NAME
SQL::SyntaxModel - An abstract syntax tree for all types of SQL
DEPENDENCIES
Perl Version: 5.006
Standard Modules: none
Nonstandard Modules:
Locale::KeyedText 0.03 (for error messages)COPYRIGHT AND LICENSE
This file is part of the SQL::SyntaxModel library (libSQLSM).
SQL::SyntaxModel is Copyright (c) 1999-2004, Darren R. Duncan. All rights reserved. Address comments, suggestions, and bug reports to perl@DarrenDuncan.net, or visit "http://www.DarrenDuncan.net" for more information.
SQL::SyntaxModel is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License (GPL) version 2 as published by the Free Software Foundation (http://www.fsf.org/). You should have received a copy of the GPL as part of the SQL::SyntaxModel distribution, in the file named "LICENSE"; if not, write to the Free Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA.
Linking SQL::SyntaxModel statically or dynamically with other modules is making a combined work based on SQL::SyntaxModel. Thus, the terms and conditions of the GPL cover the whole combination. As a special exception, the copyright holders of SQL::SyntaxModel give you permission to link SQL::SyntaxModel with independent modules, regardless of the license terms of these independent modules, and to copy and distribute the resulting combined work under terms of your choice, provided that every copy of the combined work is accompanied by a complete copy of the source code of SQL::SyntaxModel (the version of SQL::SyntaxModel used to produce the combined work), being distributed under the terms of the GPL plus this exception. An independent module is a module which is not derived from or based on SQL::SyntaxModel, and which is fully useable when not linked to SQL::SyntaxModel in any form.
Any versions of SQL::SyntaxModel that you modify and distribute must carry prominent notices stating that you changed the files and the date of any changes, in addition to preserving this original copyright notice and other credits. SQL::SyntaxModel 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.
While it is by no means required, the copyright holders of SQL::SyntaxModel would appreciate being informed any time you create a modified version of SQL::SyntaxModel that you are willing to distribute, because that is a practical way of suggesting improvements to the standard version.
SYNOPSIS
Model-Building Perl Code Examples
This module's native API is highly verbose / detailed and so a realistically complete example of its use would be too large to show here. (In fact, most real uses of the module would involve user-picked wrapper functions that aren't included.) However, here are a few example usage lines:
use SQL::SyntaxModel;
eval {
my $model = SQL::SyntaxModel->new_container();
# ... add a few Nodes
# Create user-defined data type domain that our database record primary keys are:
my $dom_entity_id = SQL::SyntaxModel->new_node( 'domain' );
$dom_entity_id->set_node_id( 1 );
$dom_entity_id->put_in_container( $model );
$dom_entity_id->add_reciprocal_links();
$dom_entity_id->set_node_ref_attribute( 'schema', $schema );
$dom_entity_id->set_parent_node_attribute_name( 'schema' );
$dom_entity_id->set_literal_attribute( 'name', 'entity_id' );
$dom_entity_id->set_enumerated_attribute( 'base_type', 'NUM_INT' );
$dom_entity_id->set_literal_attribute( 'num_precision', 9 );
# Define the table that holds our data:
my $tb_person = $pp_node->new_node( 'table' );
$tb_person->set_node_id( 1 );
$tb_person->put_in_container( $model );
$tb_person->add_reciprocal_links();
$tb_person->set_node_ref_attribute( 'schema', $schema );
$tb_person->set_parent_node_attribute_name( 'schema' );
$tb_person->set_literal_attribute( 'name', 'person' );
# Define the 'person id' column of that table:
my $tbc_person_id = $pp_node->new_node( 'table_col' );
$tbc_person_id->set_node_id( 1 );
$tbc_person_id->put_in_container( $model );
$tbc_person_id->add_reciprocal_links();
$tbc_person_id->set_node_ref_attribute( 'table', $tb_person );
$tbc_person_id->set_parent_node_attribute_name( 'table' );
$tbc_person_id->set_literal_attribute( 'name', 'person_id' );
$tbc_person_id->set_node_ref_attribute( 'domain', $dom_entity_id );
$tbc_person_id->set_literal_attribute( 'mandatory', 1 );
$tbc_person_id->set_literal_attribute( 'default_val', 1 );
$tbc_person_id->set_literal_attribute( 'auto_inc', 1 );
# ... add a lot more Nodes
# Now check that we didn't omit something important:
$model->with_all_nodes_test_mandatory_attributes();
# Now serialize all our Nodes to see if we stored what we expected:
print $model->get_all_properties_as_xml_str();
# Now explicitly destroy our Container so we don't leak memory:
$model->destroy();
};
if( $@ ) {
my $translator = Locale::KeyedText->new_translator( ['SQL::SyntaxModel::L::'], ['en'] );
my $user_text = $translator->translate_message( $@ );
print "SOMETHING'S WRONG: $user_text" );
}The above code sample is taken and slightly altered from a longer set of code in this module's test script/module: 't/SQL_SyntaxModel.t' and 'lib/t_SQL_SyntaxModel.pm'. Even that code is an incomplete sample, but it also demonstrates the use of a couple simple wrapper functions.
A Complete Example Model Serialized to XML
This is a serialization of the model that the test code makes, which should give you a better idea what kind of information is stored in a SQL::SynaxModel:
<root>
<elements />
<blueprints>
<catalog id="1">
<owner id="1" catalog="1" />
<schema id="1" catalog="1" name="gene" owner="1">
<domain id="1" schema="1" name="entity_id" base_type="NUM_INT" num_precision="9" />
<domain id="2" schema="1" name="person_name" base_type="STR_CHAR" max_chars="100" char_enc="UTF8" />
<table id="1" schema="1" name="person">
<table_col id="1" table="1" name="person_id" domain="1" mandatory="1" default_val="1" auto_inc="1" />
<table_col id="2" table="1" name="name" domain="2" mandatory="1" />
<table_col id="3" table="1" name="father_id" domain="1" mandatory="0" />
<table_col id="4" table="1" name="mother_id" domain="1" mandatory="0" />
<table_ind id="1" table="1" name="primary" ind_type="UNIQUE">
<table_ind_col id="1" table_ind="1" table_col="1" />
</table_ind>
<table_ind id="2" table="1" name="fk_father" ind_type="FOREIGN" f_table="1">
<table_ind_col id="2" table_ind="2" table_col="3" f_table_col="1" />
</table_ind>
<table_ind id="3" table="1" name="fk_mother" ind_type="FOREIGN" f_table="1">
<table_ind_col id="3" table_ind="3" table_col="4" f_table_col="1" />
</table_ind>
</table>
</schema>
</catalog>
<application id="1" name="Setup">
<catalog_link id="1" application="1" name="admin_link" target="1" />
<command id="1" application="1" name="install_app_schema" command_type="DB_CREATE" command_arg="1" />
<command id="2" application="1" name="remove_app_schema" command_type="DB_DELETE" command_arg="1" />
</application>
<application id="2" name="People Watcher">
<catalog_link id="2" application="2" name="editor_link" target="1" />
<routine id="1" routine_type="ANONYMOUS" application="2" name="fetch_all_persons" return_var_type="CURSOR">
<view id="2" view_type="MATCH" name="fetch_all_persons" routine="1" match_all_cols="1">
<view_src id="3" view="2" name="person" match_table="1" />
</view>
<routine_var id="4" routine="1" name="person_cursor" var_type="CURSOR" curs_view="2" />
<routine_stmt id="5" routine="1" stmt_type="SPROC" call_sproc="CURSOR_OPEN">
