SQL::Preproc - Embed SQL in your Perl (ala SQL preprocessors)


        use SQL::Preproc 
                subclass => 'Chart', 
                emit => 1, 
                keepsql => 1, 
                syntax => [ 'Teradata' ], 
                pponly => 1,
                relax => 1,
                alias => 1,
                debug => 1;

        use DBI;
        use DBI qw(:sql_types);
        use SQL::Preproc::ExceptContainer;
        use DBIx::Chart;

        ...some code...

        my ($host, $user, $passwd) = @ARGV;
        DECLARE CONTEXT $sqlpp_ctxt;

        CONNECT TO "Teradata:$host"
                USER $userid IDENTIFIED BY $passwd AS myconn WITH tdat_mode => ANSI;

        my $more_rows;
        WHENEVER NOTFOUND { $more_rows = undef }

                my ($ctxt, $err, $state, $errmsg) = @_;
                die $errmsg;
        DECLARE CURSOR mycursor AS 
                SELECT * FROM mytable FOR UPDATE;

        OPEN mycursor;

        my ($col1, $col2, $col3, $col4);
        while ($more_rows) {

                FETCH mycursor INTO :$col1, :$col2, :$col3, :$col4;

                if ($col1 > $col4) {
                        UPDATE mytable SET col4 = col4 + 100 
                                WHERE CURRENT OF mycursor;

        CLOSE mycursor;
        #       do a bulkload via array binding
        my @col1 = ();
        my @col2 = ();
        my @col3 = ();

        #       load some data into the arrays, then send it to the DBMS

        EXEC SQL INSERT INTO sometable VALUES(:@col1, :@col2, :@col3);
        #       now create a chart, dumping results to @_
        SELECT * FROM mytable
        WHERE WIDTH=500 
                AND HEIGHT=500
                AND FORMAT='PNG'
                AND LOGO='myimage.png'
                AND X_AXIS='Date'
                AND Y_AXIS='Stock Price'
                AND MAPNAME='stockmap'
                AND SIGNATURE='GOWI Systems, Inc.'
                AND SHOWPOINTS=1
                AND POINT='opencircle';

        open(OUTF, '>mychart.png') || die $!;
        binmode OUTF;
        print OUTF, $_[0];
        close OUTF;

        open(OUTF, '>') || die $!;
        print OUTF, $_[1];
        close OUTF;

        DISCONNECT myconn;




Perl 5.8+ (preferably 5.8.5+)

DBI 1.42

Filter::Simple 0.78

Text::Balanced 1.95

DBIx::Chart 0.01 (only when subclass => 'Chart')

DBI Driver for target database system

(Optional) Filter::ExtractSource 0.02 (to dump generated source code; an alternative to the emit directive).

(Optional) Target DBMS syntax extension module (see "Defining SEMs") defined in a SQL::Preproc::<name> package, e.g., see the included SQL::Preproc::Teradata.


Use the usual

        perl Makefile.PL
        make test
        make install

except on Windows, where you need to

        perl Makefile.PL
        nmake test
        nmake install

Notes on Testing

In order to run the test suite, you'll need at least DBD::CSV; however, DBD::CSV's limited functionality in some areas (e.g., no native TYPE metadata) limits the testing possible (e.g., no DESCRIBE testing). To use another DBI driver, define the environment variable SQLPREPROC_DSN to a DBI compatible DSN string (e.g., 'dbi:Oracle:host'). The test suite will use that value instead of DBD::CSV if defined. If your driver requires a user id and/or password, they can be assigned to the SQLPREPROC_USER and SQLPREPROC_PASSWORD environment variables, respectively.

Also note that you'll need to install DBIx::Chart and all of its dependent modules in order to run the subclass driver tests. Syntax extension testing is not yet possible.


A SQL preprocessor for Perl. Via Filter::Simple, embedded SQL statements are converted into DBI calls to perform the specified database operation.

NOTE: The current release should be considered ALPHA release software, and subject to change at the whim of the author. This release is being provided primarily because it is probably useful for a large number of basic database applications; to solicit feedback; and, frankly, because I've run out of time/resources to push it any farther at the moment, (contributors are certainly welcome...).

The biggest missing pieces at the moment are full integration of SEMs, and more thorough test cases.


For the unfamiliar, embedded SQL first appeared shortly after relational databases, primarily for use in COBOL or PL/1 programs. In recent years, it is most frequently used with C or C++, and, more recently, with Java, via SQLJ. And, now, with Perl!


