (time - $self->{last_ping_t}) < 1; } $self->{last_ping_t} = time; return $dbh if $dbh->ping; # one should never reconnect to a database during a transaction # because that would initiate a new transaction Catmandu::Error->throw("Connection to DBI backend lost, and cannot reconnect during a transaction") unless $dbh->{AutoCommit}; # reconnect and return dbh # note: mysql_auto_reconnect only works when AutoCommit is 1 $self->reconnect; } sub reconnect { my $self = $_[0]; my $dbh = $self->_dbh; $dbh->disconnect if defined($dbh); $self->_set_dbh($self->_build_dbh); $self->_dbh; } sub transaction { my ($self, $sub) = @_; if ($self->_in_transaction) { return $sub->(); } my $dbh = $self->dbh; my @res; eval { $self->_set_in_transaction(1); $dbh->begin_work; @res = $sub->(); $dbh->commit; $self->_set_in_transaction(0); 1; } or do { my $err = $@; eval {$dbh->rollback}; $self->_set_in_transaction(0); die $err; }; @res; } sub DEMOLISH { my ($self) = @_; $self->{_dbh}->disconnect if $self->{_dbh}; } 1; __END__ =pod =encoding utf8 =head1 NAME Catmandu::Store::DBI - A Catmandu::Store backed by DBI =head1 VERSION Version 0.0424 =head1 SYNOPSIS # From the command line $ catmandu import JSON to DBI --data_source SQLite:mydb.sqlite < data.json # Or via a configuration file $ cat catmandu.yml --- store: mydb: package: DBI options: data_source: "dbi:mysql:database=mydb" username: xyz password: xyz ... $ catmandu import JSON to mydb < data.json $ catmandu export mydb to YAML > data.yml $ catmandu export mydb --id 012E929E-FF44-11E6-B956-AE2804ED5190 to JSON > record.json $ catmandu count mydb $ catmandy delete mydb # From perl use Catmandu::Store::DBI; my $store = Catmandu::Store::DBI->new( data_source => 'DBI:mysql:database=mydb', # prefix "DBI:" optional username => 'xyz', # optional password => 'xyz', # optional ); my $obj1 = $store->bag->add({ name => 'Patrick' }); printf "obj1 stored as %s\n" , $obj1->{_id}; # Force an id in the store my $obj2 = $store->bag->add({ _id => 'test123' , name => 'Nicolas' }); my $obj3 = $store->bag->get('test123'); $store->bag->delete('test123'); $store->bag->delete_all; # All bags are iterators $store->bag->each(sub { ... }); $store->bag->take(10)->each(sub { ... }); =head1 DESCRIPTION A Catmandu::Store::DBI is a Perl package that can store data into DBI backed databases. The database as a whole is a 'store' L. Databases tables are 'bags' (L). Databases need to be preconfigured for accepting Catmandu data. When no specialized Catmandu tables exist in a database then Catmandu will create them automatically. See "DATABASE CONFIGURATION" below. DO NOT USE Catmandu::Store::DBI on an existing database! Tables and data can be deleted and changed. =head1 LIMITATIONS Currently only MySQL, Postgres and SQLite are supported. Text columns are also assumed to be utf-8. =head1 CONFIGURATION =over =item data_source Required. The connection parameters to the database. See L for more information. Examples: dbi:mysql:foobar <= a local mysql database 'foobar' dbi:Pg:dbname=foobar;host=myserver.org;port=5432 <= a remote PostGres database dbi:SQLite:mydb.sqlite <= a local SQLLite file based database mydb.sqlite dbi:Oracle:host=myserver.org;sid=data01 <= a remote Oracle database Drivers for each database need to be available on your computer. Install then with: cpanm DBD::mysql cpanm DBD::Pg cpanm DBD::SQLite =item user Optional. A user name to connect to the database =item password Optional. A password for connecting to the database =item default_order Optional. Default the default sorting of results when returning an iterator. Choose 'ID' to order on the configured identifier field, 'NONE' to skip all ordering, or "$field" where $field is the name of a table column. By default set to 'ID'. =back =head1 DATABASE CONFIGURATION When no tables exists for storing data in the database, then Catmandu will create them. By default tables are created for each L which contain an '_id' and 'data' column. This behavior can be changed with mapping option: my $store = Catmandu::Store::DBI->new( data_source => 'DBI:mysql:database=test', bags => { # books table books => { mapping => { # these keys will be directly mapped to columns # all other keys will be serialized in the data column title => {type => 'string', required => 1, column => 'book_title'}, isbn => {type => 'string', unique => 1}, authors => {type => 'string', array => 1} } } } ); For keys that have a corresponding table column configured, the method 'select' of class L provides a more efficiƫnt way to query records. See L for more information. =head2 Column types =over =item string =item integer =item binary =item datetime Only MySQL, PostgreSQL =item datetime_milli Only MySQL, PostgreSQL =item json Only PostgreSQL This is mapped internally to postgres field of type "jsonb". Please use the serializer L, if you choose to store the perl data structure into this type of field. Reasons: * there are several types of serializers. E.g. serializer "messagepack" produces a string that is not accepted by a jsonb field in postgres * the default serializer L converts the perl data structure to a binary json string, and the DBI client reencodes that utf8 string (because jsonb is a sort of text field), so you end up having a double encoded string. =back =head2 Column options =over =item column Name of the table column if it differs from the key in your data. =item array Boolean option, default is C<0>. Note that this is only supported for PostgreSQL. =item unique Boolean option, default is C<0>. =item index Boolean option, default is C<0>. Ignored if C is true. =item required Boolean option, default is C<0>. =back =head1 AUTO RECONNECT This library automatically connects to the underlying database, and reconnects when that connection is lost. There is one exception though: when the connection is lost in the middle of a transaction, this is skipped and a L is thrown. Reconnecting during a transaction would have returned a new transaction, and (probably?) committed the lost transaction contrary to your expectation. There is actually no way to recover from that, so throwing an error seemed liked to a "good" way to solve that. In order to avoid this situation, try to avoid a big time lap between database actions during a transaction, as your server may have thrown you out. P.S. the mysql option C<< mysql_auto_reconnect >> does NOT automatically reconnect during a transaction exactly for this reason. =head1 SEE ALSO L, L =cut