<routine_expr id="6" expr_type="VAR" p_stmt="5" routine_var="4" />
</routine_stmt>
<routine_stmt id="7" routine="1" stmt_type="RETURN">
<routine_expr id="8" expr_type="VAR" p_stmt="7" routine_var="4" />
</routine_stmt>
</routine>
<routine id="9" routine_type="ANONYMOUS" application="2" name="insert_a_person">
<routine_arg id="10" routine="9" name="arg_person_id" var_type="SCALAR" domain="1" />
<routine_arg id="11" routine="9" name="arg_person_name" var_type="SCALAR" domain="2" />
<routine_arg id="12" routine="9" name="arg_father_id" var_type="SCALAR" domain="1" />
<routine_arg id="13" routine="9" name="arg_mother_id" var_type="SCALAR" domain="1" />
<view id="14" view_type="MATCH" name="insert_a_person" routine="9">
<view_src id="15" view="14" name="person" match_table="1">
<view_src_col id="16" src="15" match_table_col="1" />
<view_src_col id="17" src="15" match_table_col="2" />
<view_src_col id="18" src="15" match_table_col="3" />
<view_src_col id="19" src="15" match_table_col="4" />
</view_src>
<view_expr id="20" expr_type="ARG" view="14" view_part="SET" set_view_col="16" routine_arg="10" />
<view_expr id="21" expr_type="ARG" view="14" view_part="SET" set_view_col="17" routine_arg="11" />
<view_expr id="22" expr_type="ARG" view="14" view_part="SET" set_view_col="18" routine_arg="12" />
<view_expr id="23" expr_type="ARG" view="14" view_part="SET" set_view_col="19" routine_arg="13" />
</view>
<routine_stmt id="24" routine="9" stmt_type="SPROC" call_sproc="INSERT" view_for_dml="14" />
</routine>
<routine id="25" routine_type="ANONYMOUS" application="2" name="update_a_person">
<routine_arg id="26" routine="25" name="arg_person_id" var_type="SCALAR" domain="1" />
<routine_arg id="27" routine="25" name="arg_person_name" var_type="SCALAR" domain="2" />
<routine_arg id="28" routine="25" name="arg_father_id" var_type="SCALAR" domain="1" />
<routine_arg id="29" routine="25" name="arg_mother_id" var_type="SCALAR" domain="1" />
<view id="30" view_type="MATCH" name="update_a_person" routine="25">
<view_src id="31" view="30" name="person" match_table="1">
<view_src_col id="32" src="31" match_table_col="1" />
<view_src_col id="33" src="31" match_table_col="2" />
<view_src_col id="34" src="31" match_table_col="3" />
<view_src_col id="35" src="31" match_table_col="4" />
</view_src>
<view_expr id="36" expr_type="ARG" view="30" view_part="SET" set_view_col="33" routine_arg="27" />
<view_expr id="37" expr_type="ARG" view="30" view_part="SET" set_view_col="34" routine_arg="28" />
<view_expr id="38" expr_type="ARG" view="30" view_part="SET" set_view_col="35" routine_arg="29" />
<view_expr id="39" expr_type="SFUNC" view="30" view_part="WHERE" call_sfunc="EQ">
<view_expr id="40" expr_type="COL" p_expr="39" src_col="32" />
<view_expr id="41" expr_type="ARG" p_expr="39" routine_arg="26" />
</view_expr>
</view>
<routine_stmt id="42" routine="25" stmt_type="SPROC" call_sproc="UPDATE" view_for_dml="30" />
</routine>
<routine id="43" routine_type="ANONYMOUS" application="2" name="delete_a_person">
<routine_arg id="44" routine="43" name="arg_person_id" var_type="SCALAR" domain="1" />
<view id="45" view_type="MATCH" name="delete_a_person" routine="43">
<view_src id="46" view="45" name="person" match_table="1">
<view_src_col id="47" src="46" match_table_col="1" />
</view_src>
<view_expr id="48" expr_type="SFUNC" view="45" view_part="WHERE" call_sfunc="EQ">
<view_expr id="49" expr_type="COL" p_expr="48" src_col="47" />
<view_expr id="50" expr_type="ARG" p_expr="48" routine_arg="44" />
</view_expr>
</view>
<routine_stmt id="51" routine="43" stmt_type="SPROC" call_sproc="DELETE" view_for_dml="45" />
</routine>
</application>
</blueprints>
<tools>
<data_storage_product id="1" product_code="SQLite_2_8_12" is_file_based="1" />
<data_storage_product id="2" product_code="Oracle_9_i" is_network_svc="1" />
<data_link_product id="1" product_code="ODBC" />
</tools>
<sites>
<catalog_instance id="1" product="1" blueprint="1" name="test">
<user id="1" catalog="1" user_type="SCHEMA_OWNER" match_owner="1" name="ronsealy" password="K34dsD" />
<user id="2" catalog="1" user_type="DATA_EDITOR" name="joesmith" password="fdsKJ4" />
</catalog_instance>
<application_instance id="1" blueprint="1" name="test Setup">
<catalog_link_instance id="1" product="1" application="1" unrealized="1" target="1" local_dsn="test" />
</application_instance>
<application_instance id="2" blueprint="2" name="test People Watcher">
<catalog_link_instance id="2" product="1" application="2" unrealized="2" target="1" local_dsn="test" />
</application_instance>
<catalog_instance id="2" product="2" blueprint="1" name="demo">
<user id="3" catalog="2" user_type="SCHEMA_OWNER" match_owner="1" name="florence" password="0sfs8G" />
<user id="4" catalog="2" user_type="DATA_EDITOR" name="thainuff" password="9340sd" />
</catalog_instance>
<application_instance id="3" blueprint="1" name="demo Setup">
<catalog_link_instance id="3" product="1" application="3" unrealized="1" target="2" local_dsn="demo" />
</application_instance>
<application_instance id="4" blueprint="2" name="demo People Watcher">
<catalog_link_instance id="4" product="1" application="4" unrealized="2" target="2" local_dsn="demo" />
</application_instance>
</sites>
<circumventions />
</root>For some additional code samples, try looking at the various modules that sub-class or use SQL::SyntaxModel. They tend to implement or use wrappers that make for much more compact code.
Comparative SQL Code Examples Generated From a Model
SQL::SyntaxModel works like an XML DOM except that it is restricted to holding specific kinds of data, which resemble SQL statements. This part of the SYNOPSIS shows some actual SQL statements that can be generated from selected portions of the above model.
This first set of Nodes describes 2 domains and 1 table, all 3 of which are conceptually named schema objects.
<domain id="1" schema="1" name="entity_id" base_type="NUM_INT" num_precision="9" />
<domain id="2" schema="1" name="person_name" base_type="STR_CHAR" max_chars="100" char_enc="UTF8" />
<table id="1" schema="1" name="person">
<table_col id="1" table="1" name="person_id" domain="1" mandatory="1" default_val="1" auto_inc="1" />
<table_col id="2" table="1" name="name" domain="2" mandatory="1" />
<table_col id="3" table="1" name="father_id" domain="1" mandatory="0" />
<table_col id="4" table="1" name="mother_id" domain="1" mandatory="0" />
<table_ind id="1" table="1" name="primary" ind_type="UNIQUE">
<table_ind_col id="1" table_ind="1" table_col="1" />
</table_ind>
<table_ind id="2" table="1" name="fk_father" ind_type="FOREIGN" f_table="1">
<table_ind_col id="2" table_ind="2" table_col="3" f_table_col="1" />
</table_ind>
<table_ind id="3" table="1" name="fk_mother" ind_type="FOREIGN" f_table="1">
<table_ind_col id="3" table_ind="3" table_col="4" f_table_col="1" />
</table_ind>
</table>The above Node group has all the necessary details needed by external code to generate the following SQL statements. There are two versions of SQL given for the same task; the first one is for SQL-2003 compliant databases, that support DOMAIN schema objects; the second example is for older databases that do not. (Both of them use a MySQL extension AUTO_INCREMENT, but SQL generated for other databases would do the same thing in a different way.)
