factor out filesystem interactions

[dbsrgits/DBIx-Class-DeploymentHandler.git] / lib / DBIx / Class / DeploymentHandler / DeployMethod / SQL / Translator.pm

package DBIx::Class::DeploymentHandler::DeployMethod::SQL::Translator;
use Moose;

# ABSTRACT: Manage your SQL and Perl migrations in nicely laid out directories

use autodie;
use Carp qw( carp croak );
use DBIx::Class::DeploymentHandler::Logger;
use Log::Contextual qw(:log :dlog), -package_logger =>
  DBIx::Class::DeploymentHandler::Logger->new({
    env_prefix => 'DBICDH'
  });

use Method::Signatures::Simple;
use Try::Tiny;

use SQL::Translator;
require SQL::Translator::Diff;

require DBIx::Class::Storage;   # loaded for type constraint
use DBIx::Class::DeploymentHandler::Types;

use File::Path 'mkpath';
use File::Spec::Functions;

with 'DBIx::Class::DeploymentHandler::HandlesDeploy';

has ignore_ddl => (
  isa      => 'Bool',
  is       => 'ro',
  default  => undef,
);

has force_overwrite => (
  isa      => 'Bool',
  is       => 'ro',
  default  => undef,
);

has schema => (
  isa      => 'DBIx::Class::Schema',
  is       => 'ro',
  required => 1,
);

has storage => (
  isa        => 'DBIx::Class::Storage',
  is         => 'ro',
  lazy_build => 1,
);

method _build_storage {
  my $s = $self->schema->storage;
  $s->_determine_driver;
  $s
}

has sql_translator_args => (
  isa => 'HashRef',
  is  => 'ro',
  default => sub { {} },
);

has script_directory => (
  isa      => 'Str',
  is       => 'ro',
  required => 1,
  default  => 'sql',
);

has _filesystem_interface => (
  is         => 'ro',
  lazy_build => 1,
  handles => {
    _ddl_initialize_consume_filenames            => '_ddl_initialize_consume_filenames',
    _ddl_schema_consume_filenames                => '_ddl_schema_consume_filenames',
    _ddl_protoschema_deploy_consume_filenames    => '_ddl_protoschema_deploy_consume_filenames',
    _ddl_protoschema_upgrade_consume_filenames   => '_ddl_protoschema_upgrade_consume_filenames',
    _ddl_protoschema_downgrade_consume_filenames => '_ddl_protoschema_downgrade_consume_filenames',
    _ddl_protoschema_produce_filename            => '_ddl_protoschema_produce_filename',
    _ddl_schema_produce_filename                 => '_ddl_schema_produce_filename',
    _ddl_schema_upgrade_consume_filenames        => '_ddl_schema_upgrade_consume_filenames',
    _ddl_schema_downgrade_consume_filenames      => '_ddl_schema_downgrade_consume_filenames',
    _ddl_schema_upgrade_produce_filename         => '_ddl_schema_upgrade_produce_filename',
    _ddl_schema_downgrade_produce_filename       => '_ddl_schema_downgrade_produce_filename',

    _read_sql_file                               => '_read_sql_file',
    _coderefs_per_files                          => '_coderefs_per_files',
    _write_data_string                           => '_write_data_string',
    _write_data_list                             => '_write_data_list',
  },
);

sub _build__filesystem_interface {
   use DBIx::Class::DeploymentHandler::Filesystem;
   DBIx::Class::DeploymentHandler::Filesystem->new(
     script_directory => $_[0]->script_directory,
     ignore_ddl => $_[0]->ignore_ddl
   )
}

has databases => (
  coerce  => 1,
  isa     => 'DBIx::Class::DeploymentHandler::Databases',
  is      => 'ro',
  default => sub { [qw( MySQL SQLite PostgreSQL )] },
);

has txn_wrap => (
  is => 'ro',
  isa => 'Bool',
  default => 1,
);

has schema_version => (
  is => 'ro',
  isa => 'Str',
  lazy_build => 1,
);

# this will probably never get called as the DBICDH
# will be passing down a schema_version normally, which
# is built the same way, but we leave this in place
method _build_schema_version { $self->schema->schema_version }

method _run_sql_array($sql) {
  my $storage = $self->storage;

