Enrico,the DB trigger approach sounds very exciting to me because I really need to be sure that there is no way to manipulate records without audit trail. I also would be very interested in the trigger code for MySQL you mentioned.OK. I'll append some sample code below. You also mentioned that you did something similar in the past, what was your approach to store information in the audit trail table? At the moment, I have two ways in front of me and I am not quite sure which way to go. Assuming my primary table (rooms) stores the fields id, room, number, size and version.My approach was one audit record per changed primary record, not one audit record per changed primary field. More like your 1st example above. But there's no need to store old and new values in the audit table. Just store new values in the audit table. You can find the old value in a previous row of the audit table (sorted by creation time of the audit table records), from when it was a new value being inserted or updated. Make sure you store all new values in the audit table, including the first INSERT. Then you can see a complete audit trail in the audit tables, even if the value has never changed. In a previous project, I made the mistake of only putting old values into the audit tables, after they changed in the primary tables. So, I had to look at the audit tables for all old values and at the primary tables for the current value, which was more complicated. Be sure to use triggers for DELETE, not just INSERT and UPDATE. Otherwise, the audit table won't have a record of whether the row was deleted. Here are a set of tables (primary and audit) and triggers (INSERT, UPDATE, and DELETE) for a typical table in my DB. This is valid MySQL code. In this code, you'll see some other conventions I always follow: - Naming conventions of: -- Tables (prefixed with app name ("ITF" in this case)) -- PKs (table name, plus suffix "_id") -- Constraints (prefixes like "pk_", fk1_", "fk2_", etc for PK and potentially multiple FKs) -- Audit tables (prefixed with app name plus "A") -- Triggers ("ai_", "au_", "ad_" prefixes for audit triggers for INSERT, UPDATE, and DELETE) - All primary tables contain these extra fields which dramatically reduce the number of times you have to go to the audit tables: -- create_user (string name of user who created row) -- create_dt (date/time of row creation) -- update_user (string name of user who last updated row) -- update_dt (date/time of last row update) - All primary tables contain this extra field, which I use to track the status of the row (active, inactive, archived, etc.). I never actually delete a row. Just mark it inactive, which makes undelete possible w/o having to go to the audit table. -- status_id - Each audit table has exactly the same fields as its primary table, plus 3 additional fields: -- audited_change_user (string name of user who made the change) -- audited_change_dt (date/time time of the change.' -- audited_change_type ('INSERT', 'UPDATE', or 'DELETE') - Audit tables have no need for unique, FK, or PK constraints. Just values. Here's the code: -- -- ITF_PRODUCT -- DROP TABLE IF EXISTS itf_product; CREATE TABLE itf_product (itf_product_id INTEGER NOT NULL AUTO_INCREMENT COMMENT 'Primary key' ,name VARCHAR(255) NOT NULL COMMENT 'Unique user-assigned and user-visible name' ,descrip VARCHAR(255) NOT NULL COMMENT 'Description' ,notes VARCHAR(255) NOT NULL COMMENT 'User notes' ,create_user VARCHAR(255) NOT NULL COMMENT 'User who created this database row. String, not foreign key, to preserve history when users are deleted.' ,create_dt DATETIME NOT NULL COMMENT 'Date and time of creation of this database row.' ,update_user VARCHAR(255) NOT NULL COMMENT 'User who last updated this database row. String, not foreign key, to preserve history when users are deleted.' ,update_dt DATETIME NOT NULL COMMENT 'Date and time of last update of this database row.' ,status_id INTEGER NOT NULL COMMENT 'The status of the data in this database row (active, inactive, archived, etc.). FK