CREATE DOMAIN entity_id AS INTEGER(9);
CREATE DOMAIN person_name AS VARCHAR(100);
CREATE TABLE person (
person_id entity_id NOT NULL DEFAULT 1 AUTO_INCREMENT,
name person_name NOT NULL,
father_id entity_id NULL,
mother_id entity_id NULL,
CONSTRAINT PRIMARY KEY (person_id),
CONSTRAINT fk_father FOREIGN KEY (father_id) REFERENCES person (person_id),
CONSTRAINT fk_mother FOREIGN KEY (mother_id) REFERENCES person (person_id)
);
CREATE TABLE person (
person_id INTEGER(9) NOT NULL DEFAULT 1 AUTO_INCREMENT,
name VARCHAR(100) NOT NULL,
father_id INTEGER(9) NULL,
mother_id INTEGER(9) NULL,
CONSTRAINT PRIMARY KEY (person_id),
CONSTRAINT fk_father FOREIGN KEY (father_id) REFERENCES person (person_id),
CONSTRAINT fk_mother FOREIGN KEY (mother_id) REFERENCES person (person_id)
);Note that, regardless of which type of SQL is generated, the details for each data type, including its name, only need to be declared once, in 'domain' Nodes; if this one copy is changed, everything using it updates automatically.
This second set of Nodes describes a routine that takes 4 arguments (each of which is an actual argument if a named stored procedure is generated, or a named bind variable if un-named client-side SQL is generated) and performs an UPDATE query against one table record; the query takes 4 arguments, using one to match a record and 3 as new record column values to set.
<routine id="25" routine_type="ANONYMOUS" application="2" name="update_a_person">
<routine_arg id="26" routine="25" name="arg_person_id" var_type="SCALAR" domain="1" />
<routine_arg id="27" routine="25" name="arg_person_name" var_type="SCALAR" domain="2" />
<routine_arg id="28" routine="25" name="arg_father_id" var_type="SCALAR" domain="1" />
<routine_arg id="29" routine="25" name="arg_mother_id" var_type="SCALAR" domain="1" />
<view id="30" view_type="MATCH" name="update_a_person" routine="25">
<view_src id="31" view="30" name="person" match_table="1">
<view_src_col id="32" src="31" match_table_col="1" />
<view_src_col id="33" src="31" match_table_col="2" />
<view_src_col id="34" src="31" match_table_col="3" />
<view_src_col id="35" src="31" match_table_col="4" />
</view_src>
<view_expr id="36" expr_type="ARG" view="30" view_part="SET" set_view_col="33" routine_arg="27" />
<view_expr id="37" expr_type="ARG" view="30" view_part="SET" set_view_col="34" routine_arg="28" />
<view_expr id="38" expr_type="ARG" view="30" view_part="SET" set_view_col="35" routine_arg="29" />
<view_expr id="39" expr_type="SFUNC" view="30" view_part="WHERE" call_sfunc="EQ">
<view_expr id="40" expr_type="COL" p_expr="39" src_col="32" />
<view_expr id="41" expr_type="ARG" p_expr="39" routine_arg="26" />
</view_expr>
</view>
<routine_stmt id="42" routine="25" stmt_type="SPROC" call_sproc="UPDATE" view_for_dml="30" />
</routine>The above Node group, *together* with the previous Node group, has details to generate the following SQL statements. There are two versions of SQL given for the same task; the first one is for databases that support named bind variables, illustrated using the Oracle style of ':foo'; the second one is for those that require positional bind variables, illustrated with the DBI style of '?'. These two SQL variants are intended to be run by the SQL client.
UPDATE person
SET name = :arg_person_name, father_id = :arg_father_id, mother_id = :arg_mother_id
WHERE person_id = :arg_person_id;
UPDATE person
SET name = ?, father_id = ?, mother_id = ?
WHERE person_id = ?;Alternately, a stored procedure (and calls to it) can be generated from the same SSM Node set, if the routine_type attribute is PROCEDURE instead of ANONYMOUS. The two SQL variants are for new or old databases respectively, like the first example.
CREATE PROCEDURE update_a_person
(arg_person_id entity_id, arg_person_name person_name, arg_father_id entity_id, arg_mother_id entity_id)
BEGIN
UPDATE person
SET name = arg_person_name, father_id = arg_father_id, mother_id = arg_mother_id
WHERE person_id = arg_person_id;
END;
CREATE PROCEDURE update_a_person
(arg_person_id INTEGER(9), arg_person_name VARCHAR(100), arg_father_id INTEGER(9), arg_mother_id INTEGER(9))
BEGIN
UPDATE person
SET name = arg_person_name, father_id = arg_father_id, mother_id = arg_mother_id
WHERE person_id = arg_person_id;
END;To go with those, here are SQL statements to invoke the server-side stored procedures, with the two variants being named-vs-positional bind variables.
CALL update_a_person (:arg_person_id, :arg_person_name, :arg_father_id, :arg_mother_id);
CALL update_a_person (?, ?, ?, ?);Finally, all DROP statements can be generated from the same Nodes as CREATE.
Note that one key feature of SQL::SyntaxModel is that all of a model's pieces are linked by references rather than by name as in SQL itself. So if you wanted to change the name of a table column, such as 'person_name' to 'the_name', then you make the change in exactly one place and all SQL generated from the model will update, both the CREATE and UPDATE statements. Alternately, if you wanted to change the data type of person ids, then you only have to make a single change, such as by setting num_precision to 6. Alternately, if you wanted to change the order of the arguments for 'update_a_person', you only have to change the order the 'routine_arg' Nodes appear, and any calls to the procedure will automatically re-order any passed values in the generated SQL.
See also the separately distributed Rosetta::Utility::SQLBuilder module, which is a reference implementation of a SQL generator for SQL::SyntaxModel.
DESCRIPTION
The SQL::SyntaxModel Perl 5 module is intended to be a powerful but easy to use replacement for SQL strings (including support for placeholders), which you can use to make queries against a database. Each SQL::SyntaxModel object can represent a non-ambiguous structured command for a database to execute, or one can be a non-ambiguous structured description of a database schema object. This class supports all types of database operations, including both data manipulation and schema manipulation, as well as managing database instances and users. You typically construct a database query by setting appropriate attributes of these objects, and you execute a database query by evaluating the same attributes. SQL::SyntaxModel objects are designed to be equivalent to SQL in both the type of information they carry and in their conceptual structure. This is analagous to how XML DOMs are objects that are equivalent to XML strings, and they can be converted back and forth at will. If you know SQL, or even just relational database theory in general, then this module should be easy to learn.
SQL::SyntaxModels are intended to represent all kinds of SQL, both DML and DDL, both ANSI standard and RDBMS vendor extensions. Unlike basically all of the other SQL generating/parsing modules I know about, which are limited to basic DML and only support table definition DDL, this class supports arbitrarily complex select statements, with composite keys and unions, and calls to stored functions; this class can also define views and stored procedures and triggers. Some of the existing modules, even though they construct complete SQL, will take/require fragments of SQL as input (such as "where" clauses) By contrast, SQL::SyntaxModel takes no SQL fragments. All of its inputs are atomic, which means it is also easier to analyse the objects for implementing a wider range of functionality than previously expected; for example, it is much easier to analyse any select statement and generate update/insert/delete statements for the virtual rows fetched with it (a process known as updateable views).
Considering that each database product has its own dialect of SQL which it implements, you would have to code SQL differently depending on which database you are using. One common difference is the syntax for specifying an outer join in a select query. Another common difference is how to specify that a table column is an integer or a boolean or a character string. Moreover, each database has a distinct feature set, so you may be able to do tasks with one database that you can't do with another. In fact, some databases don't support SQL at all, but have similar features that are accessible thorough alternate interfaces. SQL::SyntaxModel is designed to represent a normalized superset of all database features that one may reasonably want to use. "Superset" means that if even one database supports a feature, you will be able to invoke it with this class. You can also reference some features which no database currently implements, but it would be reasonable for one to do so later. "Normalized" means that if multiple databases support the same feature but have different syntax for referencing it, there will be exactly one way of referring to it with SQL::SyntaxModel. So by using this class, you will never have to change your database-using code when moving between databases, as long as both of them support the features you are using (or they are emulated). That said, it is generally expected that if a database is missing a specific feature that is easy to emulate, then code which evaluates SQL::SyntaxModels will emulate it (for example, emulating "left()" with "substr()"); in such cases, it is expected that when you use such features they will work with any database. For example, if you want a model-specified BOOLEAN data type, you will always get it, whether it is implemented on a per-database-basis as a "boolean" or an "int(1)" or a "number(1,0)". Or a model-specified "STR_CHAR" data type you will always get it, whether it is called "text" or "varchar2" or "sql_varchar".
