public static function extract_schema($host, $port, $database, $user, $password)
{
$databases = explode(',', $database);
dbsteward::notice("Connecting to mysql5 host " . $host . ':' . $port . ' database ' . $database . ' as ' . $user);
// if not supplied, ask for the password
if ($password === FALSE) {
echo "Password: "******"Analyzing database {$database}");
$db->use_database($database);
$node_schema = $doc->addChild('schema');
$node_schema['name'] = $database;
$node_schema['owner'] = 'ROLE_OWNER';
// extract global and schema permissions under the public schema
foreach ($db->get_global_grants($user) as $db_grant) {
$node_grant = $node_schema->addChild('grant');
// There are 28 permissions encompassed by the GRANT ALL statement
$node_grant['operation'] = $db_grant->num_ops == 28 ? 'ALL' : $db_grant->operations;
$node_grant['role'] = self::translate_role_name($user, $doc);
if ($db_grant->is_grantable) {
$node_grant['with'] = 'GRANT';
}
}
$enum_types = array();
$enum_type = function ($obj, $mem, $values) use(&$enum_types) {
// if that set of values is defined by a previous enum, use that
foreach ($enum_types as $name => $enum) {
if ($enum === $values) {
return $name;
}
}
// otherwise, make a new one
$name = "enum_" . md5(implode('_', $values));
$enum_types[$name] = $values;
return $name;
};
foreach ($db->get_tables() as $db_table) {
dbsteward::info("Analyze table options/partitions " . $db_table->table_name);
$node_table = $node_schema->addChild('table');
$node_table['name'] = $db_table->table_name;
$node_table['owner'] = 'ROLE_OWNER';
// because mysql doesn't have object owners
$node_table['description'] = $db_table->table_comment;
$node_table['primaryKey'] = '';
if (stripos($db_table->create_options, 'partitioned') !== FALSE && ($partition_info = $db->get_partition_info($db_table))) {
$node_partition = $node_table->addChild('tablePartition');
$node_partition['sqlFormat'] = 'mysql5';
$node_partition['type'] = $partition_info->type;
switch ($partition_info->type) {
case 'HASH':
case 'LINEAR HASH':
$opt = $node_partition->addChild('tablePartitionOption');
$opt->addAttribute('name', 'expression');
$opt->addAttribute('value', $partition_info->expression);
$opt = $node_partition->addChild('tablePartitionOption');
$opt->addAttribute('name', 'number');
$opt->addAttribute('value', $partition_info->number);
break;
case 'KEY':
case 'LINEAR KEY':
$opt = $node_partition->addChild('tablePartitionOption');
$opt->addAttribute('name', 'columns');
$opt->addAttribute('value', $partition_info->columns);
$opt = $node_partition->addChild('tablePartitionOption');
$opt->addAttribute('name', 'number');
$opt->addAttribute('value', $partition_info->number);
break;
case 'LIST':
case 'RANGE':
case 'RANGE COLUMNS':
$opt = $node_partition->addChild('tablePartitionOption');
$opt->addAttribute('name', $partition_info->type == 'RANGE COLUMNS' ? 'columns' : 'expression');
$opt->addAttribute('value', $partition_info->expression);
foreach ($partition_info->segments as $segment) {
$node_seg = $node_partition->addChild('tablePartitionSegment');
$node_seg->addAttribute('name', $segment->name);
$node_seg->addAttribute('value', $segment->value);
}
break;
}
}
foreach ($db->get_table_options($db_table) as $name => $value) {
if (strcasecmp($name, 'auto_increment') === 0 && !static::$use_auto_increment_table_options) {
// don't extract auto_increment tableOptions if we're not using them
continue;
}
$node_option = $node_table->addChild('tableOption');
$node_option['sqlFormat'] = 'mysql5';
$node_option['name'] = $name;
$node_option['value'] = $value;
}
dbsteward::info("Analyze table columns " . $db_table->table_name);
foreach ($db->get_columns($db_table) as $db_column) {
$node_column = $node_table->addChild('column');
$node_column['name'] = $db_column->column_name;
if (!empty($db_column->column_comment)) {
$node_column['description'] = $db_column->column_comment;
}
// returns FALSE if not serial, int/bigint if it is
$type = $db->is_serial_column($db_table, $db_column);
if (!$type) {
$type = $db_column->column_type;
if (stripos($type, 'enum') === 0) {
$values = $db->parse_enum_values($db_column->column_type);
$type = $enum_type($db_table->table_name, $db_column->column_name, $values);
}
if ($db_column->is_auto_increment) {
$type .= ' AUTO_INCREMENT';
}
}
if ($db_column->is_auto_update) {
$type .= ' ON UPDATE CURRENT_TIMESTAMP';
}
$node_column['type'] = $type;
// @TODO: if there are serial sequences/triggers for the column then convert to serial
if ($db_column->column_default !== NULL) {
