CODEWITHSUNDEEP
oracle-databaseplsql
Viktor
Viktor

Reputation: 1338

Record type reflection / populating a record variable of an unknown type (PL/SQL)

I have a package with a record type and a variable of that type in it: 

CREATE PACKAGE my_package AS
  TYPE t_rec IS RECORD(
    id           NUMBER,
    name         VARCHAR2(100),
    last_updated DATE
  );
  v_var t_rec;
END;

I want to create an universal procedure for populating any record variable, so it can be called like this: 

BEGIN
  populate_record('my_package.v_var', 'MY_TABLE', CHARTOROWID(:VAR1));
END;

The logic of that procedure may look like this: 

PROCEDURE populate_record(
  p_var_name   IN VARCHAR2,
  p_table_name IN VARCHAR2,
  p_rowid      IN ROWID
) IS
BEGIN
  -- determining a type of a given record variable
  -- determining names and data types of given record variable's columns
  -- selecting a row from a given table by a given rowid
  -- filling out a given record variable by mapping table columns
  --   to record columns by their names (using dynamic PL/SQL)
END;

But I have no idea how to obtain information about a given record. Are there any data dictionary views or built-in functions for querying record type columns?
Thanks.

P.S.: 11.2

UPDATED

I found a way to obtain info about records using PL/Scope:

CREATE OR REPLACE VIEW user_record_types AS
SELECT
  i.object_name package_name,
  i.name type_name
FROM
  user_identifiers i
WHERE
  i.object_type = 'PACKAGE'
  AND i.type = 'RECORD'
  AND i.usage = 'DECLARATION'
/

CREATE OR REPLACE VIEW user_record_types_columns AS
SELECT
  i1.object_name package_name,
  i1.name type_name,
  ROW_NUMBER() OVER (
    PARTITION BY
      i1.object_name,
      i1.name
    ORDER BY
      i2.usage_id
  ) column_num,
  i2.name column_name,
  i3.name column_type,
  i3.type column_type_class
FROM
  user_identifiers i1,
  user_identifiers i2,
  user_identifiers i3
WHERE
  i1.object_type = 'PACKAGE'
  AND i1.type = 'RECORD'
  AND i1.usage = 'DECLARATION'
  AND i2.object_name = i1.object_name
  AND i2.object_type = i1.object_type
  AND i2.type = 'VARIABLE'
  AND i2.usage = 'DECLARATION'
  AND i2.usage_context_id = i1.usage_id
  AND i3.object_name = i2.object_name
  AND i3.object_type = i2.object_type
  AND i3.usage = 'REFERENCE'
  AND i3.usage_context_id = i2.usage_id
/

CREATE OR REPLACE VIEW user_record_variables AS
SELECT
  i1.object_name package_name,
  i1.name variable_name,
  CASE i2.type
    WHEN 'RECORD' THEN
      i2.object_name
    WHEN 'PACKAGE' THEN
      i2.name
  END type_package_name,
  CASE i2.type
    WHEN 'RECORD' THEN
      i2.name
    WHEN 'PACKAGE' THEN
      i3.name
  END type_name
FROM
  user_identifiers i1,
  user_identifiers i2,
  user_identifiers i3
WHERE
  i1.object_type = 'PACKAGE'
  AND i1.type = 'VARIABLE'
  AND i1.usage = 'DECLARATION'
  AND i2.object_name = i1.object_name
  AND i2.object_type = i1.object_type
  AND i2.type IN ('RECORD', 'PACKAGE')
  AND i2.usage = 'REFERENCE'
  AND i2.usage_context_id = i1.usage_id
  AND i3.object_name (+) = i2.object_name
  AND i3.object_type (+) = i2.object_type
  AND i3.type (+) = 'RECORD'
  AND i3.usage (+) = 'REFERENCE'
  AND i3.usage_context_id (+) = i2.usage_id
/

But there is a problem when I use %ROWTYPEs, because there is no info about "what of" that %ROWTYPE is. So I think that PL/Scope is not a complete solution...

Upvotes: 2

Views: 1481

Answers (1)

Emmanuel
Emmanuel

Reputation: 14209

You can try to use table user_identifiers or all_identifiers if you have the access, as this other SO post shows.

You may need to recompile your packages:

alter package  my_types compile plscope_settings='IDENTIFIERS:ALL' reuse settings;

Unfortunately this is only available from 11gR1.

Upvotes: 1

Related Questions