SQL::SyntaxModel is intended to be just a stateless container for database query or schema information. It does not talk to any databases by itself and it does not generate or parse any SQL; rather, it is intended that other third party modules or code of your choice will handle this task. In fact, SQL::SyntaxModel is designed so that many existing database related modules could be updated to use it internally for storing state information, including SQL generating or translating modules, and schema management modules, and modules which implement object persistence in a database. Conceptually speaking, the DBI module itself could be updated to take SQL::SyntaxModel objects as arguments to its "prepare" method, as an alternative (optional) to the SQL strings it currently takes. Code which implements the things that SQL::SyntaxModel describes can do this in any way that they want, which can mean either generating and executing SQL, or generating Perl code that does the same task and evaling it, should they want to (the latter can be a means of emulation). This class should make all of that easy.
SQL::SyntaxModel is especially suited for use with applications or modules that make use of data dictionaries to control what they do. It is common in applications that they interpret their data dictionaries and generate SQL to accomplish some of their work, which means making sure generated SQL is in the right dialect or syntax, and making sure literal values are escaped correctly. By using this module, applications can simply copy appropriate individual elements in their data dictionaries to SQL::SyntaxModel properties, including column names, table names, function names, literal values, bind variable names, and they don't have to do any string parsing or assembling.
Now, I can only imagine why all of the other SQL generating/parsing modules that I know about have excluded privileged support for more advanced database features like stored procedures. Either the authors didn't have a need for it, or they figured that any other prospective users wouldn't need it, or they found it too difficult to implement so far and maybe planned to do it later. As for me, I can see tremendous value in various advanced features, and so I have included privileged support for them in SQL::SyntaxModel. You simply have to work on projects of a significant size to get an idea that these features would provide a large speed, reliability, and security savings for you. Look at many large corporate or government systems, such as those which have hundreds of tables or millions of records, and that may have complicated business logic which governs whether data is consistent/valid or not. Within reasonable limits, the more work you can get the database to do internally, the better. I believe that if these features can also be represented in a database-neutral format, such as what SQL::SyntaxModel attempts to do, then users can get the full power of a database without being locked into a single vendor due to all their investment in vendor-specific SQL stored procedure code. If customers can move a lot more easily, it will help encourage database vendors to keep improving their products or lower prices to keep their customers, and users in general would benefit. So I do have reasons for trying to tackle the advanced database features in SQL::SyntaxModel.
CLASSES IN THIS MODULE
This module is implemented by several object-oriented Perl 5 packages, each of which is referred to as a class. They are: SQL::SyntaxModel (the module's name-sake), SQL::SyntaxModel::Container (aka Container, aka Model), and SQL::SyntaxModel::Node (aka Node).
While all 3 of the above classes are implemented in one module for convenience, you should consider all 3 names as being "in use"; do not create any modules or packages yourself that have the same names.
The Container and Node classes do most of the work and are what you mainly use. The name-sake class mainly exists to guide CPAN in indexing the whole module, but it also provides a set of stateless utility methods and constants that the other two classes inherit, and it provides a few wrapper functions over the other classes for your convenience; you never instantiate an object of SQL::SyntaxModel itself.
STRUCTURE
The internal structure of a SQL::SyntaxModel object is conceptually a cross between an XML DOM and an object-relational database, with a specific schema. This module is implemented with two main classes that work together, Containers and Nodes. The Container object is an environment or context in which Node objects usually live. A typical application will only need to create one Container object (returned by the module's 'new' function), and then a set of Nodes which live within that Container. The Nodes are related sometimes with single or multiple cardinality to each other.
SQL::SyntaxModel is expressly designed so that its data is easy to convert between different representations, mainly in-memory data structures linked by references, and multi-table record sets stored in relational databases, and node sets in XML documents. A Container corresponds to an XML document or a complete database, and each Node corresponds to an XML node or a database record. Each Node has a specific node_type (a case-sensitive string), which corresponds to a database table or an XML tag name. See the SQL::SyntaxModel::Language documentation file to see which ones exist. The node_type is set when the Node is created and it can not be changed later.
A Node has a specific set of allowed attributes that are determined by the node_type, each of which corresponds to a database table column or an XML node attribute. Every Node of a common node_type has a unique 'id' attribute (a positive integer) by which it is referenced; that attribute corresponds to the database table's single-column primary key. Each other Node attribute is either a scalar value of some data type, or an enumerated value, or a reference to another Node of a specific node_type, which has a foreign-key constraint on it. Foreign-key constraints are enforced by this module, so you will have to add Nodes in the appropriate order, just as when adding records to a database. Any Node which is referenced in an attribute (cited in a foreign-key constraint) of another is a parent of the other; as a corollary, the second Node is a child of the first. The order of child Nodes under a parent is the same as that in which the parent-child relationship was assigned, unless you have afterwards used the move_before_sibling() method to change this.
In versions of SQL::SyntaxModel prior to 0.10, some Node types also had explicit 'order' attributes so that if each Node was converted as-is to an RDBMS record, it would be possible to retrieve the records in the same sequence; this was useful in cases where the order of Nodes was important. These redundant 'order' attributes were eliminated in version 0.10 since their maintenance was making this module more difficult to use. As a consequence, you will now have to add a column yourself to maintain the sort order when converting Nodes to RDBMS records.
When SQL::SyntaxModels are converted to XML, one referencing attribute is given higher precedence than the others and becomes the single parent XML node. For example, the XML parent of a 'table_col' Node is always a 'table' Node, even though a 'domain' Node is also referenced. While Nodes of most types always have Nodes of a single other type as their parents, there are some exceptions. Nodes of certain types, such as view or *_expr, may have either another Node of the same type as itself, or of a specific other type as its parent, depending on the context; these Nodes form trees of their own type, and it is the root Node of each tree which has a different Node type as its parent.
Finally, any Node of certain types will always have a specific pseudo-node as its single parent, which it does not reference in an attribute, and which can not be changed. All 6 pseudo-nodes have no attributes, even 'id', and only one of each exists; they are created by default with the Container they are part of, forming the top 2 levels of the Node tree, and can not be removed. They are: 'root' (the single level-1 Node which is parent to the other pseudo-nodes but no normal Nodes), 'elements' (parent to 'domain' Nodes), 'blueprints' (parent to 'catalog' and 'application' Nodes), 'tools' (parent to 'data_storage_product' and 'data_link_product' Nodes), 'sites' (parent to 'catalog_instance' and 'application_instance' Nodes), and 'circumventions' (parent to 'sql_fragment' nodes). All other Node types have normal Nodes as parents.
You should look at the POD-only file named SQL::SyntaxModel::Language, which comes with this distribution. It serves to document all of the possible Node types, with attributes, constraints, and allowed relationships with other Node types. As the SQL::SyntaxModel class itself has very few properties and methods, all being highly generic (much akin to an XML DOM), the POD of this PM file will only describe how to use said methods, and will not list all the allowed inputs or constraints to said methods. With only simple guidance in SyntaxModel.pm, you should be able to interpret Language.pod to get all the nitty gritty details. You should also look at the tutorial or example files which will be in the distribution when ready. You could also learn something from the code samples inside other modules which sub-class this one.
