| DBD::Mock - Mock database driver for testing |
DBD::Mock - Mock database driver for testing
use DBI;
# connect to your as normal, using 'Mock' as your driver name
my $dbh = DBI->connect( 'DBI:Mock:', '', '' )
|| die "Cannot create handle: $DBI::errstr\n";
# create a statement handle as normal and execute with parameters
my $sth = $dbh->prepare( 'SELECT this, that FROM foo WHERE id = ?' );
$sth->execute( 15 );
# Now query the statement handle as to what has been done with it
my $mock_params = $sth->{mock_params};
print "Used statement: ", $sth->{mock_statement}, "\n",
"Bound parameters: ", join( ', ', @{ $params } ), "\n";
Testing with databases can be tricky. If you are developing a system married to a single database then you can make some assumptions about your environment and ask the user to provide relevant connection information. But if you need to test a framework that uses DBI, particularly a framework that uses different types of persistence schemes, then it may be more useful to simply verify what the framework is trying to do -- ensure the right SQL is generated and that the correct parameters are bound. DBD::Mock makes it easy to just modify your configuration (presumably held outside your code) and just use it instead of DBD::Foo (like the DBD::Pg manpage or the DBD::mysql manpage) in your framework.
There is no distinct area where using this module makes sense. (Some people may successfully argue that this is a solution looking for a problem...) Indeed, if you can assume your users have something like the DBD::AnyData manpage or the DBD::SQLite manpage or if you do not mind creating a dependency on them then it makes far more sense to use these legitimate driver implementations and test your application in the real world -- at least as much of the real world as you can create in your tests...
And if your database handle exists as a package variable or something else easily replaced at test-time then it may make more sense to use the Test::MockObject manpage to create a fully dynamic handle. There is an excellent article by chromatic about using the Test::MockObject manpage in this and other ways, strongly recommended. (See SEE ALSO for a link)
DBD::Mock comprises a set of classes used by DBI to implement a database driver. But instead of connecting to a datasource and manipulating data found there it tracks all the calls made to the database handle and any created statement handles. You can then inspect them to ensure what you wanted to happen actually happened. For instance, say you have a configuration file with your database connection information:
[DBI] dsn = DBI:Pg:dbname=myapp user = foo password = bar
And this file is read in at process startup and the handle stored for other procedures to use:
package ObjectDirectory;
my ( $DBH );
sub run_at_startup {
my ( $class, $config ) = @_;
$config ||= read_configuration( ... );
my $dsn = $config->{DBI}{dsn};
my $user = $config->{DBI}{user};
my $pass = $config->{DBI}{password};
$DBH = DBI->connect( $dsn, $user, $pass ) || die ...;
}
sub get_database_handle {
return $DBH;
}
A procedure might use it like this (ignoring any error handling for the moment):
package My::UserActions;
sub fetch_user {
my ( $class, $login ) = @_;
my $dbh = ObjectDirectory->get_database_handle;
my $sql = q{
SELECT login_name, first_name, last_name, creation_date, num_logins
FROM users
WHERE login_name = ?
};
my $sth = $dbh->prepare( $sql );
$sth->execute( $login );
my $row = $sth->fetchrow_arrayref;
return ( $row ) ? User->new( $row ) : undef;
}
So for the purposes of our tests we just want to ensure that:
Assume whether the SQL actually works or not is irrelevant for this test :-)
To do that our test might look like:
my $config = ObjectDirectory->read_configuration( ... );
$config->{DBI}{dsn} = 'DBI:Mock:';
ObjectDirectory->run_at_startup( $config );
my $login_name = 'foobar';
my $user = My::UserActions->fetch_user( $login_name );
# Get the handle from ObjectDirectory;
# this is the same handle used in the
# 'fetch_user()' procedure above
my $dbh = ObjectDirectory->get_database_handle();
# Ask the database handle for the history
# of all statements executed against it
my $history = $dbh->{mock_all_history};
# Now query that history record to
# see if our expectations match reality
cmp_ok(scalar(@{$history}), '==', 1, 'Correct number of statements executed' ;
my $login_st = $history->[0];
like($login_st->statement,
qr/SELECT login_name.*FROM users WHERE login_name = ?/sm,
'Correct statement generated' );
my $params = $login_st->bound_params;
cmp_ok(scalar(@{$params}), '==', 1, 'Correct number of parameters bound');
is($params->[0], $login_name, 'Correct value for parameter 1' );
# Reset the handle for future operations
$dbh->{mock_clear_history} = 1;
The list of properties and what they return is listed below. But in an overall view:
execute(). It can also contain predefined results for the statement handle to 'fetch', track how many fetches were called and what its current record is.
