PApp-SQL/SQL.pm

=head1 NAME

PApp::SQL - absolutely easy yet fast and powerful sql access

=head1 SYNOPSIS

 use PApp::SQL;
 # to be written

=head1 DESCRIPTION

This module provides you with easy-to-use functions to execute sql
commands (using DBI). Despite being easy to use, they are also quite
efficient and allow you to write faster programs in less lines of code.

=over 4

=cut

package PApp::SQL;

use DBI;

#use PApp::Exception; # not yet used

BEGIN {
   use base Exporter;

   $VERSION = 0.11;
   @EXPORT = qw(
         sql_exec sql_fetch sql_fetchall sql_exists sql_insertid $sql_exec
   );
   @EXPORT_OK = qw(
         connect_cached
   );

   require XSLoader;
   XSLoader::load PApp::SQL, $VERSION;
}

our $sql_exec;  # last result of sql_exec's execute call
our $DBH;       # the default database handle
our $database;  # the current SQL::Database object, if applicable

our %dbcache;

=item $dbh = connect_cached $id, $dsn, $user, $pass, $flags, $connect

(not exported by by default)

Connect to the database given by C<($dsn,$user,$pass)>, while using the
flags from C<$flags>. These are just the same arguments as given to
C<DBI->connect>.

The database handle will be cached under the unique id
C<$id|$dsn|$user|$pass>. If the same id is requested later, the
cached handle will be checked (using ping), and the connection will
be re-established if necessary (be sure to prefix your application or
module name to the id to make it "more" unique. Things like __PACKAGE__ .
__LINE__ work fine as well).

The reason C<$id> is necessary is that you might specify special connect
arguments or special flags, or you might want to configure your $DBH
differently than maybe other applications requesting the same database
connection. If none of this is becessary for your application you can
leave $id empty (i.e. "").

If specified, C<$connect> is a callback (e.g. a coderef) that will be
called each time a new connection is being established, with the new
C<$dbh> as first argument.

Examples:

 # try your luck opening the papp database without access info
 $dbh = connect_cached __FILE__, "DBI:mysql:papp";

=cut

sub connect_cached {
   my ($id, $dsn, $user, $pass, $flags, $connect) = @_;
   # the following line is duplicated in PApp::SQL::Database::new
   $id = "$id\0$dsn\0$user\0$pass";
   unless ($dbcache{$id} && $dbcache{$id}->ping) {
      #warn "connecting to ($dsn|$user|$pass|$flags)\n";#d#
      # first, nuke our statement cache (sooory ;)
      cachesize cachesize 0;
      # then connect anew
      $dbcache{$id} =
         eval { DBI->connect($dsn, $user, $pass, $flags) }
         || eval { DBI->connect($dsn, $user, $pass, $flags) }
         || die "unable to connect to database $dsn: $DBI::errstr\n";
      $connect->($dbcache{$id}) if $connect;
   }
   $dbcache{$id};
}

=item $sth = sql_exec [dbh,] [bind-vals...,] "sql-statement", [arguments...]

C<sql_exec> is the most important and most-used function in this module.

Runs the given sql command with the given parameters and returns the
statement handle. The command and the statement handle will be cached
(with the database handle and the sql string as key), so prepare will be
called only once for each distinct sql call (please keep in mind that the
returned statement will always be the same, so, if you call C<sql_exec>
with the same dbh and sql-statement twice (e.g. in a subroutine you
called), the statement handle for the first call mustn't be used.

The database handle (the first argument) is optional. If it is missing,
C<sql_exec> first tries to use the variable C<$DBH> in the current (=
calling) package and, if that fails, it tries to use database handle in
C<$PApp::SQL::DBH>, which you can set before calling these functions.

The actual return value from the C<$sth->execute> call is stored in the
package-global (and exported) variable C<$sql_exec>.

If any error occurs C<sql_exec> will throw an exception.

Examples:

 # easy one
 my $st = sql_exec "select name, id from table where id = ?", $id;
 while (my ($name, $id) = $st->fetchrow_array) { ... };

 # the fastest way to use dbi, using bind_columns
 my $st = sql_exec \my($name, $id),
                   "select name, id from table where id = ?",
                   $id;
 while ($st->fetch) { ...}

 # now use a different dastabase:
 sql_exec $dbh, "update file set name = ?", "oops.txt";


=item sql_fetch <see sql_exec>

Execute a sql-statement and fetch the first row of results. Depending on
the caller context the row will be returned as a list (array context), or
just the first columns. In table form:

 CONTEXT        RESULT
 void           ()
 scalar         first column
 list           array

C<sql_fetch> is quite efficient in conjunction with bind variables:

 sql_fetch \my($name, $amount),
           "select name, amount from table where id name  = ?",
           "Toytest";

But of course the normal way to call it is simply:

 my($name, $amount) = sql_fetch "select ...", args...