FAULT TOLERANCE AND MULTI-THREADING SUPPORT
Disclaimer: The following claims assume that only this module's published API is used, and that you do not set object properties directly or call private methods, which Perl does not prevent. It also assumes that the module is bug free, and that any errors or warnings which appear while the code is running are thrown explicitly by this module as part of its normal functioning.
SQL::SyntaxModel is designed to ensure that the objects it produces are always internally consistant, and that the data they contain is always well-formed, regardless of the circumstances in which it is used. You should be able to fetch data from the objects at any time and that data will be self-consistant and well-formed.
This will not change regardless of what kind of bad input data you provide to object methods or module functions. Providing bad input data will cause the module to throw an exception; if you catch this and the program continues running (such as to chide the user and have them try entering correct input), then the objects will remain un-corrupted and able to accept new input or give proper output. In most cases, the object will be in the same state as it was before the public method was called with the bad input.
This module does not use package variables at all, besides constants like $VERSION, and all symbols ($@%) declared at file level are strictly constant value declarations. No object should ever step on another.
This module will allow a Node to be created piecemeal, such as when it is storing details gathered one at a time from the user, and during this time some mandatory Node properties may not be set, or pending links from this node to others may not be validated. However, until a Node has its required properties set and/or its Node links are validated, no references will be made to this Node from other Nodes; from their point of view it doesn't exist, and hence the other Nodes are all consistant.
SQL::SyntaxModel is explicitly not thread-aware (thread-safe); it contains no code to synchronize access to its objects' properties, such as semaphores or locks or mutexes. To internalize such things in an effective manner would have made the code a lot more complex than it is now, without any clear benefits. However, this module can (and should) be used in multi-threaded environments where the application/caller code takes care of synchronizing access to its objects, especially if the application uses coarse-grained read or write locks.
The author's expectation is that this module will be mainly used in circumstances where the majority of actions are reads, and there are very few writes, such as with a data dictionary; perhaps all the writes on an object may be when it is first created. An application thread would obtain a read lock/semaphore on a Container object during the period for which it needs to ensure read consistency; it would block write lock attempts but not other read locks. It would obtain a write lock during the (usually short) period it needs to change something, which blocks all other lock attempts (for read or write).
An example of this is a web server environment where each page request is being handled by a distinct thread, and all the threads share one SQL::SyntaxModel object; normally the object is instantiated when the server starts, and the worker threads then read from it for guidance in using a common database. Occasionally a thread will want to change the object, such as to correspond to a simultaneous change to the database schema, or to the web application's data dictionary that maps the database to application screens. Under this situation, the application's definitive data dictionary (stored partly or wholly in a SQL::SyntaxModel) can occupy one place in RAM visible to all threads, and each thread won't have to keep looking somewhere else such as in the database or a file to keep up with the definitive copy. (Of course, any *changes* to the in-memory data dictionary should see a corresponding update to a non-volatile copy, like in an on-disk database or file.)
Note that, while a nice thing to do may be to manage a course-grained lock in SQL::SyntaxModel, with the caller invoking lock_to_read() or lock_to_write() or unlock() methods on it, Perl's thread->lock() mechanism is purely context based; the moment lock_to_...() returns, the object has unlocked again. Of course, if you know a clean way around this, I would be happy to hear it.
NODE EVOLUTION STATES
A SQL::SyntaxModel Node object always exists in one of 3 official ordered states (which can conceptually be divided further into more states). For now we can call them "Alone" (1), "At Home" (2), and "Well Known" (3). (Hey, that rhymes!) The set of legal operations you can perform on a Node are different depending on its state, and a Node can only transition between adjacent-numbered states one at a time.
When a new Node is created, using new_node(), it starts out "Alone"; it does *not* live in a Container, and it is illegal to have any actual (Perl) references between it and any other Node. Nodes in this state can be built (have their Node Id and other attributes set or changed) piecemeal with the least processing overhead, and can be moved or exist independently of anything else that SQL::SyntaxModel manages. An "Alone" Node does not need to have its Node Id set. Any Node attributes which are conceptually references to other Nodes are stored and read as Id numbers when the Node is "Alone"; also, no confirmation has yet taken place that the referenced Nodes actually exist yet. A Node may only be individually deleted when it is "Alone"; in this state it will be garbage collected like any Perl variable when your own reference to it goes away.
When you invoke the put_in_container() method on an "Alone" Node, giving it a Container object as an argument, the Node will transition to the "At Home" state; you can move from "At Home" to "Alone" using the complementary take_from_container() method. An "At Home" Node lives in a Container, and any attributes which refer to other Nodes now must be actual references, where the existence of the other Node in the same Container is confirmed. If any conceptual references are set in a Node while it is "Alone", these will be converted into actual references by put_in_container(), which will fail if any can't be found. take_from_container() will replace references with Node Ids. A Node can only link to a Node in the same Container as itself. While a Node in "At Home" status can link to other Nodes, those Nodes can not link back to an "At Home" Node in their own child list; from their point of view, the "At Home" Node doesn't exist. In addition, an "At Home" Node can not have children of its own; it can not be referenced by any other Nodes.
When you invoke the add_reciprocal_links() method on an "At Home" Node, the Node will transition to the "Well Known" state; any other Nodes that this one references will now link back to it in their own child lists. The complementary remove_reciprocal_links() method will break those return links and transition a "Well Known" Node to an "At Home" one. A "Well Known" Node is also allowed to have children of its own.
Testing for the existence of mandatory Node attribute values is separate from the official Node state and can be invoked on a Node at any time. None of the official Node states themselves will assert that any mandatory attributes are populated. This testing is separate partly to make it easy for you to build Nodes piecemeal, though there are other practical reasons for it.
Note that all typical Node attributes can be read, set, replaced, or cleared at any time regardless of the Node state; you can set them all either when the Node is "Alone" or when it is "Well Known", as is your choice. However, the Node Id must always have a value when the Node is in a Container; if you want to make a Node "Well Known" as early as possible, you simply have to set its Node Id first.
SYNTAX
This class does not export any functions or methods, so you need to call them using object notation. This means using Class->function() for functions and $object->method() for methods. If you are inheriting this class for your own modules, then that often means something like $self->method().
All SQL::SyntaxModel functions and methods are either "getters" (which read and return or generate values but do not change the state of anything) or "setters" (which change the state of something but do not return anything on success); none do getting or setting conditionally based on their arguments. While this means there are more methods in total, I see this arrangement as being more stable and reliable, plus each method is simpler and easier to understand or use; argument lists and possible return values are also less variable and more predictable.
All "setter" functions or methods which are supposed to change the state of something will throw an exception on failure (usually from being given bad arguments); on success, they officially have no return values. A thrown exception will always include details of what went wrong (and where and how) in a machine-readable (and generally human readable) format, so that calling code which catches them can recover gracefully. The methods are all structured so that they check all preconditions prior to changing any state information, and so one can assume that upon throwing an exception, the Node and Container objects are in a consistent or recoverable state at worst, and are completely unchanged at best.
All "getter" functions or methods will officially return the value or construct that was asked for; if said value doesn't (yet or ever) exist, then this means the Perl "undefined" value. When given bad arguments, generally this module's "information" functions will return the undefined value, and all the other functions/methods will throw an exception like the "setter" functions do.
Generally speaking, if SQL::SyntaxModel throws an exception, it means one of two things: 1. Your own code is not invoking it correctly, meaning you have something to fix; 2. You have decided to let it validate some of your input data for you (which is quite appropriate).
Note also that SQL::SyntaxModel is quite strict in its own argument checking, both for internal simplicity and robustness, and so that code which *reads* data from it can be simpler. If you want your own program to be more liberal in what input it accepts, then you will have to bear the burden of cleaning up or interpreting that input, or delegating such work elsewhere. (Or perhaps someone may want to make a wrapper module to do this?)