This may be an incredibly naive implementation of a DBD. But it works for me ...
Since this is a normal DBI statement handle we need to expose our tracking information as properties (accessed like a hash) rather than methods.
1) will not allow DBI to connect. This can be used to simulate a DSN error or authentication failure. This can then be set back to false (0) to resume normal DBI operations. Here is an example of how this works:
# install the DBD::Mock driver
my $drh = DBI->install_driver('Mock');
$drh->{mock_connect_fail} = 1;
# this connection will fail
my $dbh = DBI->connect('dbi:Mock:', '', '') || die "Cannot connect";
# this connection will throw an exception
my $dbh = DBI->connect('dbi:Mock:', '', '', { RaiseError => 1 });
$drh->{mock_connect_fail} = 0;
# this will work now ...
my $dbh = DBI->connect(...);
This feature is conceptually different from the 'mock_can_connect' attribute of the C<$dbh> in that it has a driver-wide scope, where 'mock_can_connect' is handle-wide scope. It also only prevents the initial connection, any C<$dbh> handles created prior to setting 'mock_connect_fail' to true (C<1>) will still go on working just fine.
data_source() method.
DBD::Mock::StatementTrack) objects created against the database handle in the order they were created. Each history object can then report information about the SQL statement used to create it, the bound parameters, etc..
DBD::Mock::StatementTrack::Iterator object which will iterate through the current set of DBD::Mock::StatementTrack object in the history. See the DBD::Mock::StatementTrack::Iterator documentation below for more information.
my $dbh = get_handle( ... );
my $sth = $dbh->prepare( ... );
$dbh->{mock_clear_history} = 1;
$sth->execute( 'Foo' );
You will have no way to learn from the database handle that the statement parameter 'Foo' was bound.
This is useful mainly to ensure you can isolate the statement histories from each other. A typical sequence will look like:
set handle to framework
perform operations
analyze mock database handle
reset mock database handle history
perform more operations
analyze mock database handle
reset mock database handle history
...
# turn the database off
$dbh->{mock_can_connect} = 0;
# turn it back on again
$dbh->{mock_can_connect} = 1;
The statement handle checks this value as well, so something like this will fail in the expected way:
$dbh = DBI->connect( 'DBI:Mock:', '', '' );
$dbh->{mock_can_connect} = 0;
# blows up!
my $sth = eval { $dbh->prepare( 'SELECT foo FROM bar' ) });
if ( $@ ) {
# Here, $DBI::errstr = 'No connection present'
}
Turning off the database after a statement prepare will fail on the statement execute(), which is hopefully what you would expect:
$dbh = DBI->connect( 'DBI:Mock:', '', '' );
# ok!
my $sth = eval { $dbh->prepare( 'SELECT foo FROM bar' ) });
$dbh->{mock_can_connect} = 0;
# blows up!
$sth->execute;
Similarly:
$dbh = DBI->connect( 'DBI:Mock:', '', '' );
# ok!
my $sth = eval { $dbh->prepare( 'SELECT foo FROM bar' ) });
# ok!
$sth->execute;
$dbh->{mock_can_connect} = 0;
# blows up!
my $row = $sth->fetchrow_arrayref;
Note: The handle attribute Active and the handle method ping will behave according to the value of mock_can_connect. So if mock_can_connect were to be set to 0 (or off), then both Active and ping would return false values (or 0).
Here is a sample usage, partially from the test suite:
my @user_results = (
[ 'login', 'first_name', 'last_name' ],
[ 'cwinters', 'Chris', 'Winters' ],
[ 'bflay', 'Bobby', 'Flay' ],
[ 'alincoln', 'Abe', 'Lincoln' ],
);
my @generic_results = (
[ 'foo', 'bar' ],
[ 'this_one', 'that_one' ],
[ 'this_two', 'that_two' ],
);
my $dbh = DBI->connect( 'DBI:Mock:', '', '' );
$dbh->{mock_add_resultset} = \@user_results; # add first resultset
$dbh->{mock_add_resultset} = \@generic_results; # add second resultset
my ( $sth );
eval {
$sth = $dbh->prepare( 'SELECT login, first_name, last_name FROM foo' );
$sth->execute();
};
# this will fetch rows from the first resultset...
my $row1 = $sth->fetchrow_arrayref;
my $user1 = User->new( login => $row->[0],
first => $row->[1],
last => $row->[2] );
is( $user1->full_name, 'Chris Winters' );
my $row2 = $sth->fetchrow_arrayref;
my $user2 = User->new( login => $row->[0],
first => $row->[1],
last => $row->[2] );
is( $user2->full_name, 'Bobby Flay' );
...