... and it's still quite fast unless you fetch large amounts of data.

=item sql_fetchall <see sql_exec>

Similarly to C<sql_fetch>, but all result rows will be fetched (this is
of course inefficient for large results!). The context is ignored (only
list context makes sense), but the result still depends on the number of
columns in the result:

 COLUMNS        RESULT
 0              ()
 1              (row1, row2, row3...)
 many           ([row1], [row2], [row3]...)

Examples (all of which are inefficient):

 for (sql_fetchall "select id from table") { ... }

 my @names = sql_fetchall "select name from user";

 for (sql_fetchall "select name, age, place from user") {
    my ($name, $age, $place) = @$_;
 }

=item sql_exists "<table> where ...", args...

Check wether the result of the sql-statement "select xxx from
$first_argument" would be empty or not (that is, imagine the string
"select from" were prepended to your statement (it isn't)). Should work
with every database but can be quite slow, except on mysql, where this
should be quite fast.

Examples:

 print "user 7 exists!\n"
    if sql_exists "user where id = ?", 7;
 
 die "duplicate key"
    if sql_exists "user where name = ? and pass = ?", "stefan", "geheim";

=cut

=item $lastid = sql_insertid $sth

Returns the last automatically created key value. It must be executed
directly after executing the insert statement that created it. This is
what is actually returned for various databases. If your database is
missing, please send me an e-mail on how to implement this ;)

 mysql:    first C<AUTO_INCREMENT> column set to NULL
 postgres: C<oid> column (is there a way to get the last SERIAL?)
 sybase:   C<IDENTITY> column of the last insert (slow)
 informix: C<SERIAL> or C<SERIAL8> column of the last insert

Except for sybase, this does not require a server access.

=cut

sub sql_insertid($) {
   my $sth = shift or die "sql_insertid requires a statement handle";
   my $dbh = $sth->{Database};
   my $driver = $dbh->{Driver}{Name};

   $driver eq "mysql"    and return $sth->{mysql_insertid};
   $driver eq "Pg"       and return $sth->{pg_oid_status};
   $driver eq "Sybase"   and return sql_fetch($dbh, 'SELECT @@IDENTITY');
   $driver eq "Informix" and return $sth->{ix_sqlerrd}[1];

   die "sql_insertid does not spport the dbd driver '$driver', please see PApp::SQL::sql_insertid";
}

=item [old-size] = cachesize [new-size]

Returns (and possibly changes) the LRU cache size used by C<sql_exec>. The
default is somewhere around 50 (= the 50 last recently used statements
will be cached). It shouldn't be too large, since a simple linear listed
is used for the cache at the moment (which, for small (<100) cache sizes
is actually quite fast).

The function always returns the cache size in effect I<before> the call,
so, to nuke the cache (for example, when a database connection has died
or you want to garbage collect old database/statement handles), this
construct can be used:

 PApp::SQL::cachesize PApp::SQL::cachesize 0;

=cut

=item reinitialize [not exported]

Clears any internal caches (statement cache, database handle
cache). Should be called after C<fork> and other accidents that invalidate
database handles.

=cut

sub reinitialize {
   cachesize cachesize 0;
   for (values %dbcache) {
      eval { $_->disconnect };
   }
   undef %dbcache;
}

=back

=cut

reinitialize;

package PApp::SQL::Database;

=head2 THE DATABASE CLASS

Again (sigh) the problem of persistency. What do you do when you have to serialize on object
that contains (or should contain) a database handle? Short answer: you don't. Long answer:
you can embed the necessary information to recreate the dbh when needed.

The C<PApp::SQL::Database> class does that, in a relatively efficient
fashion: the overhead is currently a single method call per access (you
can cache the real dbh if you want).

=over 4

=item $db = new <same arguments as C<connect_cached>>

The C<new> call takes the same arguments as C<connect_cached> (obviously,
if you supply a connect callback it better is serializable, see
L<PApp::Callback>!) and returns a serializable database class. No database
handle is actually being created.

=item $db->dbh

Return the database handle as fast as possible (usually just a hash lookup).

=item $db->checked_dbh

Return the database handle, but first check that the database is still
available and re-open the connection if necessary.