Currently, SQL::Preproc embeds a complete code translation in place of the source SQL statements, rather than using a runtime module to implement the common functionality. While this leads to significantly more code within the translated source, it minimizes the runtime dependencies on external modules, and provides the opportunity for the user to directly inspect the generated code. A future release will provide a SQL::Preproc::Runtime module, which will provide the common functionality, and reduce the translated code size.


A few usage notes...

Modules to use/require

Due to the ability to disable/re-enable SQL::Preproc, and to process files with multiple packages defined, SQL::Preproc can't "use" all the needed modules at hte right times. Therefore, the application must explicitly use the following modules:

        use DBI;
        use SQL::Preproc::ExceptContainer;
        use DBI qw(:sql_types);         # in case the application needs type info
        use DBIx::<subclass>;           # if the application uses a DBI subclass

These should be use'd after the "use SQL::Preproc;" statement in order to capture them in any emit'd source code.

Also, when using SQL::Preproc on a package, its probably best to "use SQL::Preproc;" before the package statement.

Supported SQL Syntax

SQL statements are detected in the source using the following rules:

  1. The statement must start on its own line, and must be preceded by a line ending in any of '{', '}', ':', '?', ';' (ignoring intervening whitespace and comments) AND

  2. The statement begins with the prefix "EXEC SQL" OR

  3. The statement begins with one of the recognized introductory keywords (uppercase only), and can be subsequently parsed as a recognizable SQL statement OR

  4. The statement begins with an introductory keyword installed by any SQL::Preproc syntax extension module (aka SEM - see "Defining SEMs").

Once a potential SQL statement is detected, SQL::Preproc appends all subsequent lines of code into the statement until a line ending with a semicolon (';') terminator (again, ignoring whitespace and trailing comments) is detected. This accumulation permits support for multistatement SQL requests for those DBMS's that support them (though some syntactic morphing may be required).

Perl Variables as Placeholders

Placeholder values within SQL statements are detected as a colon followed by a valid Perl variable name (as detected by Text::Balanced::extract_variable()). Placeholder behavior follows the following rules when the statement is executed:


For CALL statements only, $scalar-name is assigned the value of the associated OUT parameter.



$scalar is assigned the value of the associated output column

as IN/INOUT CALL parameter

$scalar is bound as a source data parameter and, for INOUT parameters, receives the output parameter value (via bind_param_inout()).

in DESCRIBE..INTO clause

$scalar is assigned an arrayref of hashrefs of the returned column metadata of the form

                Name => $sth->{NAME},
                Type => $sth->{TYPE},
                Precision => $sth->{PRECISION},
                Scale => $sth->{Scale}
        #    ...other column metadata...
as a source data placeholder

the value of $scalar is bound to the associated placeholder


in SELECT..INTO clause

@array is assigned the arrayrefs of all returned rows, ala $sth->fetchall_arrayref(). Only a single placeholder may be specified.

in a FETCH...INTO clause

@array is assigned the associated output column values from a single returned row, ala $sth->fetchrow_array(). Only a single placeholder may be specified.

in DESCRIBE..INTO clause

@array is assigned the hashrefs of the returned column metadata of the form

                Name => $sth->{NAME},
                Type => $sth->{TYPE},
                Precision => $sth->{PRECISION},
                Scale => $sth->{Scale}
        #    ...other column metadata...

Only a single placeholder may be specified.

as a source data placeholder

@array is bound to the associated placeholder for a bulk execute operation, i.e., $sth->bind_param_array() and $sth->execute_array(). Not supported for source data placeholders for SELECT...INTO, cursor, or CALL statements.


in SELECT..INTO clause

%hash is assigned the associated output column values for all returned rows, keyed by the column name. Only a single placeholder may be specified. NOTE that this is different than fetchall_hashref(), which returns a hashref keyed by a selected column's value.

in FETCH..INTO clause

%hash is assigned the associated output column values for a single returned row, keyed by column name, ala $sth->fetchrow_hashref(). Only a single placeholder may be specified.

in DESCRIBE..INTO clause

Receives the returned column metadata, keyed by the returned column names, e.g.,

            $sth->{NAME} =>
                Type => $sth->{TYPE},
                Precision => $sth->{PRECISION},
                Scale => $sth->{Scale}
        #    ...other column metadata...

Only a single placeholder may be specified.

as a source data placeholder

Not supported.

Any combination of references, array elemements, or hash entries may be specified, e.g., :$$scalarref, :@$arrayref, :%$hashref, :$$myhash->{some_entry} are all valid placeholder syntax.

