Now with content!

[dbsrgits/SQL-Translator.git] / lib / SQL / Translator / Manual.pod

=head1 NAME

SQL::Translator::Manual

=head1 SYNOPSIS

SQL::Translator (AKA "SQLFairy") is a collection of modules for
transforming (mainly) SQL DDL files into a variety of other formats,
including other SQL dialects, documentation, images, and code.  In
this manual, we will attempt to address how to use SQLFairy for common
tasks.  For a lower-level discussion of how the code works, please
read the documentation for L<SQL::Translator>.

It may prove helpful to have a general understanding of the SQLFairy
code before continuing.  The code can be broken into three conceptual
groupings:

=over 4

=item * Parsers

The parsers are responsible for reading the input files and describing 
them to the Schema object middleware.

=item * Producers

The producers create the output as described by the Schema middleware.

=item * Schema objects

The Schema objects bridge the communication between the Parsers and
Producers by representing any parsed file through a standard set of
generic objects to represent concepts like Tables, Fields (columns),
Indices, Constraints, etc.

=back

It's not necessary to understand how to write or manipulate any
of these for most common tasks, but you should aware of the concepts
as they will be referenced later in this document. 

=head1 SQLFAIRY SCRIPTS

Most common tasks can be accomplished through the use of the script
interfaces to the SQL::Translator code.  All SQLFairy scripts begin
with "sqlt."  Here are the scripts and a description of what they each
do:

=over 4

=item * sqlt

This is the main interface for text-to-text translations, e.g.,
converting a MySQL schema to Oracle.  

=item * sqlt-diagram

This is a tailored interface for the Diagram producer and its many
myriad options.

=item * sqlt-diff

This script will examine two schemas and report the SQL commands
(ALTER, CREATE) needed to turn the first schema into the second. 

=item * sqlt-dumper

This script generates a Perl script that can be used to connect to a
database and dump the data in each table in different formats, similar
to the "mysqldump" program.

=item * sqlt-graph

This is an interface to the GraphViz visualization tool and its myriad
options.

=item * sqlt.cgi

This is a CGI script that presents an HTML form for uploading or
pasting a schema and choosing an output and the output options.

=back

To read the full documentation for each script, use "perldoc" (or
execute any of the command-line scripts with the "--help" flag).

=head1 CONVERTING SQL DIALECTS

Probably the most common task SQLFairy is used for is to convert one
dialect of SQL to another.  If you have a text description of an SQL
database (AKA a "DDL" -- "Data Definition Language"), then you should
use the "sqlt" script with switches to indicate the parser and
producer and the name of the text file as the final argument.  For
example, to convert the "foo.sql" MySQL schema to a version suitable
for PostgreSQL, you would do the following:

  $ sqlt -f MySQL -t PostgreSQL foo.sql > foo-pg.sql

The "from" and "to" options are case-sensitive and must match exactly
the names of the Parser and Producer classes in SQL::Translator.  For
a complete listing of your options, execute "sqlt" with the "--list"
flag.

=head1 EXTRACT SQL SCHEMAS DIRECTLY FROM DATABASE

It is possible to extract some schemas directly from the database
without parsing a text file (the "foo.sql" in the above example).
This can prove significantly faster than parsing a text file.  To
do this, use the "DBI" parser and provide the necessary arguments to
connect to the database and indicate the producer class, like so:

  $ sqlt -f DBI --dsn dbi:mysql:FOO --db-user guest \
    --db-password p4ssw0rd -t PostgreSQL > foo

The "--list" option to "sqlt" will show the databases supported by
DBI parsers.

=head1 HANDLING NON-SQL DATA

Certain structured document formats can be easily thought of as
tables.  SQLFairy can parse Microsoft Excel spreadsheets and
arbitrarily delimited text files just as if they were schemas which
contained only one table definition.  The column names are normalized
to something sane for most databases (whitespace is converted to
underscores and non-word characters are removed), and the data in each
field is scanned to determine the appropriate data type (character,
integer, or float) and size.  For instance, to convert a
comma-separated file to an SQLite database, do the following:

  $ sqlt -f xSV --fs ',' -t SQLite foo.csv > foo-sqlite.sql

Additionally, there are non-SQL represenations of relational schemas
such as XML and XMI.  Currently the XMI support in SQLFairy is
experimental and not released.  Additionally, the only XML supported
is our own version;  however, it would be fairly easy to add an XML
parser for something like the TorqueDB (http://db.apache.org/torque/)
project.  The actual parsing of XML should be trivial given the number
of XML parsers available, so all that would be left would be to map
the specific concepts in the source file to the Schema objects in
SQLFairy.  