$node_column['default'] = mysql5::escape_default_value($db_column->column_default);
} elseif (strcasecmp($db_column->is_nullable, 'YES') === 0) {
$node_column['default'] = 'NULL';
}
$node_column['null'] = strcasecmp($db_column->is_nullable, 'YES') === 0 ? 'true' : 'false';
}
// get all plain and unique indexes
dbsteward::info("Analyze table indexes " . $db_table->table_name);
foreach ($db->get_indices($db_table) as $db_index) {
// don't process primary key indexes here
if (strcasecmp($db_index->index_name, 'PRIMARY') === 0) {
continue;
}
// implement unique indexes on a single column as unique column, but only if the index name is the column name
if ($db_index->unique && count($db_index->columns) == 1 && strcasecmp($db_index->columns[0], $db_index->index_name) === 0) {
$column = $db_index->columns[0];
$node_column = dbx::get_table_column($node_table, $column);
if (!$node_column) {
throw new Exception("Unexpected: Could not find column node {$column} for unique index {$db_index->index_name}");
} else {
$node_column = $node_column[0];
}
$node_column['unique'] = 'true';
} else {
$node_index = $node_table->addChild('index');
$node_index['name'] = $db_index->index_name;
$node_index['using'] = strtolower($db_index->index_type);
$node_index['unique'] = $db_index->unique ? 'true' : 'false';
$i = 1;
foreach ($db_index->columns as $column_name) {
$node_index->addChild('indexDimension', $column_name)->addAttribute('name', $column_name . '_' . $i++);
}
}
}
// get all primary/foreign keys
dbsteward::info("Analyze table constraints " . $db_table->table_name);
foreach ($db->get_constraints($db_table) as $db_constraint) {
if (strcasecmp($db_constraint->constraint_type, 'primary key') === 0) {
$node_table['primaryKey'] = implode(',', $db_constraint->columns);
} elseif (strcasecmp($db_constraint->constraint_type, 'foreign key') === 0) {
// mysql sees foreign keys as indexes pointing at indexes.
// it's therefore possible for a compound index to point at a compound index
if (!$db_constraint->referenced_columns || !$db_constraint->referenced_table_name) {
throw new Exception("Unexpected: Foreign key constraint {$db_constraint->constraint_name} does not refer to any foreign columns");
}
if (count($db_constraint->referenced_columns) == 1 && count($db_constraint->columns) == 1) {
// not a compound index, define the FK inline in the column
$column = $db_constraint->columns[0];
$ref_column = $db_constraint->referenced_columns[0];
$node_column = dbx::get_table_column($node_table, $column);
if (!$node_column) {
throw new Exception("Unexpected: Could not find column node {$column} for foreign key constraint {$db_constraint->constraint_name}");
}
$node_column['foreignSchema'] = $db_constraint->referenced_table_schema;
$node_column['foreignTable'] = $db_constraint->referenced_table_name;
$node_column['foreignColumn'] = $ref_column;
unset($node_column['type']);
// inferred from referenced column
$node_column['foreignKeyName'] = $db_constraint->constraint_name;
// RESTRICT is the default, leave it implicit if possible
if (strcasecmp($db_constraint->delete_rule, 'restrict') !== 0) {
$node_column['foreignOnDelete'] = str_replace(' ', '_', $db_constraint->delete_rule);
}
if (strcasecmp($db_constraint->update_rule, 'restrict') !== 0) {
$node_column['foreignOnUpdate'] = str_replace(' ', '_', $db_constraint->update_rule);
}
} elseif (count($db_constraint->referenced_columns) > 1 && count($db_constraint->referenced_columns) == count($db_constraint->columns)) {
$node_fkey = $node_table->addChild('foreignKey');
$node_fkey['columns'] = implode(', ', $db_constraint->columns);
$node_fkey['foreignSchema'] = $db_constraint->referenced_table_schema;
$node_fkey['foreignTable'] = $db_constraint->referenced_table_name;
$node_fkey['foreignColumns'] = implode(', ', $db_constraint->referenced_columns);
$node_fkey['constraintName'] = $db_constraint->constraint_name;
// RESTRICT is the default, leave it implicit if possible
if (strcasecmp($db_constraint->delete_rule, 'restrict') !== 0) {
$node_fkey['onDelete'] = str_replace(' ', '_', $db_constraint->delete_rule);
}
if (strcasecmp($db_constraint->update_rule, 'restrict') !== 0) {
$node_fkey['onUpdate'] = str_replace(' ', '_', $db_constraint->update_rule);
}
} else {
var_dump($db_constraint);
throw new Exception("Unexpected: Foreign key constraint {$db_constraint->constraint_name} has mismatched columns");
}
} elseif (strcasecmp($db_constraint->constraint_type, 'unique') === 0) {
dbsteward::warning("Ignoring UNIQUE constraint '{$db_constraint->constraint_name}' because they are implemented as indices");
