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 DBD::Pg or DBD::mysql) in your framework.
DBD::Mock
DBD::Foo
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 DBD::AnyData or DBD::SQLite 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 Test::MockObject to create a fully dynamic handle. There is an excellent article by chromatic about using Test::MockObject 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:
A database handle contains the history of all statements created against it. Other properties set for the handle (e.g., 'PrintError', 'RaiseError') are left alone and can be queried as normal, but they do not affect anything. (A future feature may track the sequence/history of these assignments but if there is no demand it probably will not get implemented.)
A statement handle contains the statement it was prepared with plus all bound parameters or parameters passed via 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.
execute()
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.
Returns an array reference with all history (a.k.a. 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
Returns a 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.
DBD::Mock::StatementTrack::Iterator
If set to a true value all previous statement history operations will be erased. This includes the history of currently open handles, so if you do something like:
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 ...
This statement allows you to simulate a downed database connection. This is useful in testing how your application/tests will perform in the face of some kind of catastrophic event such as a network outage or database server failure. It is a simple boolean value which defaults to on, and can be set like this:
# 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).
Active
ping
mock_can_connect
This stocks the database handle with a record set, allowing you to seed data for your application to see if it works properly.. Each recordset is a simple arrayref of arrays with the first arrayref being the fieldnames used. Every time a statement handle is created it asks the database handle if it has any resultsets available and if so uses it.
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. (This behavior could change.)
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:
rows
$sth->rows
$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.
This attribute can be used to set a current DBD::Mock::Session object. For more information on this, see the DBD::Mock::Session docs below.
This attribute is incremented each time an INSERT statement is passed to prepare on a per-handle basis. It's starting value can be set with the 'mock_start_insert_id' attribute (see below).
This attribute can be used to set a start value for the 'mock_last_insert_id' attribute. It can also be used to effectively reset the 'mock_last_insert_id' attribute as well.
DBI provides some simple parsing capabilities for 'SELECT' statements to ensure that placeholders are bound properly. And typically you may simply want to check after the fact that a statement is syntactically correct, or at least what you expect.
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.
warn
die
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):
parse()
my $parser = SQL::Parser->new( 'mysql', { RaiseError => 1 } ); $dbh->{mock_add_parser} = $parser;
Returns true if the handle is a 'SELECT' and has more records to fetch, false otherwise. (From the DBI.)
The SQL statement this statement handle was prepared with. So if the handle were created with:
prepare
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...)
Fields used by the statement. As said elsewhere we do no analysis or parsing to find these, you need to define them beforehand. That said, you do not actually need this very often.
Note that this returns the same thing as the normal statement property 'FIELD'.
Returns an arrayref of parameters bound to this statement in the order specified by the bind type. For instance, if you created and stocked a handle with:
my $sth = $dbh->prepare( 'SELECT * FROM foo WHERE id = ? AND is_active = ?' ); $sth->bind_param( 2, 'yes' ); $sth->bind_param( 1, 7783 );
[ 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' );
An arrayref of arrayrefs representing the records the mock statement was stocked with.
Number of records the mock statement was stocked with; if never stocked it is still 0. (Some weirdos might expect undef...)
This returns the same value as mock_num_records. And is what is returned by the rows method of the statement handle.
Current record the statement is on; returns 0 in the instances when you have not yet called execute() and if you have not yet called a fetch method after the execute.
fetch
Whether execute() has been called against the statement handle. Returns 'yes' if so, 'no' if not.
Whether finish() has been called against the statement handle. Returns 'yes' if so, 'no' if not.
finish()
Returns 'yes' if all the records in the recordset have been returned. If no fetch() was executed against the statement, or If no return data was set this will return 'no'.
fetch()
Returns a 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.
mock_*
Takes the following parameters:
return_data: Arrayref of return data records
fields: Arrayref of field names
bound_params: Arrayref of bound parameters
Gets/sets the SQL statement used.
Gets/sets the fields to use for this statement.
Gets/set the bound parameters to use for this statement.