CONSTRUCTOR WRAPPER FUNCTIONS
These functions are stateless and can be invoked off of either the module name, or any package name in this module, or any object created by this module; they are thin wrappers over other methods and exist strictly for convenience.
new_container()
my $model = SQL::SyntaxModel->new_container();
my $model2 = SQL::SyntaxModel::Container->new_container();
my $model3 = SQL::SyntaxModel::Node->new_container();
my $model4 = $model->new_container();
my $model5 = $node->new_container();This function wraps SQL::SyntaxModel::Container->new().
new_node( NODE_TYPE )
my $node = SQL::SyntaxModel->new_node( 'table' );
my $node2 = SQL::SyntaxModel::Container->new_node( 'table' );
my $node3 = SQL::SyntaxModel::Node->new_node( 'table' );
my $node4 = $model->new_node( 'table' );
my $node5 = $node->new_node( 'table' );This function wraps SQL::SyntaxModel::Node->new( NODE_TYPE ).
CONTAINER CONSTRUCTOR FUNCTIONS AND METHODS
This function/method is stateless and can be invoked off of either the Container class name or an existing Container object, with the same result.
new()
my $model = SQL::SyntaxModel::Container->new();
my $model2 = $model->new();This "getter" function/method will create and return a single SQL::SyntaxModel::Container (or subclass) object.
CONTAINER OBJECT METHODS
These methods are stateful and may only be invoked off of Container objects.
destroy()
$model->destroy();This "setter" method will destroy the Container object that it is invoked from, and it will also destroy all of the Nodes inside that Container. This method exists because all Container objects (having 1 or more Node) contain circular references between the Container and all of its Nodes. You need to invoke this method when you are done with a Container, or you will leak the memory it uses when your external references to it go out of scope. This method can be invoked at any time and will not throw any exceptions. When it has completed, all external references to the Container or any of its Nodes will each point to an empty (but still blessed) Perl hash. See the CAVEATS documentation.
get_node( NODE_TYPE, NODE_ID )
my $catalog_node = $model->get_node( 'catalog', 1 );This "getter" method returns a reference to one of this Container's member Nodes, which has a Node Type of NODE_TYPE, and a Node Id of NODE_ID. You may not request a pseudo-node (it doesn't actually exist).
get_child_nodes([ NODE_TYPE ])
my $ra_node_list = $model->get_child_nodes();
my $ra_node_list = $model->get_child_nodes( 'catalog' );This "getter" method returns a list of this Container's child Nodes, in a new array ref. A Container's child Nodes are defined as being all Nodes in the Container whose Node Type defines them as always having a pseudo-Node parent. If the optional argument NODE_TYPE is defined, then only child Nodes of that Node Type are returned; otherwise, all child Nodes are returned. All Nodes are returned in the same order they were added.
with_all_nodes_test_mandatory_attributes()
my $model->with_all_nodes_test_mandatory_attributes();This "getter" method implements a type of deferrable data validation. It will iterate through every Node in this Container and invoke its test_mandatory_attributes() method; the order that the Nodes are tested is not defined, but every one will get tested regardless of its state or connectedness with the other Nodes. That said, a failure with any one Node will cause the testing of the whole set to abort, as the offending Node throws an exception which this method doesn't catch; any untested Nodes could also have failed. Only when you can call this method without any exceptions being thrown will all Nodes have passed the tests.
NODE CONSTRUCTOR FUNCTIONS AND METHODS
This function/method is stateless and can be invoked off of either the Node class name or an existing Node object, with the same result.
new( NODE_TYPE )
my $node = SQL::SyntaxModel::Node->new( 'table' );
my $node2 = $node->new( 'table' );This "getter" function/method will create and return a single SQL::SyntaxModel::Node (or subclass) object whose Node Type is given in the NODE_TYPE (enum) argument, and all of whose other properties are defaulted to an "empty" state. A Node's type can only be set on instantiation and can not be changed afterwards; only specific values are allowed, which you can see in the SQL::SyntaxModel::Language documentation file. This new Node does not yet live in a Container, and will have to be put in one later before you can make full use of it. However, you can read or set or clear any or all of this new Node's attributes (including the Node Id) prior to putting it in a Container, making it easy to build one piecemeal before it is actually "used". A Node can not have any actual Perl references between it and other Nodes until it is in a Container, and as such you can delete it simply by letting your own reference to it be garbage collected.
NODE OBJECT METHODS
These methods are stateful and may only be invoked off of Node objects. For some of these, it doesn't matter whether the Node is in a Container or not, nor whether its links to other Nodes are reciprocated or not. For others, one or both of these conditions must be true or false for the method to be invoked, or it will throw an exception (like for bad input).
delete_node()
This "setter" method will destroy the Node object that it is invoked from, if it can. You are only allowed to delete Nodes that are not inside Containers, and which don't have child Nodes; failing this, you must remove the children and then take this Node from its Container first. Technically, this method doesn't actually do anything (pure-Perl version) other than validate that you are allowed to delete; when said conditions are met, the Node will be garbage collected as soon as you lose your reference to it.
get_node_type()
my $type = $node->get_node_type();This "getter" method returns the Node Type scalar (enum) property of this Node. You can not change this property on an existing Node, but you can set it on a new one.
get_node_id()
This "getter" method will return the integral Node Id property of this Node, if it has one.
clear_node_id()
This "setter" method will erase this Node's Id property if it can. A Node's Id may only be cleared if the Node is not in a Container.
set_node_id( NEW_ID )
This "setter" method will set or replace this Node's Id property if it can. If this Node is in a Container, then the replacement will fail if some other Node with the same Node Type and Node Id already exists in the same Container.
expected_literal_attribute_type( ATTR_NAME )
This "getter" method will return an enumerated value that explains which literal data type that values for this Node's literal attribute named in the ATTR_NAME argument must be.
get_literal_attribute( ATTR_NAME )
This "getter" method will return the value for this Node's literal attribute named in the ATTR_NAME argument.
get_literal_attributes()
This "getter" method will fetch all of this Node's literal attributes, returning them in a Hash ref.
clear_literal_attribute( ATTR_NAME )
This "setter" method will clear this Node's literal attribute named in the ATTR_NAME argument.
clear_literal_attributes()
This "setter" method will clear all of this Node's literal attributes.
set_literal_attribute( ATTR_NAME, ATTR_VALUE )
This "setter" method will set or replace this Node's literal attribute named in the ATTR_NAME argument, giving it the new value specified in ATTR_VALUE.
set_literal_attributes( ATTRS )
This "setter" method will set or replace multiple Node literal attributes, whose names and values are specified by keys and values of the ATTRS hash ref argument; this method will invoke set_literal_attribute() for each key/value pair.
expected_enumerated_attribute_type( ATTR_NAME )
This "getter" method will return an enumerated value that explains which enumerated data type that values for this Node's enumerated attribute named in the ATTR_NAME argument must be.
get_enumerated_attribute( ATTR_NAME )
This "getter" method will return the value for this Node's enumerated attribute named in the ATTR_NAME argument.
get_enumerated_attributes()
This "getter" method will fetch all of this Node's enumerated attributes, returning them in a Hash ref.
clear_enumerated_attribute( ATTR_NAME )
This "setter" method will clear this Node's enumerated attribute named in the ATTR_NAME argument.
clear_enumerated_attributes()
This "setter" method will clear all of this Node's enumerated attributes.
set_enumerated_attribute( ATTR_NAME, ATTR_VALUE )
This "setter" method will set or replace this Node's enumerated attribute named in the ATTR_NAME argument, giving it the new value specified in ATTR_VALUE.
set_enumerated_attributes( ATTRS )
This "setter" method will set or replace multiple Node enumerated attributes, whose names and values are specified by keys and values of the ATTRS hash ref argument; this method will invoke set_enumerated_attribute() for each key/value pair.
expected_node_ref_attribute_type( ATTR_NAME )
This "getter" method will return an enumerated value that explains which Node Type that values for this Node's node attribute named in the ATTR_NAME argument must be.