=cut

sub new($$;@) {
   my $class = shift;
   my ($id, $dsn, $user, $pass, $flags, $connect) = @_;
   # the following line is duplicated in PApp::SQL::Database::new
   my $id2 = "$id\0$dsn\0$user\0$pass";
   bless [$id2, $flags, $connect], $class;
}

# the following two functions better be fast!
sub dbh($) {
   $dbcache{$_[0][0]} || $_[0]->checked_dbh;
}

sub checked_dbh($) {
   my $dbh = $dbcache{$_[0][0]};
   $dbh && $dbh->ping
      ? $dbh
      : PApp::SQL::connect_cached((split /\x00/, $_[0][0]), $_[0][1], $_[0][2]);
}

=item $db->dsn

Return the DSN (L<DBI>) fo the database object (e.g. for error messages).

=cut

sub dsn($) {
   my $self = shift;
   (split /\x00/, $self->[0])[1];
}

=back

=cut

1;

=head1 BUGS

As of this writing, sql_fetch and sql_fetchall are not very well tested
(they were just re-written in C).

sql_exists could be faster (it is written very ugly to not change the
current package).

=head1 SEE ALSO

L<PApp>.

=head1 AUTHOR

 Marc Lehmann <pcg@goof.com>
 http://www.goof.com/pcg/marc/

=cut

Revision:	1.9
Committed:	Sat Feb 3 17:48:02 2001 UTC (23 years, 3 months ago) by root
Branch:	MAIN
Changes since 1.8:	+4 -2 lines
Log Message:	* empty log message *
#	User	Rev	Content
1	root	1.1	=head1 NAME
2
3			PApp::SQL - absolutely easy yet fast and powerful sql access
4
5			=head1 SYNOPSIS
6
7			use PApp::SQL;
8			# to be written
9
10			=head1 DESCRIPTION
11
12			This module provides you with easy-to-use functions to execute sql
13			commands (using DBI). Despite being easy to use, they are also quite
14			efficient and allow you to write faster programs in less lines of code.
15
16			=over 4
17
18			=cut
19
20			package PApp::SQL;
21
22			use DBI;
23
24			#use PApp::Exception; # not yet used
25
26			BEGIN {
27			use base Exporter;
28
29	root	1.4	$VERSION = 0.11;
30	root	1.1	@EXPORT = qw(
31			sql_exec sql_fetch sql_fetchall sql_exists sql_insertid $sql_exec
32			);
33			@EXPORT_OK = qw(
34			connect_cached
35			);
36
37			require XSLoader;
38			XSLoader::load PApp::SQL, $VERSION;
39			}
40
41			our $sql_exec; # last result of sql_exec's execute call
42			our $DBH; # the default database handle
43			our $database; # the current SQL::Database object, if applicable
44
45			our %dbcache;
46
47			=item $dbh = connect_cached $id, $dsn, $user, $pass, $flags, $connect
48
49			(not exported by by default)
50
51			Connect to the database given by C<($dsn,$user,$pass)>, while using the
52			flags from C<$flags>. These are just the same arguments as given to
53			C<DBI->connect>.
54
55	root	1.6	The database handle will be cached under the unique id
56			C<$id\|$dsn\|$user\|$pass>. If the same id is requested later, the
57			cached handle will be checked (using ping), and the connection will
58			be re-established if necessary (be sure to prefix your application or
59			module name to the id to make it "more" unique. Things like __PACKAGE__ .
60			__LINE__ work fine as well).
61
62			The reason C<$id> is necessary is that you might specify special connect
63			arguments or special flags, or you might want to configure your $DBH
64			differently than maybe other applications requesting the same database
65			connection. If none of this is becessary for your application you can
66			leave $id empty (i.e. "").
67	root	1.1
68			If specified, C<$connect> is a callback (e.g. a coderef) that will be
69			called each time a new connection is being established, with the new
70			C<$dbh> as first argument.
71
72			Examples:
73
74			# try your luck opening the papp database without access info
75			$dbh = connect_cached __FILE__, "DBI:mysql:papp";
76
77			=cut
78
79			sub connect_cached {
80			my ($id, $dsn, $user, $pass, $flags, $connect) = @_;
81			# the following line is duplicated in PApp::SQL::Database::new
82			$id = "$id\0$dsn\0$user\0$pass";
83			unless ($dbcache{$id} && $dbcache{$id}->ping) {
84			#warn "connecting to ($dsn\|$user\|$pass\|$flags)\n";#d#
85	root	1.5	# first, nuke our statement cache (sooory ;)
86	root	1.1	cachesize cachesize 0;
87			# then connect anew
88			$dbcache{$id} =
89			eval { DBI->connect($dsn, $user, $pass, $flags) }
90			\|\| eval { DBI->connect($dsn, $user, $pass, $flags) }
91	root	1.5	\|\| die "unable to connect to database $dsn: $DBI::errstr\n";
92	root	1.1	$connect->($dbcache{$id}) if $connect;
93			}
94			$dbcache{$id};
95			}
96