} elseif (strcasecmp($db_constraint->constraint_type, 'check') === 0) {
// @TODO: implement CHECK constraints
} else {
throw new exception("unknown constraint_type {$db_constraint->constraint_type}");
}
}
foreach ($db->get_table_grants($db_table, $user) as $db_grant) {
dbsteward::info("Analyze table permissions " . $db_table->table_name);
$node_grant = $node_table->addChild('grant');
$node_grant['operation'] = $db_grant->operations;
$node_grant['role'] = self::translate_role_name($user, $doc);
if ($db_grant->is_grantable) {
$node_grant['with'] = 'GRANT';
}
}
}
foreach ($db->get_sequences() as $db_seq) {
$node_seq = $node_schema->addChild('sequence');
$node_seq['name'] = $db_seq->name;
$node_seq['owner'] = 'ROLE_OWNER';
$node_seq['start'] = $db_seq->start_value;
$node_seq['min'] = $db_seq->min_value;
$node_seq['max'] = $db_seq->max_value;
$node_seq['inc'] = $db_seq->increment;
$node_seq['cycle'] = $db_seq->cycle ? 'true' : 'false';
// the sequences table is a special case, since it's not picked up in the tables loop
$seq_table = $db->get_table(mysql5_sequence::TABLE_NAME);
foreach ($db->get_table_grants($seq_table, $user) as $db_grant) {
$node_grant = $node_seq->addChild('grant');
$node_grant['operation'] = $db_grant->operations;
$node_grant['role'] = self::translate_role_name($doc, $user);
if ($db_grant->is_grantable) {
$node_grant['with'] = 'GRANT';
}
}
}
foreach ($db->get_functions() as $db_function) {
dbsteward::info("Analyze function " . $db_function->routine_name);
$node_fn = $node_schema->addChild('function');
$node_fn['name'] = $db_function->routine_name;
$node_fn['owner'] = 'ROLE_OWNER';
$node_fn['returns'] = $type = $db_function->dtd_identifier;
if (strcasecmp($type, 'enum') === 0) {
$node_fn['returns'] = $enum_type($db_function->routine_name, 'returns', $db->parse_enum_values($db_function->dtd_identifier));
}
$node_fn['description'] = $db_function->routine_comment;
if (isset($db_function->procedure) && $db_function->procedure) {
$node_fn['procedure'] = 'true';
}
// $node_fn['procedure'] = 'false';
$eval_type = $db_function->sql_data_access;
// srsly mysql? is_deterministic varchar(3) not null default '', contains YES or NO
$determinism = strcasecmp($db_function->is_deterministic, 'YES') === 0 ? 'DETERMINISTIC' : 'NOT DETERMINISTIC';
$node_fn['cachePolicy'] = mysql5_function::get_cache_policy_from_characteristics($determinism, $eval_type);
$node_fn['mysqlEvalType'] = str_replace(' ', '_', $eval_type);
// INVOKER is the default, leave it implicit when possible
if (strcasecmp($db_function->security_type, 'definer') === 0) {
$node_fn['securityDefiner'] = 'true';
}
foreach ($db_function->parameters as $param) {
$node_param = $node_fn->addChild('functionParameter');
// not supported in mysql functions, even though it's provided?
// $node_param['direction'] = strtoupper($param->parameter_mode);
$node_param['name'] = $param->parameter_name;
$node_param['type'] = $type = $param->dtd_identifier;
if (strcasecmp($type, 'enum') === 0) {
$node_param['type'] = $enum_type($db_function->routine_name, $param->parameter_name, $db->parse_enum_values($param->dtd_identifier));
}
if (isset($param->direction)) {
$node_param['direction'] = $param->direction;
}
}
$node_def = $node_fn->addChild('functionDefinition', $db_function->routine_definition);
$node_def['language'] = 'sql';
$node_def['sqlFormat'] = 'mysql5';
}
foreach ($db->get_triggers() as $db_trigger) {
dbsteward::info("Analyze trigger " . $db_trigger->name);
$node_trigger = $node_schema->addChild('trigger');
foreach ((array) $db_trigger as $k => $v) {
$node_trigger->addAttribute($k, $v);
}
$node_trigger->addAttribute('sqlFormat', 'mysql5');
}
foreach ($db->get_views() as $db_view) {
dbsteward::info("Analyze view " . $db_view->view_name);
if (!empty($db_view->view_name) && empty($db_view->view_query)) {
throw new Exception("Found a view in the database with an empty query. User '{$user}' problaby doesn't have SELECT permissions on tables referenced by the view.");
}
$node_view = $node_schema->addChild('view');
$node_view['name'] = $db_view->view_name;
$node_view['owner'] = 'ROLE_OWNER';
$node_view->addChild('viewQuery', $db_view->view_query)->addAttribute('sqlFormat', 'mysql5');
}
foreach ($enum_types as $name => $values) {
$node_type = $node_schema->addChild('type');
$node_type['type'] = 'enum';
$node_type['name'] = $name;
foreach ($values as $v) {
$node_type->addChild('enum')->addAttribute('name', $v);
}
}
}
xml_parser::validate_xml($doc->asXML());
return xml_parser::format_xml($doc->saveXML());
}