Gets/sets the data to return when asked (that is, when someone calls 'fetch' on the statement handle).
Gets/sets the current record number.
Returns true if the statement is a SELECT and has more records to fetch, false otherwise. (This is from the DBI, see the 'Active' docs under 'ATTRIBUTES COMMON TO ALL HANDLES'.)
Sets the state of the tracker 'executed' flag.
If set to 'yes' tells the tracker that the statement is finished. This resets the current record number to '0' and clears out the array ref of returned records.
Returns true if the current record number is greater than the number of records set to return.
Returns the number of fields set in the 'fields' parameter.
Returns the number of records in the current result set.
Returns the number of parameters set in the 'bound_params' parameter.
Sets bound parameter $param_num to $value. Returns the arrayref of currently-set bound parameters. This corresponds to the 'bind_param' statement handle call.
$param_num
$value
Pushes @params onto the list of already-set bound parameters.
@params
Tells the tracker that the statement has been executed and resets the current record number to '0'.
If the statement has been depleted (all records returned) returns undef; otherwise it gets the current recordfor returning, increments the current record number and returns the current record.
Tries to give an decent depiction of the object state for use in debugging.
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.
new
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 undef.
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' ]] } ));
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 $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', an error will be raised (and propagated through the normal DBI error handling based on your values for RaiseError and PrintError).
$dbh
As can also be seen in the example above, 'statement' fields can come in many forms. The simplest is a string, which will be compared using eq against the currently running statement. The next is a reg-exp reference, this too will get compared against the currently running statement. The last option is a CODE ref, this is sort of a catch-all to allow for a wide range of SQL comparison approaches (including using modules like SQL::Statement or SQL::Parser for detailed functional comparisons). The first argument to the CODE ref will be the currently active SQL statement to compare against, the second argument is a reference to the current state HASH (in case you need to alter the results, or store extra information). The CODE is evaluated in boolean context and throws and exception if it is false.
eq
new ($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.
$session_name
@session_states
verify ($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.
$SQL
All functionality listed here is highly experimental and should be used with great caution (if at all).
Basically this feature allows you to alias attributes to other attributes. So for instance, you can alias a commonly expected attribute like 'mysql_insertid' to something DBD::Mock already has like 'mock_last_insert_id'. While you can also just set 'mysql_insertid' yourself, this functionality allows it to take advantage of things like the autoincrementing of the 'mock_last_insert_id' attribute.
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).
$sth
DBD::Mock:_set_mock_attribute_aliases
When writing the test suite I encountered some odd behavior with some $dbh attributes. I still need to get deeper into how DBD's work to understand what it is that is actually doing wrong.
Each DBD has its own quirks and issues, it would be nice to be able to handle those issues with DBD::Mock in some way. I have an number of ideas already, but little time to sit down and really flesh them out. If you have any suggestions or thoughts, feel free to email me with them.
I would like to have the DBD::Mock::StatementTrack object handle more of the mock_* attributes. This would encapsulate much of the mock_* behavior in one place, which would be a good thing.
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 Devel::Cover to test the code coverage of my tests, below is the Devel::Cover report on this module test suite.
---------------------------- ------ ------ ------ ------ ------ ------ ------ File stmt branch cond sub pod time total ---------------------------- ------ ------ ------ ------ ------ ------ ------ DBD/Mock.pm 89.3 83.9 82.6 93.9 0.0 100.0 87.6 ---------------------------- ------ ------ ------ ------ ------ ------ ------ Total 89.3 83.9 82.6 93.9 0.0 100.0 87.6 ---------------------------- ------ ------ ------ ------ ------ ------ ------
DBI
DBD::NullP, which provided a good starting point
Test::MockObject, which provided the approach
Test::MockObject article - http://www.perl.com/pub/a/2002/07/10/tmo.html
Copyright (c) 2004 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>
To install DBD::Mock, copy and paste the appropriate command in to your terminal.
cpanm
cpanm DBD::Mock
CPAN shell
perl -MCPAN -e shell install DBD::Mock
For more information on module installation, please visit the detailed CPAN module installation guide.