97			=item $sth = sql_exec [dbh,] [bind-vals...,] "sql-statement", [arguments...]
98
99			C<sql_exec> is the most important and most-used function in this module.
100
101			Runs the given sql command with the given parameters and returns the
102			statement handle. The command and the statement handle will be cached
103			(with the database handle and the sql string as key), so prepare will be
104			called only once for each distinct sql call (please keep in mind that the
105			returned statement will always be the same, so, if you call C<sql_exec>
106			with the same dbh and sql-statement twice (e.g. in a subroutine you
107			called), the statement handle for the first call mustn't be used.
108
109			The database handle (the first argument) is optional. If it is missing,
110			C<sql_exec> first tries to use the variable C<$DBH> in the current (=
111			calling) package and, if that fails, it tries to use database handle in
112			C<$PApp::SQL::DBH>, which you can set before calling these functions.
113
114			The actual return value from the C<$sth->execute> call is stored in the
115			package-global (and exported) variable C<$sql_exec>.
116
117			If any error occurs C<sql_exec> will throw an exception.
118
119			Examples:
120
121			# easy one
122			my $st = sql_exec "select name, id from table where id = ?", $id;
123			while (my ($name, $id) = $st->fetchrow_array) { ... };
124
125			# the fastest way to use dbi, using bind_columns
126			my $st = sql_exec \my($name, $id),
127			"select name, id from table where id = ?",
128			$id;
129			while ($st->fetch) { ...}
130
131			# now use a different dastabase:
132			sql_exec $dbh, "update file set name = ?", "oops.txt";
133
134
135			=item sql_fetch <see sql_exec>
136
137			Execute a sql-statement and fetch the first row of results. Depending on
138			the caller context the row will be returned as a list (array context), or
139			just the first columns. In table form:
140
141			CONTEXT RESULT
142			void ()
143			scalar first column
144			list array
145
146			C<sql_fetch> is quite efficient in conjunction with bind variables:
147
148			sql_fetch \my($name, $amount),
149			"select name, amount from table where id name = ?",
150			"Toytest";
151
152			But of course the normal way to call it is simply:
153
154			my($name, $amount) = sql_fetch "select ...", args...
155
156			... and it's still quite fast unless you fetch large amounts of data.
157
158			=item sql_fetchall <see sql_exec>
159
160			Similarly to C<sql_fetch>, but all result rows will be fetched (this is
161			of course inefficient for large results!). The context is ignored (only
162			list context makes sense), but the result still depends on the number of
163			columns in the result:
164
165			COLUMNS RESULT
166			0 ()
167			1 (row1, row2, row3...)
168			many ([row1], [row2], [row3]...)
169
170			Examples (all of which are inefficient):
171
172			for (sql_fetchall "select id from table") { ... }
173
174			my @names = sql_fetchall "select name from user";
175
176			for (sql_fetchall "select name, age, place from user") {
177			my ($name, $age, $place) = @$_;
178			}
179
180			=item sql_exists "<table> where ...", args...
181
182			Check wether the result of the sql-statement "select xxx from
183			$first_argument" would be empty or not (that is, imagine the string
184			"select from" were prepended to your statement (it isn't)). Should work
185			with every database but can be quite slow, except on mysql, where this
186			should be quite fast.
187
188			Examples:
189
190			print "user 7 exists!\n"
191			if sql_exists "user where id = ?", 7;
192
193			die "duplicate key"
194			if sql_exists "user where name = ? and pass = ?", "stefan", "geheim";
195
196			=cut
197
198	root	1.3	=item $lastid = sql_insertid $sth
199
200	root	1.8	Returns the last automatically created key value. It must be executed
201			directly after executing the insert statement that created it. This is
202			what is actually returned for various databases. If your database is
203			missing, please send me an e-mail on how to implement this ;)
204
205			mysql: first C<AUTO_INCREMENT> column set to NULL
206			postgres: C<oid> column (is there a way to get the last SERIAL?)
207			sybase: C<IDENTITY> column of the last insert (slow)
208			informix: C<SERIAL> or C<SERIAL8> column of the last insert
209