my $sth_generic = $dbh->prepare( 'SELECT foo, bar FROM baz' );
$sth_generic->execute;
# this will fetch rows from the second resultset...
my $row = $sth->fetchrow_arrayref;
You can also associate a resultset with a particular SQL statement instead of adding them in the order they will be fetched:
$dbh->{mock_add_resultset} = {
sql => 'SELECT foo, bar FROM baz',
results => [
[ 'foo', 'bar' ],
[ 'this_one', 'that_one' ],
[ 'this_two', 'that_two' ],
],
};
This will return the given results when the statement 'SELECT foo, bar FROM baz' is prepared. Note that they will be returned every time the statement is prepared, not just the first. It should also be noted that if you want, for some reason, to change the result set bound to a particular SQL statement, all you need to do is add the result set again with the same SQL statement and DBD::Mock will overwrite it.
It should also be noted that the rows method will return the number of records stocked in the result set. So if your code/application makes use of the $sth->rows method for things like UPDATE and DELETE calls you should stock the result set like so:
$dbh->{mock_add_resultset} = {
sql => 'UPDATE foo SET baz = 1, bar = 2',
# this will appear to have updated 3 rows
results => [[ 'rows' ], [], [], []],
};
# or ...
$dbh->{mock_add_resultset} = {
sql => 'DELETE FROM foo WHERE bar = 2',
# this will appear to have deleted 1 row
results => [[ 'rows' ], []],
};
Now I admit this is not the most elegant way to go about this, but it works for me for now, and until I can come up with a better method, or someone sends me a patch ;) it will do for now.
If you want a given statement to fail, you will have to use the hashref method and add a 'failure' key. That key can be handed an arrayref with the error number and error string, in that order. It can also be handed a hashref with two keys - errornum and errorstring. If the 'failure' key has no useful value associated with it, the errornum will be '1' and the errorstring will be 'Unknown error'.
$dbh simply by setting it to undef.
prepare on a per-handle basis. It's starting value can be set with the 'mock_start_insert_id' attribute (see below).
This attribute also can be used with an ARRAY ref parameter, it's behavior is slightly different in that instead of incrementing the value for every prepare it will only increment for each execute. This allows it to be used over multiple execute calls in a single $sth. It's usage looks like this:
$dbh->{mock_last_insert_id} = [ 'Foo', 10 ];
my $sth = $dbh->prepare('INSERT INTO Foo (foo, bar) VALUES(?, ?)');
$sth->execute(1, 2);
# $dbh->{mock_last_insert_id} == 10
$sth->execute(3, 4);
# $dbh->{mock_last_insert_id} == 11
For more examples, please refer to the test file t/025_mock_last_insert_id.t.
But other times you may want to parse the statement as it is prepared rather than after the fact. There is a hook in this mock database driver for you to provide your own parsing routine or object.
The syntax is simple:
$dbh->{mock_add_parser} = sub {
my ( $sql ) = @_;
unless ( $sql =~ /some regex/ ) {
die "does not contain secret fieldname";
}
};
You can also add more than one for a handle. They will be called in order, and the first one to fail will halt the parsing process:
$dbh->{mock_add_parser} = \&parse_update_sql;
$dbh->{mock_add-parser} = \&parse_insert_sql;
Depending on the 'PrintError' and 'RaiseError' settings in the database handle any parsing errors encountered will issue a warn or die. No matter what the statement handle will be undef.
Instead of providing a subroutine reference you can use an object. The only requirement is that it implements the method parse() and takes a SQL statement as the only argument. So you should be able to do something like the following (untested):
my $parser = SQL::Parser->new( 'mysql', { RaiseError => 1 } );
$dbh->{mock_add_parser} = $parser;
In order to capture begin_work(), commit(), and rollback(), DBD::Mock will create statements for them, as if you had issued them in the appropriate SQL command line program. They will go through the standard prepare()-execute() cycle, meaning that any custom SQL parsers will be triggered and DBD::Mock::Session will need to know about these statements.
prepared with. So if the handle were created with:
my $sth = $dbh->prepare( 'SELECT * FROM foo' );
This would return:
SELECT * FROM foo
The original statement is unmodified so if you are checking against it in tests you may want to use a regex rather than a straight equality check. (However if you use a phrasebook to store your SQL externally you are a step ahead...)
