Sequel PostgreSQL Triggers¶ ↑

Sequel PostgreSQL Triggers is a small enhancement to Sequel allowing a user to easily handle the following types of columns:

• Timestamp Columns (Created At/Updated At)

• Counter/Sum Caches

• Immutable Columns

• Touch Propogation

• Foreign Key Arrays (Referential Integrity Checks)

It handles these internally to the database via triggers, so even if other applications access the database (without using Sequel), things will still work (unless the database superuser disables triggers).

To use this, load the pg_triggers extension into the Sequel::Database object:

DB.extension :pg_triggers

Then you can call the pgt_* methods it adds on your Sequel::Database object:

DB.pgt_created_at(:table_name, :created_at)

Most commonly, this is used in migrations, with a structure similar to:

Sequel.migration do
  up do
    extension :pg_triggers

    pgt_created_at(:table_name,
                   :created_at,
                   :function_name=>:table_name_set_created_at,
                   :trigger_name=>:set_created_at)
  end

  down do
    drop_trigger(:table_name, :set_created_at)
    drop_function(:table_name_set_created_at)
  end
end

Note that you only need to load this extension when defining the triggers, you don’t need to load this extension when your application is running.

To use any of these methods before PostgreSQL 9.0, you have to add the plpgsql procedural language to PostgreSQL, which you can do with:

DB.create_language(:plpgsql)

If you want to load this extension globally for all PostgreSQL databases, you can do:

require 'sequel_postgresql_triggers'

However, global modification is discouraged and only remains for backwards compatibility.

All of the public methods this extension adds take the following options in their opts hash:

:function_name

The name of the function to use. This is important to specify if you want an easy way to drop the function.

:trigger_name

The name of the trigger to use. This is important to specify if you want an easy way to drop the trigger.

Methods that create trigger functions which can result in additional triggers firing (due to INSERT/UPDATE/DELETE queries) also take the following option:

:trigger_depth_limit

If given, should be an integer specifying the maximum trigger depth that the trigger will operate under. Any higher depth will result in the trigger silently exiting. This can be used to prevent unbounded trigger recursion. A value of true is interpreted as 1.

pgt_created_at takes the table and column given and makes it so that upon insertion, the column is set to the CURRENT_TIMESTAMP, and that upon update, the column’s value is always set to the previous value. This is sort of like an immutable column, but it doesn’t bring up an error if you try to change it, it just ignores it.

Arguments:

table

name of table

column

column in table that should be a created at timestamp column

opts

option hash

Similar to pgt_created_at, takes a table and column and makes it so that upon insertion, the column is set to CURRENT_TIMESTAMP. It differs that upon update, the column is also set to CURRENT_TIMESTAMP.

Arguments:

table

name of table

column

column in table that should be a updated at timestamp column

opts

options hash

This takes many arguments and sets up a counter cache so that when the counted table is inserted to or deleted from, records in the main table are updated with the count of the corresponding records in the counted table. The counter cache column must have a default of 0 for this to work correctly.

Use pgt_sum_cache with a Sequel expression in summed_column to handle any custom logic such as a counter cache that only counts certain rows.

Arguments:

main_table

name of table holding counter cache column

main_table_id_column

column in main table matching counted_table_id_column in counted_table

counter_column

column in main table containing the counter cache

counted_table

name of table being counted

counted_table_id_column

column in counted_table matching main_table_id_column in main_table

opts

options hash

Similar to pgt_counter_cache, except instead of storing a count of records in the main table, it stores the sum on one of the columns in summed table. The sum cache column must have a default of 0 for this to work correctly.

Use a Sequel expression in summed_column to handle any custom logic such as a counter cache that only counts certain rows, or a sum cache that sums the length of a string column.

Arguments:

main_table

name of table holding counter cache column

main_table_id_column

column in main table matching summed_table_id_column in summed_table

sum_column

column in main table containing the sum cache

summed_table

name of table being summed

summed_table_id_column

column in summed_table matching main_table_id_column in main_table

summed_column

column in summed_table being summed or a Sequel expression to be evaluated in the context of summed_table

opts

options hash

Similar to pgt_sum_cache, except instead of a one-to-many relationship, it supports a many-to-many relationship with a single join table. The sum cache column must have a default of 0 for this to work correctly. Use a Sequel expression in summed_column to handle any custom logic. See pgt_sum_cache for details.