  $sql = [grep {
    $_ && # remove blank lines
    !/^(BEGIN|BEGIN TRANSACTION|COMMIT)/ # strip txn's
  } map {
    s/^\s+//; s/\s+$//; # trim whitespace
    join '', grep { !/^--/ } split /\n/ # remove comments
  } @$sql];

  Dlog_trace { "Running SQL $_" } $sql;
  foreach my $line (@{$sql}) {
    $storage->_query_start($line);
    # the whole reason we do this is so that we can see the line that was run
    try {
      $storage->dbh_do (sub { $_[1]->do($line) });
    }
    catch {
      die "$_ (running line '$line')"
    };
    $storage->_query_end($line);
  }
  return join "\n", @$sql
}

method _run_sql($filename) {
  log_debug { "Running SQL from $filename" };
  return $self->_run_sql_array($self->_read_sql_file($filename));
}

method _run_perl($filename, $versions) {
  log_debug { "Running Perl from $filename" };
  my $filedata = do { local( @ARGV, $/ ) = $filename; <> };

  no warnings 'redefine';
  my $fn = eval "$filedata";
  use warnings;
  Dlog_trace { "Running Perl $_" } $fn;

  if ($@) {
    carp "$filename failed to compile: $@";
  } elsif (ref $fn eq 'CODE') {
    $fn->($self->schema, $versions)
  } else {
    carp "$filename should define an anonymouse sub that takes a schema but it didn't!";
  }
}

method _run_sql_and_perl($filenames, $sql_to_run, $versions) {
  my @files   = @{$filenames};
  my $guard   = $self->schema->txn_scope_guard if $self->txn_wrap;

  $self->_run_sql_array($sql_to_run) if $self->ignore_ddl;

  my $sql = ($sql_to_run)?join ";\n", @$sql_to_run:'';
  FILENAME:
  for my $filename (@files) {
    if ($self->ignore_ddl && $filename =~ /^[^_]*-auto.*\.sql$/) {
      next FILENAME
    } elsif ($filename =~ /\.sql$/) {
       $sql .= $self->_run_sql($filename)
    } elsif ( $filename =~ /\.pl$/ ) {
       $self->_run_perl($filename, $versions)
    } else {
      croak "A file ($filename) got to deploy that wasn't sql or perl!";
    }
  }

  $guard->commit if $self->txn_wrap;

  return $sql;
}

sub deploy {
  my $self = shift;
  my $version = (shift @_ || {})->{version} || $self->schema_version;
  log_info { "deploying version $version" };
  my $sqlt_type = $self->storage->sqlt_type;
  my $sql;
  if ($self->ignore_ddl) {
     $sql = $self->_sql_from_yaml({},
       '_ddl_protoschema_deploy_consume_filenames', $sqlt_type
     );
  }
  return $self->_run_sql_and_perl($self->_ddl_schema_consume_filenames(
    $sqlt_type,
    $version,
  ), $sql, [$version]);
}

sub initialize {
  my $self         = shift;
  my $args         = shift;
  my $version      = $args->{version}      || $self->schema_version;
  log_info { "initializing version $version" };
  my $storage_type = $args->{storage_type} || $self->storage->sqlt_type;

  my @files = @{$self->_ddl_initialize_consume_filenames(
    $storage_type,
    $version,
  )};

  for my $filename (@files) {
    # We ignore sql for now (till I figure out what to do with it)
    if ( $filename =~ /^(.+)\.pl$/ ) {
      my $filedata = do { local( @ARGV, $/ ) = $filename; <> };

      no warnings 'redefine';
      my $fn = eval "$filedata";
      use warnings;

      if ($@) {
        carp "$filename failed to compile: $@";
      } elsif (ref $fn eq 'CODE') {
        $fn->()
      } else {
        carp "$filename should define an anonymous sub but it didn't!";
      }
    } else {
      croak "A file ($filename) got to initialize_scripts that wasn't sql or perl!";
    }
  }
}

method _sqldiff_from_yaml($from_version, $to_version, $db, $direction) {
  my $dir       = $self->script_directory;
  my $sqltargs = {
    add_drop_table => 1,
    ignore_constraint_names => 1,
    ignore_index_names => 1,
    %{$self->sql_translator_args}
  };

  my $source_schema;
  {
    my $prefilename = $self->_ddl_protoschema_produce_filename($from_version, $dir);