Supplying Placeholder Type Information

This feature is not yet implemented.

In the scenarios described above, the placeholders are bound without explicit type information (i.e., default binding). (Note that in most other strongly-type languages, embedded SQL derives a placeholder's binding type information from the defined type of the associated variable.) In the event a query with placeholders requires explicitly defined type information, SQL::Preproc provides a typing mechanism via the USING clause query prefix:

USING ( perl-variable type-info [ , ... ] ) sql-statement

e.g., assume a target table of

        CREATE TABLE sometable (
                column1 integer,
                column2 varchar(30),
                column3 float

and an INSERT statement of

        INSERT INTO sometable VALUES (:$col1, :$col2, :$col3);

If explicit type information is required, it can be provided via

        USING (col1 int, col2 varchar(30), col3 float)
                INSERT INTO sometable VALUES (:$col1, :$col2, :$col3);

Note that, if the USING clause omits some placeholder variables, those variables will use the default binding.

Acceptable types are any of the following DBI type codes without the 'SQL_' prefix [precision, scale denoted by (P), (P,S)]:


Supplying Statement Attributes

In some (hopefully rare) circumstances, an application needs to specify some DBI statement level attributes with their SQL. SQL::Preproc supports this using the EXEC SQL prefixed form of SQL statements. If the EXEC SQL prefix is immediately followed by an attributes hash list (i.e., { <attribute => value [, ... ] }, it will be supplied to the associated DBI do(), prepare(), or execute() call. E.g.

        EXEC SQL { chart_type_map => [ 
                        { NAME => 'x', TYPE => SQL_INTEGER },
                        { NAME => 'y', TYPE => SQL_INTEGER } ] 
                SELECT * INTO :$image, :$imagemap from simpline;

will result in a DBI call of

        $sth = $dbh->prepare('SELECT * from simpline',
                { chart_type_map => [ 
                        { NAME => 'x', TYPE => SQL_INTEGER },
                        { NAME => 'y', TYPE => SQL_INTEGER } ] 

This method mimics the "escape" syntax used, e.g., by ODBC and JDBC, for normalizing variations on SQL dialect and type formatting. Note that, in the event the SQL statement to be executed begins with such an escape clause, and no statement attributes are needed, the SQL must be preceded by an empty hash list, e.g.,

        EXEC SQL {}
                {fn some_special_sql_dialect};

Admittedly, this method is a bit clumsy, but, assuming statement attributes are rarely needed, it minimizes both the parsing required, and the risk of colliding with some database system's SQL dialect.

Default Recognized SQL Statements

The following statements are recognized by default by SQL::Preproc. Note that wherever a name value is specified, either a literal value or a scalar variable may be used, and the resulting scalar value must match the regex /^[_A-Za-z]\w*$/.


Begins a section of code to be executed as a single transaction, up to any COMMIT WORK or ROLLBACK WORK statement. (Causes current $dbh to set AutoCommit => 0)

CLOSE name

Closes the specified cursor by calling the associated cursor's $sth->finish().


Commits any open transaction on the current connection. AutoCommit is turned back on after this statement is executed.

CONNECT TO dsn [ USER user [ IDENTIFIED BY password ] [ AS name ] [ WITH attributes ]

Connects to the specified dsn as user, password, with an optional attributes hash, optionally assigned the given name.

The dsn, user, password, or name values may be provided as either a scalar variable reference, a string literal, or a parenthesized expression:

The attributes is specified as a hashref list, i.e., attribute => value.

dsn is the usual DBI DSN, although the 'dbi:' prefix is optional.

user is the userid for the connection.

password is the password for the given userid.

name is a name assigned to the connection, to permit disambiguating multiple concurrent connections within the same application. If no name is given, then the default name 'default' is supplied.

attributes is a hash list of driver-specific attributes/value pairs which are supplied to the DBI connect() method.

After a successful CONNECT, the connection is made the "current" connection, which will be used for all SQL operations up to a SET CONNECTION statement changes the current connection.


Identifies a scalar variable to be used as the SQL::Preproc runtime context hashref. The application is responsible for declaring $scalar, and/or passing $scalar to subroutines. If DECLARE CONTEXT is omitted, then SQL::Preproc defaults to $sqlpp_ctxt.