get_node_ref_attribute( ATTR_NAME )
This "getter" method will return the value for this Node's node attribute named in the ATTR_NAME argument. The value will be a Node ref if the current Node is in a Container, and an Id number if it isn't.
get_node_ref_attributes()
This "getter" method will fetch all of this Node's node attributes, returning them in a Hash ref. The values will be Node refs if the current Node is in a Container, and Id numbers if it isn't.
clear_node_ref_attribute( ATTR_NAME )
This "setter" method will clear this Node's node attribute named in the ATTR_NAME argument. If the other Node being referred to has a reciprocal link to the current one in its child list, that will also be cleared.
clear_node_ref_attributes()
This "setter" method will clear all of this Node's node attributes; see the clear_node_ref_attribute() documentation for the semantics.
set_node_ref_attribute( ATTR_NAME, ATTR_VALUE )
This "setter" method will set or replace this Node's node attribute named in the ATTR_NAME argument, giving it the new value specified in ATTR_VALUE (if it is different). If the attribute was previously valued, this method will first invoke clear_node_ref_attribute() on it. When setting a new value, if the current Node is in a Container and expects Nodes it links to reciprocate, then it will also add the current Node to the other Node's child list.
set_node_ref_attributes( ATTRS )
This "setter" method will set or replace multiple Node node attributes, whose names and values are specified by keys and values of the ATTRS hash ref argument; this method will invoke set_node_ref_attribute() for each key/value pair.
expected_attribute_major_type( ATTR_NAME )
This "getter" method will return an enumerated value that explains which major data type that values for this Node's attribute named in the ATTR_NAME argument must be. There are 4 possible return values: 'ID' (the Node Id), 'LITERAL' (a literal attribute), 'ENUM' (an enumerated attribute), and 'NODE' (a node ref attribute).
get_attribute( ATTR_NAME )
my $curr_val = $node->get_attribute( 'name' );This "getter" method will return the value for this Node's attribute named in the ATTR_NAME argument.
get_attributes()
my $rh_attrs = $node->get_attributes();This "getter" method will fetch all of this Node's attributes, returning them in a Hash ref.
clear_attribute( ATTR_NAME )
This "setter" method will clear this Node's attribute named in the ATTR_NAME argument.
clear_attributes()
This "setter" method will clear all of this Node's attributes.
set_attribute( ATTR_NAME, ATTR_VALUE )
This "setter" method will set or replace this Node's attribute named in the ATTR_NAME argument, giving it the new value specified in ATTR_VALUE.
set_attributes( ATTRS )
$node->set_attributes( $rh_attrs );This "setter" method will set or replace multiple Node attributes, whose names and values are specified by keys and values of the ATTRS hash ref argument; this method will invoke set_attribute() for each key/value pair.
get_parent_node_attribute_name()
This "getter" method returns the name of this Node's node attribute which is designated to reference this Node's primary parent Node, if there is one.
get_parent_node()
my $parent = $node->get_parent_node();This "getter" method returns the primary parent Node of the current Node, if there is one. The semantics are like "if the current Node is in a Container and its 'parent node attribute name' is defined, then return the Node ref value of the named node attribute, if it has one".
clear_parent_node_attribute_name()
This "setter" method will clear this Node's 'parent node attribute name' property, if it has one. The actual node attribute being referred to is not affected.
set_parent_node_attribute_name( ATTR_NAME )
This "setter" method will set or replace this Node's 'parent node attribute name' property, giving it the new value specified in ATTR_NAME. No actual node attribute is affected. Note that only a subset (usually one) of a Node's node attributes may be named as the holder of its primary parent.
estimate_parent_node_attribute_name( NEW_PARENT[, ONLY_NOT_VALUED] )
This "getter" method will try to find a way to make the Node given in its NEW_PARENT argument into the primary parent of the current Node. It returns the name of the first appropriate Node attribute which takes a Node of the same Node Type as NEW_PARENT; if one can not be found, the undefined value is returned. By default, the current value of the found attribute is ignored; but if the optional argument ONLY_NOT_VALUED is true, then an otherwise acceptible attribute name will not be returned if it already has a value.
get_container()
my $model = $node->get_container();This "getter" method returns the Container object which this Node lives in, if any.
put_in_container( NEW_CONTAINER )
This "setter" method will put the current Node into the Container given as the NEW_CONTAINER argument if it can, which moves the Node from "Alone" to "At Home" status.
take_from_container()
This "setter" method will take the current Node from its Container if it can, which moves the Node from "At Home" to "Alone" status.
are_reciprocal_links()
This "getter" method returns a true boolean value if the current Node is in "Well Known" status, and false otherwise.
add_reciprocal_links()
This "setter" method will move the current Node from "At Home" to "Well Known" status if possible.
remove_reciprocal_links()
This "setter" method will move the current Node from "Well Known" to "At Home" status if possible.
move_before_sibling( SIBLING[, PARENT] )
This "setter" method allows you to change the order of child Nodes under a common parent Node; specifically, it moves the current Node to a position just above/before the sibling Node specified in the SIBLING Node ref argument, if it can. You can only invoke it on a Node that is "Well Known", since that is the only time it exists in its parent's child list at all. Since a Node can have multiple parent Nodes (and the sibling likewise), the optional PARENT argument lets you specify which parent's child list you want to move in; if you do not provide an PARENT value, then the current Node's primary parent Node is used, if possible. This method will throw an exception if the current Node and the specified sibling or parent Nodes are not appropriately related to each other (parent <-> child). If you want to move the current Node to follow the sibling instead, then invoke this method on the sibling.
get_child_nodes([ NODE_TYPE ])
my $ra_node_list = $node->get_child_nodes();
my $ra_node_list = $node->get_child_nodes( 'table' );This "getter" method returns a list of this object's child Nodes, in a new array ref. If the optional argument NODE_TYPE is defined, then only child Nodes of that Node Type are returned; otherwise, all child Nodes are returned. All Nodes are returned in the same order they were added.
add_child_node( NEW_CHILD )
$node->add_child_node( $child );This "setter" method allows you to add a new child Node to this object, which is provided as the single NEW_CHILD Node ref argument. The new child Node is appended to the list of existing child Nodes, and the current Node becomes the new or first primary parent Node of NEW_CHILD.
add_child_nodes( LIST )
$model->add_child_nodes( [$child1,$child2] );
$model->add_child_nodes( $child );This "setter" method takes an array ref in its single LIST argument, and calls add_child_node() for each element found in it.
test_mandatory_attributes()
This "getter" method implements a type of deferrable data validation. It will look at all of this Node's attributes which must have a value set before this Node is ready to be used, and throw an exception if any are not. This method confines its tests to the specific Node being tested, and does not perform more complicated tests that require looking at multiple Nodes together.
CONTAINER OR NODE METHODS FOR DEBUGGING
The following 3 "getter" methods can be invoked either on Container or Node objects, and will return a tree-arranged structure having the contents of a Node and all its children (to the Nth generation). The previous statement assumes that all the 'children' have a true are_reciprocal_links property, which means that a Node's parent is aware of it; if that property is false for a Node, the assumption is that said Node is still being constructed, and neither it nor its children will be included in the output. If you invoke the 3 methods on a Node, then that Node will be the root of the returned structure. If you invoke them on a Container, then a few pseudo-nodes will be output with all the normal Nodes in the Container as their children.
get_all_properties()
$rh_node_properties = $node->get_all_properties();
$rh_node_properties = $container->get_all_properties();This method returns a deep copy of all of the properties of this object as non-blessed Perl data structures. These data structures are also arranged in a tree, but they do not have any circular references. The main purpose, currently, of get_all_properties() is to make it easier to debug or test this class; it makes it easier to see at a glance whether the other class methods are doing what you expect. The output of this method should also be easy to serialize or unserialize to strings of Perl code or xml or other things, should you want to compare your results easily by string compare (see "get_all_properties_as_perl_str()" and "get_all_properties_as_xml_str()").