This takes a single options hash argument, supporting the following options in addition to the standard options:

:main_table

name of table holding sum cache column

:main_table_id_column

primary key column in main table referenced by main_table_fk_column (default: :id)

:sum_column

column in main table containing the sum cache, must be NOT NULL and default to 0

:summed_table

name of table being summed

:summed_table_id_column

primary key column in summed_table referenced by summed_table_fk_column (default: :id)

:summed_column

column in summed_table being summed or a Sequel expression to be evaluated in the context of summed_table, must be NOT NULL

:join_table

name of table which joins main_table with summed_table

:join_trigger_name

name of trigger for join table

:join_function_name

name of trigger function for join table

:main_table_fk_column

column in join_table referencing main_table_id_column, must be NOT NULL

:summed_table_fk_column

column in join_table referencing summed_table_id_column, must be NOT NULL

This takes a table name and one or more column names, and adds an update trigger that raises an exception if you try to modify the value of any of the columns.

Arguments:

table

name of table

*columns

All columns in the table that should be immutable. Can end with options hash.

This takes several arguments and sets up a trigger that watches one table for changes, and touches timestamps of related rows in a separate table.

Arguments:

main_table

name of table that is being watched for changes

touch_table

name of table that needs to be touched

column

name of timestamp column to be touched

expr

hash or array that represents the columns that define the relationship

opts

options hash

This takes a single options hash, and sets up triggers on both tables involved. The table with the foreign key array has insert/update triggers to make sure newly inserted/updated rows reference valid rows in the referenced table. The table being referenced has update/delete triggers to make sure the value before update or delete is not still being referenced.

Note that this will not catch all referential integrity violations, but it should catch the most common ones.

Options:

:table

table with foreign key array

:column

foreign key array column

:referenced_table

table referenced by foreign key array

:referenced_column

column referenced by foreign key array (generally primary key)

:referenced_function_name

function name for trigger function on referenced table

:referenced_trigger_name

trigger name for referenced table

This takes 2 arguments, a table and a hash of column default values, and sets up an insert trigger that will override user submitted or database default values and use the values given when setting up the trigger. This is mostly useful in situations where multiple database accounts are used where one account has insert permissions but not update permissions, and you want to ensure that inserted rows have specific column values to enforce security requirements.

Arguments:

table

The name of the table

defaults

A hash of default values to enforce, where keys are column names and values are the default values to enforce

These methods setup an auditing function where updates and deletes log the previous values to a central auditing table in JSON format.

This creates an audit table and a trigger function that will log previous values to the audit table. This returns the name of the trigger function created, which should be passed to pgt_json_audit_log.

Arguments:

table

The name of the table storing the audit logs.

Options:

function_opts

Options to pass to create_function when creating the trigger function.

The audit log table will store the following columns:

txid

The 64-bit transaction ID for the transaction that made the modification (txid_current())

at

The timestamp of the transaction that made the modification (CURRENT_TIMESTAMP)

user

The database user name that made the modification (CURRENT_USER)

schema

The schema containing the table that was modified (TG_TABLE_SCHEMA)

table

The table that was modified (TG_TABLE_NAME)

action

The type of modification, either DELETE or UPDATE (TG_OP)

prior

A jsonb column with the contents of the row before the modification (to_jsonb(OLD))

This adds a trigger to the table that will log previous values to the audting table for updates and deletes.

Arguments:

table

The name of the table to audit

function

The name of the trigger function to call to log changes

Note that it is probably a bad idea to use the same table argument to both pgt_json_audit_log_setup and pgt_json_audit_log.

If you have defined counter or sum cache triggers using this library before version 1.6.0, you should drop them and regenerate them if you want the triggers to work correctly with queries that use INSERT ... ON CONFLICT DO NOTHING.

When restoring a data-only migration with pg_dump, you may need to use --disable-triggers for it to restore correctly, and you will need to manually enforce data integrity if you are doing partial restores and not full restores.

This library is released under the MIT License. See the MIT-LICENSE file for details.

Jeremy Evans <code@jeremyevans.net>