Note that this returns the same thing as the normal statement property 'FIELD'.
my $sth = $dbh->prepare( 'SELECT * FROM foo WHERE id = ? AND is_active = ?' ); $sth->bind_param( 2, 'yes' ); $sth->bind_param( 1, 7783 );
This would return:
[ 7738, 'yes' ]
The same result will occur if you pass the parameters via execute() instead:
my $sth = $dbh->prepare( 'SELECT * FROM foo WHERE id = ? AND is_active = ?' ); $sth->execute( 7783, 'yes' );
rows method of the statement handle.
execute() and if you have not yet called a fetch method after the execute.
execute() has been called against the statement handle. Returns 'yes' if so, 'no' if not.
finish() has been called against the statement handle. Returns 'yes' if so, 'no' if not.
fetch() was executed against the statement, or If no return data was set this will return 'no'.
DBD::Mock::StatementTrack object which tracks the actions performed by this statement handle. Most of the actions are separately available from the properties listed above, so you should never need this.
This module can be used to emulate Apache::DBI style DBI connection pooling. Just as with Apache::DBI, you must enable DBD::Mock::Pool before loading DBI.
use DBD::Mock qw(Pool); # followed by ... use DBI;
While this may not seem to make a lot of sense in a single-process testing scenario, it can be useful when testing code which assumes a multi-process Apache::DBI pooled environment.
Under the hood this module does most of the work with a DBD::Mock::StatementTrack object. This is most useful when you are reviewing multiple statements at a time, otherwise you might want to use the mock_* statement handle attributes instead.
$param_num to $value. Returns the arrayref of currently-set bound parameters. This corresponds to the 'bind_param' statement handle call.
@params onto the list of already-set bound parameters.
This object can be used to iterate through the current set of DBD::Mock::StatementTrack objects in the history by fetching the 'mock_all_history_iterator' attribute from a database handle. This object is very simple and is meant to be a convience to make writing long test script easier. Aside from the constructor (new) this object has only one method.
next
Calling next will return the next DBD::Mock::StatementTrack object in the history. If there are no more DBD::Mock::StatementTrack objects available, then this method will return false.
reset
This will reset the internal pointer to the begining of the statement history.
The DBD::Mock::Session object is an alternate means of specifying the SQL statements and result sets for DBD::Mock. The idea is that you can specify a complete 'session' of usage, which will be verified through DBD::Mock. Here is an example:
my $session = DBD::Mock::Session->new('my_session' => (
{
statement => "SELECT foo FROM bar", # as a string
results => [[ 'foo' ], [ 'baz' ]]
},
{
statement => qr/UPDATE bar SET foo \= \'bar\'/, # as a reg-exp
results => [[]]
},
{
statement => sub { # as a CODE ref
my ($SQL, $state) = @_;
return $SQL eq "SELECT foo FROM bar";
},
results => [[ 'foo' ], [ 'bar' ]]
},
{
# with bound parameters
statement => "SELECT foo FROM bar WHERE baz = ? AND borg = ?",
# check exact bound param value,
# then check it against regexp
bound_params => [ 10, qr/\d+/ ],
results => [[ 'foo' ], [ 'baz' ]]
}
));
As you can see, a session is essentially made up a list of HASH references we call 'states'. Each state has a 'statement' and a set of 'results'. If DBD::Mock finds a session in the 'mock_session' attribute, then it will pass the current C<$dbh> and SQL statement to that DBD::Mock::Session. The SQL statement will be checked against the 'statement' field in the current state. If it passes, then the 'results' of the current state will get feed to DBD::Mock through the 'mock_add_resultset' attribute. We then advance to the next state in the session, and wait for the next call through DBD::Mock. If at any time the SQL statement does not match the current state's 'statement', or the session runs out of available states, an error will be raised (and propagated through the normal DBI error handling based on your values for RaiseError and PrintError).
Also, as can be seen in the the session element, bound parameters can also be supplied and tested. In this statement, the SQL is compared, then when the statement is executed, the bound parameters are also checked. The bound parameters much match in both number of parameters and the parameters themselves, or an error will be raised.