get_all_properties_as_perl_str([ NO_INDENTS ])
$perl_code_str = $container->get_all_properties_as_perl_str();
$perl_code_str = $container->get_all_properties_as_perl_str( 1 );
$perl_code_str = $node->get_all_properties_as_perl_str();
$perl_code_str = $node->get_all_properties_as_perl_str( 1 );This method is a wrapper for get_all_properties() that serializes its output into a pretty-printed string of Perl code, suitable for humans to read. You should be able to eval this string and produce the original structure. By default, contents of lists are indented under the lists they are in (easier to read); if the optional boolean argument NO_INDENTS is true, then all output lines will be flush with the left, saving a fair amount of memory in what the resulting string consumes. (That said, even the indents are tabs, which take up much less space than multiple spaces per indent level.)
get_all_properties_as_xml_str([ NO_INDENTS ])
$xml_doc_str = $container->get_all_properties_as_xml_str();
$xml_doc_str = $container->get_all_properties_as_xml_str( 1 );
$xml_doc_str = $node->get_all_properties_as_xml_str();
$xml_doc_str = $node->get_all_properties_as_xml_str( 1 );This method is a wrapper for get_all_properties() that serializes its output into a pretty-printed string of XML, suitable for humans to read. By default, child nodes are indented under their parent nodes (easier to read); if the optional boolean argument NO_INDENTS is true, then all output lines will be flush with the left, saving a fair amount of memory in what the resulting string consumes. (That said, even the indents are tabs, which take up much less space than multiple spaces per indent level.)
INFORMATION FUNCTIONS AND METHODS
These "getter" functions/methods are all intended for use by programs that want to dynamically interface with SQL::SyntaxModel, especially those programs that will generate a user interface for manual editing of data stored in or accessed through SQL::SyntaxModel constructs. It will allow such programs to continue working without many changes while SQL::SyntaxModel itself continues to evolve. In a manner of speaking, these functions/methods let a caller program query as to what 'schema' or 'business logic' drive this class. These functions/methods are all deterministic and stateless; they can be used in any context and will always give the same answers from the same arguments, and no object properties are used. You can invoke them from any kind of object that SQL::SyntaxModel implements, or straight off of the class name itself, like a 'static' method. All of these functions return the undefined value if they match nothing.
valid_enumerated_types([ ENUM_TYPE ])
This function by default returns a list of the valid enumerated types that SQL::SyntaxModel recognizes; if the optional ENUM_TYPE argument is given, it just returns true if that matches a valid type, and false otherwise.
valid_enumerated_type_values( ENUM_TYPE[, ENUM_VALUE] )
This function by default returns a list of the values that SQL::SyntaxModel recognizes for the enumerated type given in the ENUM_TYPE argument; if the optional ENUM_VALUE argument is given, it just returns true if that matches an allowed value, and false otherwise.
valid_node_types([ NODE_TYPE ])
This function by default returns a list of the valid Node Types that SQL::SyntaxModel recognizes; if the optional NODE_TYPE argument is given, it just returns true if that matches a valid type, and false otherwise.
valid_node_type_literal_attributes( NODE_TYPE[, ATTR_NAME] )
This function by default returns a Hash ref where the keys are the names of the literal attributes that SQL::SyntaxModel recognizes for the Node Type given in the NODE_TYPE argument, and where the values are the literal data types that values for those attributes must be; if the optional ATTR_NAME argument is given, it just returns the literal data type for the named attribute.
valid_node_type_enumerated_attributes( NODE_TYPE[, ATTR_NAME] )
This function by default returns a Hash ref where the keys are the names of the enumerated attributes that SQL::SyntaxModel recognizes for the Node Type given in the NODE_TYPE argument, and where the values are the enumerated data types that values for those attributes must be; if the optional ATTR_NAME argument is given, it just returns the enumerated data type for the named attribute.
valid_node_type_node_ref_attributes( NODE_TYPE[, ATTR_NAME] )
This function by default returns a Hash ref where the keys are the names of the node attributes that SQL::SyntaxModel recognizes for the Node Type given in the NODE_TYPE argument, and where the values are the Node Types that values for those attributes must be; if the optional ATTR_NAME argument is given, it just returns the Node Type for the named attribute.
major_type_of_node_type_attribute( NODE_TYPE, ATTR_NAME )
This "getter" function returns the major type for the attribute of NODE_TYPE Nodes named ATTR_NAME, which is one of 'ID', 'LITERAL', 'ENUM' or 'NODE'.
valid_node_type_parent_attribute_names( NODE_TYPE[, ATTR_NAME] )
This function by default returns an Array ref which lists the names of the node attributes that are allowed to reference the primary parent of a Node whose type is specified in the NODE_TYPE argument; if the optional ATTR_NAME argument is given, it just returns true the named attribute may reference the primary parent of a NODE_TYPE Node.
node_types_with_pseudonode_parents([ NODE_TYPE ])
This function by default returns a Hash ref where the keys are the names of the Node Types whose primary parents can only be pseudo-nodes, and where the values name the pseudo-nodes they are the children of; if the optional NODE_TYPE argument is given, it just returns the pseudo-node for that Node Type.
mandatory_node_type_literal_attribute_names( NODE_TYPE[, ATTR_NAME] )
This function by default returns a list of the mandatory literal attributes of the Node Type specified in the NODE_TYPE argument; if the optional ATR_NAME argument is given, it just returns true if that attribute is mandatory.
mandatory_node_type_enumerated_attribute_names( NODE_TYPE[, ATTR_NAME] )
This function by default returns a list of the mandatory enumerated attributes of the Node Type specified in the NODE_TYPE argument; if the optional ATR_NAME argument is given, it just returns true if that attribute is mandatory.
mandatory_node_type_node_ref_attribute_names( NODE_TYPE[, ATTR_NAME] )
This function by default returns a list of the mandatory node attributes of the Node Type specified in the NODE_TYPE argument; if the optional ATR_NAME argument is given, it just returns true if that attribute is mandatory.
BUGS
This module is currently in pre-alpha development status, meaning that some parts of it will be changed in the near future, perhaps in incompatible ways; however, I believe that the largest short-term changes are already done. This module will indeed execute and do a variety of things, but it isn't yet recommended for any kind of serious use. The current state is analagous to 'developer releases' of operating systems; you can study it with the intent of using it in the future, but you should hold off writing any volume of code against it which you aren't prepared to rewrite later as the API changes. Also, the module hasn't been tested as much as I would like, but it has tested the more commonly used areas. All of the code included with the other modules that sub-class this one has been executed, which tests most internal functions and data. All of this said, I plan to move this module into alpha development status within the next few releases, once I start using it in a production environment myself.
CAVEATS
All SQL::SyntaxModel::Container objects contain circular references by design (or more specifically, when 1 or more Node is in one). When you are done with a Container object, you should explicitly call its "destroy()" method prior to letting your references to it go out of scope, or you will leak the memory it used. Up to and including SQL::SyntaxModel v0.12 I had implemented a cludge that wrapped the actual Container object in a second object that would automatically destroy its contents when it went out of scope. While that saved users from doing manual destruction, it introduced potentially worse problems, such as the Container being destroyed too early (and it added complexity regardless); as of v0.13 I did away with the cludge.
SEE ALSO
perl(1), SQL::SyntaxModel::L::*, SQL::SyntaxModel::Language, SQL::SyntaxModel::API_C, Rosetta::Utility::SQLBuilder, SQL::SyntaxModel::ByTree, SQL::SyntaxModel::SkipID, Rosetta, Rosetta::Framework, DBI, SQL::Statement, SQL::Translator, SQL::YASP, SQL::Generator, SQL::Schema, SQL::Abstract, SQL::Snippet, SQL::Catalog, DB::Ent, DBIx::Abstract, DBIx::AnyDBD, DBIx::DBSchema, DBIx::Namespace, DBIx::SearchBuilder, TripleStore, and various other modules.