    # should probably be a croak
    carp("No previous schema file found ($prefilename)")
       unless -e $prefilename;

    my $t = SQL::Translator->new({
       %{$sqltargs},
       debug => 0,
       trace => 0,
       parser => 'SQL::Translator::Parser::YAML',
    });

    my $out = $t->translate( $prefilename )
      or croak($t->error);

    $source_schema = $t->schema;

    $source_schema->name( $prefilename )
      unless  $source_schema->name;
  }

  my $dest_schema;
  {
    my $filename = $self->_ddl_protoschema_produce_filename($to_version, $dir);

    # should probably be a croak
    carp("No next schema file found ($filename)")
       unless -e $filename;

    my $t = SQL::Translator->new({
       %{$sqltargs},
       debug => 0,
       trace => 0,
       parser => 'SQL::Translator::Parser::YAML',
    });

    my $out = $t->translate( $filename )
      or croak($t->error);

    $dest_schema = $t->schema;

    $dest_schema->name( $filename )
      unless $dest_schema->name;
  }

  my $transform_files_method =  "_ddl_protoschema_${direction}_consume_filenames";
  my $transforms = $self->_coderefs_per_files(
    $self->$transform_files_method([$from_version, $to_version])
  );
  $_->($source_schema, $dest_schema) for @$transforms;

  return [SQL::Translator::Diff::schema_diff(
     $source_schema, $db,
     $dest_schema,   $db,
     $sqltargs
  )];
}

method _sql_from_yaml($sqltargs, $from_file, $db) {
  my $schema    = $self->schema;
  my $version   = $self->schema_version;

  my @sql;

  my $actual_file = $self->$from_file($version);
  for my $yaml_filename (@{
     DlogS_trace { "generating SQL from Serialized SQL Files: $_" }
        (ref $actual_file?$actual_file:[$actual_file])
  }) {
     my $sqlt = SQL::Translator->new({
       add_drop_table          => 0,
       parser                  => 'SQL::Translator::Parser::YAML',
       %{$sqltargs},
       producer => $db,
     });

     push @sql, $sqlt->translate($yaml_filename);
     if(!@sql) {
       carp("Failed to translate to $db, skipping. (" . $sqlt->error . ")");
       return undef;
     }
  }
  return \@sql;
}

sub _prepare_install {
  my $self      = shift;
  my $sqltargs  = { %{$self->sql_translator_args}, %{shift @_} };
  my $from_file = shift;
  my $to_file   = shift;
  my $dir       = $self->script_directory;
  my $databases = $self->databases;
  my $version   = $self->schema_version;

  foreach my $db (@$databases) {
    my $sql = $self->_sql_from_yaml($sqltargs, $from_file, $db ) or next;

    my $filename = $self->$to_file($db, $version, $dir);
    if (-e $filename ) {
      if ($self->force_overwrite) {
         carp "Overwriting existing DDL file - $filename";
         unlink $filename;
      } else {
         die "Cannot overwrite '$filename', either enable force_overwrite or delete it"
      }
    }

    $self->_write_data_list($filename, $sql);
  }
}

sub _resultsource_install_filename {
  my ($self, $source_name) = @_;
  return sub {
    my ($self, $type, $version) = @_;
    my $dirname = catfile( $self->script_directory, $type, 'deploy', $version );
    mkpath($dirname) unless -d $dirname;

    return catfile( $dirname, "001-auto-$source_name.sql" );
  }
}

sub _resultsource_protoschema_filename {
  my ($self, $source_name) = @_;
  return sub {
    my ($self, $version) = @_;
    my $dirname = catfile( $self->script_directory, '_source', 'deploy', $version );
    mkpath($dirname) unless -d $dirname;

    return catfile( $dirname, "001-auto-$source_name.yml" );
  }
}

sub install_resultsource {
  my ($self, $args) = @_;
  my $source          = $args->{result_source}
    or die 'result_source must be passed to install_resultsource';
  my $version         = $args->{version}
    or die 'version must be passed to install_resultsource';
  log_info { 'installing_resultsource ' . $source->source_name . ", version $version" };
  my $rs_install_file =
    $self->_resultsource_install_filename($source->source_name);