This statement was introduced in order to permit SQL::Preproc runtime context to be shared between multiple Perl packages/objects, so that, e.g., an application may create connections in one object, and then use those connections in other objects/packages. The context is a hashref with the following contents:

            dbhs => { },            # hashref of connection handles, keyed by name
            sths => { },            # hashref of prepared statement handles, keyed by name
            cursors => { },         # hashref of cursor statement handles, keyed by name
            current_dbh => $dbh,    # current connection handle
            current_sth => $sth,    # current statement handle
            SQLERROR => [ ],        # stack of error exception objects
            NOTFOUND => [ ],        # stack of no rows found exception objects
                handler_idx => 0 || -1, # indicates which entry is used on the 
                                                                # exception handler stacks
            rows => $scalar         # rowcount of last DBI statement execution

If the DECLARE'd scalar is not recognized as a hashref at runtime, it will be instantiated with the above structure.

DECLARE CURSOR name AS select-statement

Declares a cursor, possibly updatable.

select-statement must be an explicitly specified SELECT statement. It may contain placeholders, but cannot contain an INTO clause, nor can it be an expression requiring runtime evaluation.

Note that the statement will be prepared and executed on the current connection when the cursor was declared, which may be different than the current connection when the cursor is opened. Updatable cursor syntax is database dependent.

DESCRIBE name [ INTO target ]

Deposits the returned column metadata information for the named cursor statement into the target placeholder, as defined in the "Perl Variables as Placeholders" section above. If no target is provided, the metadata is stored in @_ as an array of hashrefs of metdata.

Note that for some DBI drivers, the column metadata will not be available until after the cursor has been opened (i.e., the statement has been executed).


Disconnects either the current connection (if either name or ALL is not specified), or the named connection, or ALL connections.

EXECUTE name, EXEC name

Executes the named PREPARE'd statement. Placeholder values within the named statement are applied as described in the "Perl Variables as Placeholders" section above. The named statement cannot be a data returning statement.

Note that the statement will be executed on the current connection when the statement was prepared, which may be different than the current connection when the statement is executed.


Immediately executes the specified statement, which is supplied as either a string expression (either a string literal, or any expression which evaluates to a string). The resulting statement cannot be a data returning statement. The statement is executed on the current connection, and is not subject to any SEM translation processing.

FETCH name [ INTO placeholder-list ]

Fetches the current row of the named cursor into the the specied placeholder list (as described in the "Perl Variables as Placeholders" section above); if no INTO clause is specified, the returned column values are stored in @_. The cursor is first advanced to the next row, and then the row is retrieved.

OPEN name

Opens the named cursor by executing the associated SELECT statement. Placeholder data for the SELECT statement is applied as described in the "Perl Variables as Placeholders" section above. When a cursor is OPENed, it is positioned before the first row of the result set.

Note that the cursor will be opened on the current connection when the statement was prepared, which may be different than the current connection when the cursor is opened.

PREPARE name FROM statement

Prepares the specified statement, supplied as a literal SQL statement, and assigns it the specified name. Returned column metadata can be retrieved via the DESCRIBE statement. The statement is prepared on the current connection, and may be processed by any defined SEM's.

Note that the statement will be executed on the current connection when the statement was prepared, which may be different than the current connection when the statement is executed.

RAISE [ SQLERROR | NOT FOUND ] [ param-list ]

Raises the specified exception, passing in the optional parameter list. SQL::Preproc will prepend the DECLARE'd context variable to the param-list before invoking the specified exception handler. The usual exception handler rules apply (i.e., RAISE inside an exception handler will invoke the default handler.


Rolls back any open transaction on the current connection. AutoCommit is turned back on after this statement is executed.

SELECT column-list [ INTO placeholder-list ] select-body

Executes the SELECT statement on the current connection, placing the results into the placeholder-list as described in the "Perl Variables as Placeholders" section above; if no INTO clause is provided, then the returned column values are stored in @_. select-body is the remainder of the SELECT statement , e.g., FROM/WHERE/GROUP/etc. clauses.


Sets the current connection to the named connection. The current connection is always used for succeding SQL statements.

WHENEVER [ SQLERROR | NOT FOUND ] { perl-statements }

Defines an exception handler. A SQLERROR exception is raised whenever a DBI operation returns a defined, non-zero error code, or SQL::Preproc detects a runtime anomaly, e.g., attempting to execute a SQL statement when no connection is available. A NOT FOUND condition is raised whenever a SELECT or OPEN cursor statement returns no rows, or the application attempts to FETCH beyond the end of a cursor. perl-statements is 1 or more Perl statements. SQL::Preproc treats the exception handler as a closure at runtime (hence the required opening and closing braces), and passes in the context hash, the error code, the SQL state, and the error message text when the handler is invoked (NOTE: The error information does not apply for NOT FOUND conditions).