210			Except for sybase, this does not require a server access.
211	root	1.3
212			=cut
213
214			sub sql_insertid($) {
215			my $sth = shift or die "sql_insertid requires a statement handle";
216			my $dbh = $sth->{Database};
217			my $driver = $dbh->{Driver}{Name};
218
219	root	1.8	$driver eq "mysql" and return $sth->{mysql_insertid};
220			$driver eq "Pg" and return $sth->{pg_oid_status};
221			$driver eq "Sybase" and return sql_fetch($dbh, 'SELECT @@IDENTITY');
222	root	1.3	$driver eq "Informix" and return $sth->{ix_sqlerrd}[1];
223
224			die "sql_insertid does not spport the dbd driver '$driver', please see PApp::SQL::sql_insertid";
225	root	1.1	}
226
227			=item [old-size] = cachesize [new-size]
228
229			Returns (and possibly changes) the LRU cache size used by C<sql_exec>. The
230			default is somewhere around 50 (= the 50 last recently used statements
231			will be cached). It shouldn't be too large, since a simple linear listed
232			is used for the cache at the moment (which, for small (<100) cache sizes
233			is actually quite fast).
234
235			The function always returns the cache size in effect I<before> the call,
236			so, to nuke the cache (for example, when a database connection has died
237			or you want to garbage collect old database/statement handles), this
238			construct can be used:
239
240			PApp::SQL::cachesize PApp::SQL::cachesize 0;
241
242			=cut
243
244			=item reinitialize [not exported]
245
246	root	1.9	Clears any internal caches (statement cache, database handle
247			cache). Should be called after C<fork> and other accidents that invalidate
248			database handles.
249	root	1.1
250			=cut
251
252			sub reinitialize {
253			cachesize cachesize 0;
254			for (values %dbcache) {
255			eval { $_->disconnect };
256			}
257			undef %dbcache;
258			}
259
260			=back
261
262			=cut
263
264	root	1.7	reinitialize;
265
266	root	1.1	package PApp::SQL::Database;
267
268			=head2 THE DATABASE CLASS
269
270			Again (sigh) the problem of persistency. What do you do when you have to serialize on object
271			that contains (or should contain) a database handle? Short answer: you don't. Long answer:
272			you can embed the necessary information to recreate the dbh when needed.
273
274			The C<PApp::SQL::Database> class does that, in a relatively efficient
275			fashion: the overhead is currently a single method call per access (you
276			can cache the real dbh if you want).
277
278			=over 4
279
280			=item $db = new <same arguments as C<connect_cached>>
281
282			The C<new> call takes the same arguments as C<connect_cached> (obviously,
283			if you supply a connect callback it better is serializable, see
284			L<PApp::Callback>!) and returns a serializable database class. No database
285			handle is actually being created.
286
287			=item $db->dbh
288
289			Return the database handle as fast as possible (usually just a hash lookup).
290
291			=item $db->checked_dbh
292
293			Return the database handle, but first check that the database is still
294			available and re-open the connection if necessary.
295
296			=cut
297
298			sub new($$;@) {
299			my $class = shift;
300			my ($id, $dsn, $user, $pass, $flags, $connect) = @_;
301			# the following line is duplicated in PApp::SQL::Database::new
302			my $id2 = "$id\0$dsn\0$user\0$pass";
303			bless [$id2, $flags, $connect], $class;
304			}
305
306			# the following two functions better be fast!
307			sub dbh($) {
308			$dbcache{$_[0][0]} \|\| $_[0]->checked_dbh;
309			}
310
311			sub checked_dbh($) {
312			my $dbh = $dbcache{$_[0][0]};
313			$dbh && $dbh->ping
314			? $dbh
315			: PApp::SQL::connect_cached((split /\x00/, $_[0][0]), $_[0][1], $_[0][2]);
316			}
317
318			=item $db->dsn
319
320			Return the DSN (L<DBI>) fo the database object (e.g. for error messages).
321
322			=cut
323
324			sub dsn($) {
325			my $self = shift;
326	root	1.9	(split /\x00/, $self->[0])[1];
327	root	1.1	}
328
329			=back
330
331			=cut
332
333			1;
334
335			=head1 BUGS
336
337			As of this writing, sql_fetch and sql_fetchall are not very well tested
338			(they were just re-written in C).
339
340			sql_exists could be faster (it is written very ugly to not change the
341			current package).
342
343			=head1 SEE ALSO
344
345			L<PApp>.
346
347			=head1 AUTHOR
348
349			Marc Lehmann <pcg@goof.com>
350			http://www.goof.com/pcg/marc/
351
352			=cut
353