  my $files = [
     $self->$rs_install_file(
      $self->storage->sqlt_type,
      $version,
    )
  ];
  $self->_run_sql_and_perl($files, '', [$version]);
}

sub prepare_resultsource_install {
  my $self = shift;
  my $source = (shift @_)->{result_source};
  log_info { 'preparing install for resultsource ' . $source->source_name };

  my $install_filename = $self->_resultsource_install_filename($source->source_name);
  my $proto_filename = $self->_resultsource_protoschema_filename($source->source_name);
  $self->prepare_protoschema({
      parser_args => { sources => [$source->source_name], }
  }, $proto_filename);
  $self->_prepare_install({}, $proto_filename, $install_filename);
}

sub prepare_deploy {
  log_info { 'preparing deploy' };
  my $self = shift;
  $self->prepare_protoschema({
      # Exclude __VERSION so that it gets installed separately
      parser_args => { sources => [grep { $_ ne '__VERSION' } $self->schema->sources], }
  }, '_ddl_protoschema_produce_filename');
  $self->_prepare_install({}, '_ddl_protoschema_produce_filename', '_ddl_schema_produce_filename');
}

sub prepare_upgrade {
  my ($self, $args) = @_;
  log_info {
     "preparing upgrade from $args->{from_version} to $args->{to_version}"
  };
  $self->_prepare_changegrade(
    $args->{from_version}, $args->{to_version}, $args->{version_set}, 'upgrade'
  );
}

sub prepare_downgrade {
  my ($self, $args) = @_;
  log_info {
     "preparing downgrade from $args->{from_version} to $args->{to_version}"
  };
  $self->_prepare_changegrade(
    $args->{from_version}, $args->{to_version}, $args->{version_set}, 'downgrade'
  );
}

method _prepare_changegrade($from_version, $to_version, $version_set, $direction) {
  my $schema    = $self->schema;
  my $databases = $self->databases;
  my $dir       = $self->script_directory;

  my $schema_version = $self->schema_version;
  my $diff_file_method = "_ddl_schema_${direction}_produce_filename";
  foreach my $db (@$databases) {
    my $diff_file = $self->$diff_file_method($db, $version_set, $dir );
    if(-e $diff_file) {
      if ($self->force_overwrite) {
         carp("Overwriting existing $direction-diff file - $diff_file");
         unlink $diff_file;
      } else {
         die "Cannot overwrite '$diff_file', either enable force_overwrite or delete it"
      }
    }

    $self->_write_data_list($diff_file,
      $self->_sqldiff_from_yaml($from_version, $to_version, $db, $direction)
    );
  }
}

sub downgrade_single_step {
  my $self = shift;
  my $version_set = (shift @_)->{version_set};
  Dlog_info { "downgrade_single_step'ing $_" } $version_set;

  my $sqlt_type = $self->storage->sqlt_type;
  my $sql_to_run;
  if ($self->ignore_ddl) {
     $sql_to_run = $self->_sqldiff_from_yaml(
       $version_set->[0], $version_set->[1], $sqlt_type, 'downgrade',
     );
  }
  my $sql = $self->_run_sql_and_perl($self->_ddl_schema_downgrade_consume_filenames(
    $sqlt_type,
    $version_set,
  ), $sql_to_run, $version_set);

  return ['', $sql];
}

sub upgrade_single_step {
  my $self = shift;
  my $version_set = (shift @_)->{version_set};
  Dlog_info { "upgrade_single_step'ing $_" } $version_set;

  my $sqlt_type = $self->storage->sqlt_type;
  my $sql_to_run;
  if ($self->ignore_ddl) {
     $sql_to_run = $self->_sqldiff_from_yaml(
       $version_set->[0], $version_set->[1], $sqlt_type, 'upgrade',
     );
  }
  my $sql = $self->_run_sql_and_perl($self->_ddl_schema_upgrade_consume_filenames(
    $sqlt_type,
    $version_set,
  ), $sql_to_run, $version_set);
  return ['', $sql];
}

sub prepare_protoschema {
  my $self      = shift;
  my $sqltargs  = { %{$self->sql_translator_args}, %{shift @_} };
  my $to_file   = shift;
  my $filename
    = $self->$to_file($self->schema_version);