The default exception handler for SQLERROR is

        die $sqlpp_ctxt->{current_dbh}->errstr;

The default exception handler for NOT FOUND is to silently ignore the condition and continue processing; this can be problematic for cursor FETCH loops, so be sure to either define a NOT FOUND handler, or use other information to terminate those loops.

Note that exception handlers are subject to Perl's normal scoping rules, i.e., an exception handler defined within a scope block is not available outside that block. Multiple exception handlers can be defined within multiple scope levels; only the most current handler definition for a condition will be applied. Note that handlers cannot be nested, i.e., a handler cannot be defined inside the body of a handler. Any exceptions raised within an exception handler are handled by the default handlers.

Note that some operations within exception handlers can lead to general chaos, e.g., "goto" labels outside the current scope.

Standard SQL Statements

By default, the following statements are recognized and processed only to extract placeholder variables, and are otherwise executed immediately:


Syntax extension modules may interpret these statements differently.

Default Targets for SELECT/FETCH Operations

If FETCH or SELECT are specified without an expicit INTO Clause, the column values are returned in @_. Note that only the first row's values will be returned for a SELECT.

Configuration Flags

The following flags can be specified in the "use SQL::Preproc" statement:

alias => scalar | undef

When set to a 'true' value (the default), line number aliasing is enabled. When enabled, SQL::Preproc emits #line linenumber into the output source stream, matching the linenumber from which the translated SQL statement was taken from the input stream. Note that the linenumber values may not perfectly align to the original source, especially if SQL::Preproc is disabled (via no SQL::Preproc), and then re-enabled.

debug => scalar | undef

When set to a non-zero value, causes diagnostic information to be dumped to STDERR during the preprocessing phase.

emit => number | filename-string

When set to a defined, nonzero numeric value, causes the translated output from the SQL::Preproc filter to be dumped to STDOUT. When set to a filename string, opens the specified file for writing, and prints translated output to that file. filename-string may be 'STDOUT' or 'STDERR', in which case the output will be sent to STDOUT or STDERR, respectively. This is useful, e.g., for debugging SQL::Preproc behavior, or saving the resulting sourcecode for distribution in systems lacking Filter::Simple or SQL::Preproc support. Also see Filter::ExtractSource for an alternative method of capturing filtered source code.

keepsql => scalar | undef

When set to a defined, nonzero value, causes SQL::Preproc to output the original SQL statements as Perl comments preceding the translated output code.

pponly => scalar | undef

When set to a non-zero value, causes an exit immediately after the preprocessing phase, i.e., the resulting code is not further processed by perl. Useful for diagnosing errors when used in combination with emit.

relax => scalar | undef

When set to a non-zero value, relaxes the generated error checking code. Under normal circumstances, nearly all embedded SQL statements are preceded by various safety checks (e.g., does a connection exist ? Does the referenced cursor exist ?, etc.). Enabling "relaxed" mode causes SQL::Preproc to skip generation of this "safety net" code to both improve performance and reduce generated code size.

subclass => string

When set to a defined string, causes SQL::Preproc to use the string as a DBIx subclassing module name, rather than DBI directly. Note that this can be an issue for subclasses that do not process regular SQL or handle the usual DBI interface methods/attributes. An example supported subclass is 'Chart' (i.e., DBIx::Chart).

syntax => [ SEM-name [,...] ]

When set to a SEM name (e.g., 'Oracle', 'MySQL', etc.), causes SQL::Preproc to attempt to load a SQL::Preproc SEM with the same name (e.g., SQL::Preproc::Oracle, SQL::Preproc::MySQL, etc.). This feature permits driver and/or database specific SQL syntax to be recognized and preprocessed. (See the next section, "Defining SEMs", for details on creating and using SEMs).

Defining SEMs

This feature is not currently fully integrated, but is documented to solicit comments.

In order to support variations in SQL syntax, SQL::Preproc can be configured to load additional syntax preprocessing modules aka SEMs (as described above).

To create a SQL::Preproc SEM, create a new package named "SQL::Preproc::<extension-name>". The module should define

a new() constructor method

This method is the standard object constructor. A SEM object is created for each defined SEM. The constructor should return an instance of the class.

A SQL::Preproc callback is passed to the SEM constructor to permit the constructor to install new keyword hooks, or hooks to existing keywords (see example below).