To convert a schema in SQLFairy's XML dialect to Oracle, do the following:

  $ sqlt -f XML-SQLFairy -t Oracle foo.xml > foo-oracle.sql 

=head1 SERIALIZING SCHEMAS

via XML, YAML and Storable
Parsing a schema is generally the most computationally expensive
operation performed by SQLFairy, so it may behoove you to serialize a
parsed schema if you need to perform repeated conversions.  For
example, as part of a build process the author converts a MySQL schema
first to YAML, then to PostgreSQL, Oracle, SQLite and Sybase.
Additionally, a variety of documention in HTML and images is produced.
This can be accomplished like so:

  $ sqlt -f MySQL -t YAML schema-mysql.sql > schema.yaml
  $ sqlt -f YAML -t Oracle schema.yaml > schema-oracle.sql
  $ sqlt -f YAML -t PostgreSQL schema.yaml > schema-postgresql.sql
  $ ...

SQLFairy has three serialization producers, none of which is superior
to the other in their description of a schema.  

=over 4

=item * XML-SQLFairy

This is the aforementioned XML format.  It is essentially a direct
mapping of the Schema objects into XML.  This can also provide a very
convenient bridge to describing a schema to a non-Perl application.
Providing a producer argument to "sqlt" of just "XML" will default to
using "XML-SQLFairy."

=item * Storable

This producer stores the Schema object using Perl's Storable.pm module
available on CPAN.

=item * YAML

This producer serialized the Schema object with the very readable
structured data format of YAML (http://www.yaml.org/).  Earlier
examples show serializing to YAML.

=back

=head1 VISUALIZING SQL SCHEMAS

The visualization tools in SQLFairy can graphically represent the
tables, fields, datatypes and sizes, constraints, and foreign key
relationships in a very compact and intuitive format.  This can be
very beneficial in understanding and document large or small schemas.
Two producers in SQLFairy will create pseudo-E/R (entity-relationship)
diagrams:

=over 4

=item * Diagram

The first visualization tool in SQLFairy, this producer uses libgd to
draw a picture of the schema.  The tables are evenly distributed in
definition order running in columns (i.e., no graphing algorithms are
used), so the end result may result in many crossed lines showing the
foreign key relationships.  Please read the documentation of the
"sqlt-diagram" script for all the options available to this producer.

=item * GraphViz

The layout of the GraphViz producer is far superior to the Diagram
producer as it uses the Graphviz binary from Bell Labs to create very
professional-looking graphs.  There are several different layout
algorithms and node shapes available.  Please see the documentation of
the "sqlt-graph" script for more information.

=back

=head1 AUTOMATED CODE-GENERATION 

Given that so many applications interact with SQL databases, it's no
wonder that people have automated code to deal with this interaction.
Class::DBI from CPAN is one such module that allows a developer to
describe the relationships between tables and fields in class
declarations and then generates all the SQL to interact (SELECT,
UPDATE, DELETE, INSERT statements) at runtime.  Obviously, the schema
already describes itself, so it only makes sense that you should be
able to generate this kind of code directly from the schema.  The
"ClassDBI" producer in SQLFairy does just this, creating a Perl module
that inherits from Class::DBI and sets up most of the code needed to
interact with the database.  Here is an example of how to do this:

  $ sqlt -f MySQL -t ClassDBI foo.sql > Foo.pm

Then simply edit Foo.pm as needed and include it in your code.

=head1 DOCUMENTING WITH SQL::TRANSLATOR

SQLFairy offers two producers to help document schemas:

=over 4

=item * HTML

This producer creates a single HTML document which uses HTML
formatting to describe the Schema objects and to create hyperlinks on
foreign key relationships.  This can be a surprisingly useful
documentation aid as it creates a very readable format that allows one
to jump easily to specific tables and fields.  It's also possible to
plugin your own CSS to further control the presentation of the HTML.

=item * POD

This is arguably not that useful of a producer by itself, but the
number of POD-conversion tools could be used to further transform the
POD into something more interesting.  The schema is basically
represented in POD sections where tables are broken down into fields,
indices, constraints, foreign keys, etc.

=back

=head1 TEMPLATE-BASED MANIPULATION OF SCHEMA OBJECTS

All of the producers which create text output could have been coded
using a templating system to mix in the dynamic output with static
text.  CPAN offers several diverse templating systems, but few are as
powerful as Template Toolkit (http://www.template-toolkit.org/).  You
can easily create your own producer without writing any Perl code at
all simply by writing a template using Template Toolkit's syntax.  The
template will be passed a reference to the Schema object briefly
described at the beginning of this document and mentioned many times
throughout.  For example, you could create a template that simply
prints the name of each table and field that looks like this:

  # file: schema.tt
  [% FOREACH table IN schema.get_tables %]
  Table: [% table.name %]
  Fields:
  [% FOREACH field IN table.get_fields -%]
    [% field.name %]
  [% END -%]
  [% END %]

And the process it like so:

  $ sqlt -f YAML -t TTSchema --template schema.tt foo.yaml

To create output like this:

  Table: foo
  Fields:
    foo_id
    foo_name

For more information on Template Toolkit, please install the
"Template" module and read the POD.

=head1 PLUGIN YOUR OWN PARSERS AND PRODUCERS

Now that you have seen how the parsers and producers interact via the
Schema objects, you may wish to create your own versions to plugin.

=head1 AUTHOR

Ken Y. Clark E<lt>kclark@cpan.orgE<gt>.