  # we do this because the code that uses this sets parser args,
  # so we just need to merge in the package
  $sqltargs->{parser_args}{package} = $self->schema;
  my $sqlt = SQL::Translator->new({
    parser                  => 'SQL::Translator::Parser::DBIx::Class',
    producer                => 'SQL::Translator::Producer::YAML',
    %{ $sqltargs },
  });

  my $yml = $sqlt->translate;

  croak("Failed to translate to YAML: " . $sqlt->error)
    unless $yml;

  if (-e $filename ) {
    if ($self->force_overwrite) {
       carp "Overwriting existing DDL-YML file - $filename";
       unlink $filename;
    } else {
       die "Cannot overwrite '$filename', either enable force_overwrite or delete it"
    }
  }

  $self->_write_data_string($filename, $yml);
}

__PACKAGE__->meta->make_immutable;

1;

# vim: ts=2 sw=2 expandtab

__END__

=head1 DESCRIPTION

This class is the meat of L<DBIx::Class::DeploymentHandler>.  It takes care
of generating serialized schemata  as well as sql files to move from one
version of a schema to the rest.  One of the hallmark features of this class
is that it allows for multiple sql files for deploy and upgrade, allowing
developers to fine tune deployment.  In addition it also allows for perl
files to be run at any stage of the process.

For basic usage see L<DBIx::Class::DeploymentHandler::HandlesDeploy>.  What's
documented here is extra fun stuff or private methods.

=head1 DIRECTORY LAYOUT

Arguably this is the best feature of L<DBIx::Class::DeploymentHandler>.
It's spiritually based upon L<DBIx::Migration::Directories>, but has a
lot of extensions and modifications, so even if you are familiar with it,
please read this.  I feel like the best way to describe the layout is with
the following example:

 $sql_migration_dir
 |- _source
 |  |- deploy
 |     |- 1
 |     |  `- 001-auto.yml
 |     |- 2
 |     |  `- 001-auto.yml
 |     `- 3
 |        `- 001-auto.yml
 |- SQLite
 |  |- downgrade
 |  |  `- 2-1
 |  |     `- 001-auto.sql
 |  |- deploy
 |  |  `- 1
 |  |     `- 001-auto.sql
 |  `- upgrade
 |     |- 1-2
 |     |  `- 001-auto.sql
 |     `- 2-3
 |        `- 001-auto.sql
 |- _common
 |  |- downgrade
 |  |  `- 2-1
 |  |     `- 002-remove-customers.pl
 |  `- upgrade
 |     `- 1-2
 |     |  `- 002-generate-customers.pl
 |     `- _any
 |        `- 999-bump-action.pl
 `- MySQL
    |- downgrade
    |  `- 2-1
    |     `- 001-auto.sql
    |- initialize
    |  `- 1
    |     |- 001-create_database.pl
    |     `- 002-create_users_and_permissions.pl
    |- deploy
    |  `- 1
    |     `- 001-auto.sql
    `- upgrade
       `- 1-2
          `- 001-auto.sql

So basically, the code

 $dm->deploy(1)

on an C<SQLite> database that would simply run
C<$sql_migration_dir/SQLite/deploy/1/001-auto.sql>.  Next,

 $dm->upgrade_single_step([1,2])

would run C<$sql_migration_dir/SQLite/upgrade/1-2/001-auto.sql> followed by
C<$sql_migration_dir/_common/upgrade/1-2/002-generate-customers.pl>, and
finally punctuated by
C<$sql_migration_dir/_common/upgrade/_any/999-bump-action.pl>.

C<.pl> files don't have to be in the C<_common> directory, but most of the time
they should be, because perl scripts are generally database independent.

Note that unlike most steps in the process, C<initialize> will not run SQL, as
there may not even be an database at initialize time.  It will run perl scripts
just like the other steps in the process, but nothing is passed to them.
Until people have used this more it will remain freeform, but a recommended use
of initialize is to have it prompt for username and password, and then call the
appropriate C<< CREATE DATABASE >> commands etc.

=head2 Directory Specification

The following subdirectories are recognized by this DeployMethod:

=over 2

=item C<_source> This directory can contain the following directories:

=over 2

=item C<deploy> This directory merely contains directories named after schema
versions, which in turn contain C<yaml> files that are serialized versions
of the schema at that version.  These files are not for editing by hand.