In addition to the default supported SQL keywords defined in the "Default Recognized SQL Statements" section above, SQL::Preproc supports 2 special keywords: 'EXECSQL' to represent the 'EXEC SQL' introductory keyword, and 'sqlpp_MULTISTMT' to represent multistatement SQL requests.


        package SQL::Preproc::Foo;

        my %my_translations = (
                'FOO', 'my_foo_xlator',
                'BAR', 'my_bar_xlator',
        sub new {
                my ($class, $sqlpp_callback) = @_;
                my $self = { };
                bless $self, $class;

                &$sqlpp_callback($self, $_, $my_translations{$_})
                        foreach (keys %my_translations);
                return $self;
        sub my_foo_xlator {...}

        sub my_bar_xlator {...}
translation methods

These are the methods called whenever SQL::Preproc detects a registered keyword hook for the SEM. The method is called as an object method, and is passed the matching keyword, the complete statement text, the currently DECLARE'd preprocessor context variable name, and an arrayref of placeholder variables extracted from the statement.

This method is responsible for recognizing any specialized SQL statement syntax. If the statement does not match any of its syntax rules, the method returns undef; otherwise, the method generates the appropriate Perl/DBI code to execute the statement.


        sub my_foo_xlator {
                my ($self, $keyword, $stmt, $ctxt, $phs) = @_;
                return undef unless
                        (...parse $stmt here...);
                ...translate $stmt...
                return $xlated_perl_code;

If a SEM returns a 'false' value (undef or zero) from a translation method, SQL::Preproc will pass the statement to the next available SEM registered to the detected keyword. If no other SEM is registered, or can translate the statement, the default translation is attempted. If the default translation cannot be generated, then the original statement is emitted without translation.

Multiple SEMs

SEMs are applied to SQL statements in the order they are defined in the syntax qualifier (See "FETCH Operations" in Default Targets for SELECT).

When SQL::Preproc detects a possible embedded SQL statement, it successively passes the statement to each associated SEM until one of them returns a 'true' value, indicating it successfully translated the statement. If no SEM can translate the statement, the default translation is attempted.

Note that, since translation occurs prior to execution of the resulting script, it is not possible to dynamically apply SEM's at runtime based on current connection information.

Application Access to Runtime Elements

Driver-specific statement and connection metadata can be directly manipulated by the application to do the usual DBI things via the DECLARE'd CONTEXT hashref variable. The context hash contains the following members:


Current connection handle. Set after CONNECT or SET CONNECTION statements.


Name of the current connection.


Current statement handle. Set after a statement is prepared and/or executed, including when a cursor is opened.


Hashref of connection handles, keyed by the associated connection name.


Hashref mapping PREPAREd statement names to statement handles.


Hashref of statement handles, keyed by the associated cursor name, as specified in the DECLARE CURSOR SQL statement.


Hashref mapping of DECLARE'd cursor names to the DBI driver generated cursor names. Note that this value is set on OPEN (i.e., when the statement handle is executed), not DECLARE (i.e., when the statement handle is created).


Hashref of the cursor names indicating the cursor state; set when the cursor is OPEN'ed, and cleared when the cursor is CLOSEd.


Hashref mapping cursor names to the list of any associated placeholder variable names. Used for binding placeholders when the cursor is OPEN'ed.


Hashref mapping PREPAREd statement or cursor names to the name of the connection on which they're open. Used to purge the statement handles when the associated connection is closed.


Hashref mapping PREPAREd statement names to the list of any associated placeholder variable names. Used for binding placeholders when the statement is EXECUTEd.


Arrayref (stack, actually) of SQL::Preproc::ExceptContainer objects. WHENEVER SQLERROR statements create a SQL::Preproc::ExceptContainer object, which is pushed on this arrayref. When any subsequent SQL statement raises an error, the topmost object is invoked to handle the exception. When a SQL::Preproc::ExceptContainer object goes out of scope, it is removed from the stack.


Arrayref (stack, actually) of SQL::Preproc::ExceptContainer objects. WHENEVER NOT FOUND statements create a SQL::Preproc::ExceptContainer object, which is pushed on this arrayref. When any subsequent data-returning SQL statement returns with no rows found, the topmost object is invoked to handle the exception. When a SQL::Preproc::ExceptContainer object goes out of scope, it is removed from the stack.


Contains the current stack index to use for selecting exception handlers. Usually -1, but when an exception is caught, this is changed to 0 in order to force use of the default handlers when within a handler.