As can also be seen in the example above, 'statement' fields can come in many forms. The simplest is a string, which will be compared using eq against the currently running statement. The next is a reg-exp reference, this too will get compared against the currently running statement. The last option is a CODE ref, this is sort of a catch-all to allow for a wide range of SQL comparison approaches (including using modules like SQL::Statement or SQL::Parser for detailed functional comparisons). The first argument to the CODE ref will be the currently active SQL statement to compare against, the second argument is a reference to the current state HASH (in case you need to alter the results, or store extra information). The CODE is evaluated in boolean context and throws and exception if it is false.
new ($session_name, @session_states)
A $session_name can be optionally be specified, along with at least one @session_states. If you don't specify a $session_name, then a default one will be created for you. The @session_states must all be HASH references as well, if this conditions fail, an exception will be thrown.
verify_statement ($dbh, $SQL)
This will check the $SQL against the current state's 'statement' value, and if it passes will add the current state's 'results' to the $dbh. If for some reason the 'statement' value is bad, not of the prescribed type, an exception is thrown. See above for more details.
verify_bound_params ($dbh, $params)
If the 'bound_params' slot is available in the current state, this will check the $params against the current state's 'bound_params' value. Both number of parameters and the parameters themselves must match, or an error will be raised.
reset
Calling this method will reset the state of the session object so that it can be reused.
All functionality listed here is highly experimental and should be used with great caution (if at all).
$dbh->{mock_add_resultset} = {
sql => 'SELECT foo FROM bar',
results => DBD::Mock->NULL_RESULTSET,
failure => [ 5, 'Ooops!' ],
};
The 5 is the DBI error number, and 'Ooops!' is the error string passed to DBI. This basically allows you to force an error condition to occur when a given SQL statement is execute. We are currently working on allowing more control on the 'when' and 'where' the error happens, look for it in future releases.
Right now this feature is highly experimental, and has been added as a first attempt to automatically handle some of the DBD specific attributes which are commonly used/accessed in DBI programming. The functionality is off by default so as to not cause any issues with backwards compatability, but can easily be turned on and off like this:
# turn it on $DBD::Mock::AttributeAliasing++; # turn it off $DBD::Mock::AttributeAliasing = 0;
Once this is turned on, you will need to choose a database specific attribute aliasing table like so:
DBI->connect('dbi:Mock:MySQL', '', '');
The 'MySQL' in the DSN will be picked up and the MySQL specific attribute aliasing will be used.
Right now only MySQL is supported by this feature, and even that support is very minimal. Currently the MySQL $dbh and $sth attributes 'mysql_insertid' are aliased to the $dbh attribute 'mock_last_insert_id'. It is possible to add more aliases though, using the DBD::Mock:_set_mock_attribute_aliases function (see the source code for details).
$dbh attributes. I still need to get deeper into how DBD's work to understand what it is that is actually doing wrong.
I would also like to add the ability to bind a subroutine (or possibly an object) to the result set, so that the results can be somewhat more dynamic and allow for a more realistic interaction.
I use the Devel::Cover manpage to test the code coverage of my tests, below is the the Devel::Cover manpage report on this module test suite.
---------------------------- ------ ------ ------ ------ ------ ------ ------ File stmt bran cond sub pod time total ---------------------------- ------ ------ ------ ------ ------ ------ ------ DBD/Mock.pm 90.9 85.5 76.0 94.1 0.0 100.0 88.4 ---------------------------- ------ ------ ------ ------ ------ ------ ------ Total 90.9 85.5 76.0 94.1 0.0 100.0 88.4 ---------------------------- ------ ------ ------ ------ ------ ------ ------
the DBD::NullP manpage, which provided a good starting point
the Test::MockObject manpage, which provided the approach
Test::MockObject article - http://www.perl.com/pub/a/2002/07/10/tmo.html
Perl Code Kata: Testing Databases - http://www.perl.com/pub/a/2005/02/10/database_kata.html
We have created a DBD::Mock google group for discussion/questions about this module.
http://groups-beta.google.com/group/DBDMock
bind_param_inoutbegin_work(), commit() and rollback() methods.fetchall_hashref(), fetchrow_hashref() and selectcol_arrayref() methods and tests.mock_last_insert_ids patch and test
Copyright (c) 2004 & 2005 Stevan Little, Chris Winters. All rights reserved.
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
Chris Winters <chris@cwinters.com>
Stevan Little <stevan@iinteractive.com>
Rob Kinyon <rob.kinyon@gmail.com>
| DBD::Mock - Mock database driver for testing |