=back

=item C<_preprocess_schema> This directory can contain the following
directories:

=over 2

=item C<downgrade> This directory merely contains directories named after
migrations, which are of the form C<$from_version-$to_version>.  Inside of
these directories you may put Perl scripts which are to return a subref
that takes the arguments C<< $from_schema, $to_schema >>, which are
L<SQL::Translator::Schema> objects.

=item C<upgrade> This directory merely contains directories named after
migrations, which are of the form C<$from_version-$to_version>.  Inside of
these directories you may put Perl scripts which are to return a subref
that takes the arguments C<< $from_schema, $to_schema >>, which are
L<SQL::Translator::Schema> objects.

=back

=item C<$storage_type> This is a set of scripts that gets run depending on what
your storage type is.  If you are not sure what your storage type is, take a
look at the producers listed for L<SQL::Translator>.  Also note, C<_common>
is a special case.  C<_common> will get merged into whatever other files you
already have.  This directory can contain the following directories itself:

=over 2

=item C<initialize> Gets run before the C<deploy> is C<deploy>ed.  Has the
same structure as the C<deploy> subdirectory as well; that is, it has a
directory for each schema version.  Unlike C<deploy>, C<upgrade>, and C<downgrade>
though, it can only run C<.pl> files, and the coderef in the perl files get
no arguments passed to them.

=item C<deploy> Gets run when the schema is C<deploy>ed.  Structure is a
directory per schema version, and then files are merged with C<_common> and run
in filename order.  C<.sql> files are merely run, as expected.  C<.pl> files are
run according to L</PERL SCRIPTS>.

=item C<upgrade> Gets run when the schema is C<upgrade>d.  Structure is a directory
per upgrade step, (for example, C<1-2> for upgrading from version 1 to version
2,) and then files are merged with C<_common> and run in filename order.
C<.sql> files are merely run, as expected.  C<.pl> files are run according
to L</PERL SCRIPTS>.

=item C<downgrade> Gets run when the schema is C<downgrade>d.  Structure is a directory
per downgrade step, (for example, C<2-1> for downgrading from version 2 to version
1,) and then files are merged with C<_common> and run in filename order.
C<.sql> files are merely run, as expected.  C<.pl> files are run according
to L</PERL SCRIPTS>.


=back

=back

Note that there can be an C<_any> in the place of any of the versions (like
C<1-2> or C<1>), which means those scripts will be run B<every> time.  So if
you have an C<_any> in C<_common/upgrade>, that script will get run for every
upgrade.

=head1 PERL SCRIPTS

A perl script for this tool is very simple.  It merely needs to contain an
anonymous sub that takes a L<DBIx::Class::Schema> and the version set as it's
arguments.

A very basic perl script might look like:

 #!perl

 use strict;
 use warnings;

 sub {
   my $schema = shift;

   # [1] for deploy, [1,2] for upgrade or downgrade, probably used with _any
   my $versions = shift;

   $schema->resultset('Users')->create({
     name => 'root',
     password => 'root',
   })
 }

=attr ignore_ddl

This attribute will, when set to true (default is false), cause the DM to use
L<SQL::Translator> to use the C<_source>'s serialized SQL::Translator::Schema
instead of any pregenerated SQL.  If you have a development server this is
probably the best plan of action as you will not be putting as many generated
files in your version control.  Goes well with with C<databases> of C<[]>.

=attr force_overwrite

When this attribute is true generated files will be overwritten when the
methods which create such files are run again.  The default is false, in which
case the program will die with a message saying which file needs to be deleted.

=attr schema

The L<DBIx::Class::Schema> (B<required>) that is used to talk to the database
and generate the DDL.

=attr storage

The L<DBIx::Class::Storage> that is I<actually> used to talk to the database
and generate the DDL.  This is automatically created with L</_build_storage>.

=attr sql_translator_args

The arguments that get passed to L<SQL::Translator> when it's used.

=attr script_directory

The directory (default C<'sql'>) that scripts are stored in

=attr databases

The types of databases (default C<< [qw( MySQL SQLite PostgreSQL )] >>) to
generate files for

=attr txn_wrap

Set to true (which is the default) to wrap all upgrades and deploys in a single
transaction.

=attr schema_version

The version the schema on your harddrive is at.  Defaults to
C<< $self->schema->schema_version >>.