Contains the rowcount returned by the various DBI statement execution methods.


Contains the ArrayTupleStatus arrayref populated by execute_array(). Should be inspected in any SQLERROR whenever a bulk operation is performed (i.e., whenever an array variable is bound as an input placeholder).

Applications are free to access and/or modify these items as needed at runtime. E.g., if an application chooses to use some other method for creating a DBI connection, it may avoid using the CONNECT TO... syntax, instead performing something like

        $sqlpp_ctxt->{current_dbh} = $sqlpp_ctxt->{dbhs}{$myname} = 

        ...embedded SQL referencing $myname connection...

Some elements are useful within exception handlers, e.g., current_sth and/or tuple_status in a SQLERROR handler.


There are limits to what can be handled by SQL::Preproc, though it should easily support most common uses of DBI.

Be sure to review the lexical rules described in "Supported SQL Syntax" section above.

Multistatement SQL

Multistatement SQL requests are currently parsed, but not supported, due to DBI's lack of a standard interface for handling multistatement resultsets. Until this issue is resolved, multistatement requests are only supportable by SEMs providing driver-specific implementations.

Avoid $__except_N, $__expr_N Variable Names

WHENEVER clauses result in the creation of a lexically scoped variable with a name like $__except_N, where N is an "uniquifier" integer. Applications should avoid using similarly named variables in order to avoid name conflicts.

Likewise, EXECUTE IMMEDIATE can take any sort of expression, including eval{ }s and heredocs. In order to simplify processing, the argument is assigned to a $__expr_N scalar before applying the value to the DBI do() function.

Generic Runtime Error Codes and States

Currently, runtime exceptions raised due to SQL::Preproc (rather than DBI) always report an error code of -1, and as SQLSTATE of 'S1000'. While the error message text will be detailed enough to describe the error, a future release will attempt to provide more useful/standards conformant error code and SQLSTATE values.

Updatable Cursors

Support for updatable cursors varies widely among DBI drivers; as such, the default implementation provided by SQL::Preproc may not function properly with your DBI driver, and may require a SEM. SQL::Preproc's default behavior assumes

  1. the cursor SELECT statement includes any needed qualifiers to make it updatable, including repeating the DECLARE'd cursor name if needed.

  2. the DBI driver (or database system) synthesizes its own internal name for the cursor, which is provided in the $sth->{CursorName} attribute after execution, and may be different than the DECLARE'd cursor name.

  3. Positioned updates use the WHERE CURRENT OF cursorname suffix, where cursorname is the DECLARE'd cursoname.

  4. SQL::Preproc remaps the synthesized CursorName to the DECLARE'd cursor name.

  5. Currently, UPDATE/DELETE...WHERE CURRENT OF cursor-name cannot be used as a PREPARE'd statement, or within an EXECUTE IMMEDIATE statement expression, due to the need for SQL::Preproc to remap the cursor-name at execution time.

  6. The positioned UPDATE/DELETE must be executed on the same connection on which the cursor was DECLARE'd.

Cursors and Transactions

The behavior of cursors with respect to transaction boundaries varies among database systems. At present, SQL::Preproc makes no assumptions regarding survivability of cursors following transaction commit or rollback.

Stored Procedure CALLs

Like updatable cursors, stored procedure behavior varies between DBI drivers and database systems, and may require a SEM to override SQL::Preproc's default behavior. SQL::Preproc's default implementation

  1. Always uses bind_param_inout() to bind IN/INOUT parameters. IN/INOUT parameters are recognized as :$scalar. OUT parameters are bound via bind_col(), and are recognized as :scalar-name, without any Perl variable introductory qualifier (i.e., no [$%@]).

  2. Replaces the IN/INOUT placeholders with the ? placeholder syntax, and removes the colon prefix from the :scalar-name OUT parameters. Note this latter behavior may conflict with naming requirements of some database systems.

  3. Assigns the positional argument of either bind_param_inout() or bind_col() based on its sequence within the argument list, reading left to right, with the IN/INOUT sequence being separate from the OUT sequence, i.e., the IN/INOUT argument sequence counter is separate from the OUT argument sequence counter.

  4. Supports only scalar variables to be bound for any of IN, INOUT, or OUT parameters.

  5. Does not assume any returned results, other than values returned in INOUT or OUT parameters.

PREPARE'd Statement Restrictions

At present, due to various differences in statement placeholder semantics, some statements cannot be PREPARE'd:

SELECT (or other data returning statements)

The mix of input and output placeholders creates PREPARE issues. As an alternative, use DECLARE CURSOR syntax instead, or simply execute directly.


Stored procedure placeholder semantics (as described above) are not currently handled dur PREPARE processing.


As mentioned above, the remapping of cursor names creates issues with PREPARE.

SEMs and Subclassed DBI

Syntax extension modules and subclassed DBI may be mutually exclusive, if the subclassed DBI uses SQL syntax extensions that may raise errors in the SEM (e.g., DBIx::Chart's "RETURNING..." syntax for rendering charts). In future, SQL::Preproc will permit chaining of SEMs, so that a subclass SEM can filter out its extensions before passing them off to a database-specific SEM.

Perl Language Restrictions

Syntax Error Reporting

SQL::Preproc implements a minimal parser, based on Filter::Simple and Text::Balanced. Misplaced punctuation, and especially missing quote delimiters, are not readily detected, though Perl itself may provide reasonably useful clues about where the syntax error is located.

Cannot Mix Embedded SQL With Trailing Control Statements

SQL::Preproc requires a semicolon ';' to detect the end of an embedded SQL statement. In addition, several embedded SQL statemente generate multiple Perl statements, including control statements. Therefore, trailing Perl control statement syntax is not compatible with embedded SQL. E.g.,

        EXECUTE some_prepared_stmt
                foreach (0..$max_rows);

is not compatible, but

        foreach (0..$max_rows) {
                EXECUTE some_prepared_stmt;

is compatible.

General Identifiers May Be Detected As SQL

Due to the limited parse of the detected SQL statements, under some circumstances, a "false" SQL detect may occur if, e.g., a subroutine has been defined with the same name and case as a registered keyword. E.g, if a subroutine SELECT has been defined, and is referenced in an expression such as

        $variable = $expression ? function1($arg) :
                SELECT ($arg);

then SQL::Preproc will assume SELECT ($arg) is embedded SQL and provide a translation for it. The simplest solution is to avoid defining any subroutine name which might collide with a SQL introductory keyword, especially if it will be used as either a "naked" identifier, or following any of the predecent characters used by SQL::Preproc to detect an SQL statement (any of ;, {, }, :).

Avoid Using SQL Keywords That Are Perl Keywords

Similar to the above case, registering (via a SEM) and using a SQL keyword that is also a Perl keyword (e.g., BEGIN, END), (without using an introductory EXEC SQL prefix), could lead to parsing problems.

Cannot Embed Perl Comments Within Embedded SQL

SQL::Preproc currently does not filter out Perl comments within embedded SQL statements, so

        EXEC SQL
        # this isn't allowed!!!
                INSERT INTO mytable VALUES(...);

will probably cause some problems.

Heredoc Limitations

Due to an apparent bug in Text::Balanced 1.65, heredocs are parsed entirely by SQL::Preproc. The only supported here syntax is

        some heredoc


        some heredoc


        some heredoc

Debugging SQL::Preproc

As a source filter, debugging SQL::Preproc requires an extra pinch of syntactic sugar for debugging. Near the top of the @INC/SQL/ file, is a commented line


Uncomment this line to force the Perl debugger to stop within SQL::Preproc at startup, rather than at the first line of the program being filtered.

If you encounter unexplainable errors, emit the generated code using

        use SQL::Preproc
                emit => '',
                keepsql => 1,
                alias => undef,

This will write the translated code to '', keeping the original SQL as comments, and turns off the line number aliasing. Try running '' under the perl debugger and see if that isolates the error. And if you still have problems, edit the output file to remove any confidential info, and forward to me.


  1. bundle syntax handlers for popular databases (Oracle, MySQL, Pg, etc.) (I'd be happy to consider adding SEMs to the bundle if anyone is interested in providing them).

  2. support for multi-statement requests


Filter::Simple, Text::Balanced, DBI, DBIx::Chart


Special thanks to Damian Conway for his assistance/patience in tweaking Filter::Simple to support SQL::Preproc's requirements.


COPYRIGHT(C) 2004,2005, Dean Arnold, Presicient Corp., USA. All rights reserved.

Permission is granted to use this software according to the terms of the Artistic License, as specified in the Perl README file, with the exception that commercial distribution, either electronic or via physical media, as either a standalone package, or incorporated into a third party product, requires prior written approval of the author.

This software is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

Presicient Corp. reserves the right to provide support for this software to individual sites under a separate (possibly fee-based) agreement.