to itf_dict value of category STATUS_LOOKUP.' ,CONSTRAINT pk_product PRIMARY KEY (itf_product_id) ) COMMENT='Products' ENGINE=InnoDB ; ALTER TABLE itf_product ADD CONSTRAINT fk1_product FOREIGN KEY fk1_product (status_id) REFERENCES itf_dict (itf_dict_id) ; -- Disallow multiple products with same name. -- No, we really want to disallow only for records with STATUS=ACTIVE, -- which can't be done here with a UNIQUE constraint, so this is enforced -- in the Java BO code instead of here. -- CREATE UNIQUE INDEX ak1_product ON itf_product (name); GRANT SELECT, INSERT, UPDATE, DELETE ON itf_product to itfweb; DROP TABLE IF EXISTS itfa_product; CREATE TABLE itfa_product (audited_change_user VARCHAR(255) NOT NULL COMMENT 'User who made this change. String, not foreign key, to preserve history when users are deleted.' ,audited_change_dt DATETIME NOT NULL COMMENT 'Date and time of this change.' ,audited_change_type VARCHAR(255) NOT NULL COMMENT 'INSERT, UPDATE, or DELETE. No need for lookup in ITF_DICT since there are only these 3 types of triggers.' ,itf_product_id INTEGER NULL COMMENT 'Copied from audited table.' ,name VARCHAR(255) NULL COMMENT 'Copied from audited table.' ,descrip VARCHAR(255) NULL COMMENT 'Copied from audited table.' ,notes VARCHAR(255) NULL COMMENT 'Copied from audited table.' ,create_user VARCHAR(255) NULL COMMENT 'Copied from audited table.' ,create_dt DATETIME NULL COMMENT 'Copied from audited table.' ,update_user VARCHAR(255) NULL COMMENT 'Copied from audited table.' ,update_dt DATETIME NULL COMMENT 'Copied from audited table.' ,status_id INTEGER NULL COMMENT 'Copied from audited table.' ) COMMENT='Audit table. No need for constraints.' ENGINE=InnoDB ; GRANT SELECT, INSERT, UPDATE, DELETE ON itfa_product to itfweb; -- DROP TRIGGER itft_ai_product; DELIMITER ;; CREATE TRIGGER itft_ai_product BEFORE INSERT ON itf_product FOR EACH ROW BEGIN -- Purpose: Inserts an audit record into audit table. -- -- MODIFICATION HISTORY -- Person Date Comments -- --------- ---------- ------------------------------------------- -- Fred Stluka 2/12/2007 Original version. -- INSERT INTO itfa_product (audited_change_user ,audited_change_dt ,audited_change_type ,itf_product_id ,name ,descrip ,notes ,create_user ,create_dt ,update_user ,update_dt ,status_id ) VALUES (NEW.update_user ,SYSDATE() ,'INSERT' ,1 + (SELECT IFNULL(MAX(itf_product_id),0) from itf_product) -- Can't just use NEW.itf_product_id. AUTO_INCREMENT -- hasn't yet generated a non-zero value. This works OK -- as long as the highest generated value hasn't been -- deleted from the table. -- -- Can't use: 1 + (SELECT MAX(itf_product_id) from itf_product) -- because it causes the first INSERT to fail. -- MAX comes up NULL and gets added to 1 which produces NULL. -- IFNULL fixes that. -- -- Could perhaps use an AFTER trigger instead of a BEFORE -- trigger, and NEW.itf_product_id would work?? ,NEW.name ,NEW.descrip ,NEW.notes ,NEW.create_user ,NEW.create_dt ,NEW.update_user ,NEW.update_dt ,NEW.status_id ); END; ;; DELIMITER ; -- DROP TRIGGER itft_au_product; DELIMITER ;; CREATE TRIGGER itft_au_product BEFORE UPDATE ON itf_product FOR EACH ROW BEGIN -- Purpose: Inserts an audit record into audit table. -- -- MODIFICATION HISTORY -- Person Date Comments -- --------- ---------- ------------------------------------------- -- Fred Stluka 2/12/2007 Original version. -- INSERT INTO itfa_product (audited_change_user ,audited_change_dt ,audited_change_type ,itf_product_id ,name ,descrip ,notes ,create_user ,create_dt ,update_user ,update_dt ,status_id ) VALUES (NEW.update_user ,SYSDATE() ,'UPDATE' ,NEW.itf_product_id ,NEW.name ,NEW.descrip ,NEW.notes ,NEW.create_user ,NEW.create_dt ,NEW.update_user ,NEW.update_dt ,NEW.status_id ); END; ;; DELIMITER ; -- DROP TRIGGER itft_ad_product; DELIMITER ;; CREATE TRIGGER itft_ad_product BEFORE DELETE ON itf_product FOR EACH ROW BEGIN -- Purpose: Inserts an audit record into audit table. -- -- MODIFICATION HISTORY -- Person Date Comments -- --------- ---------- ------------------------------------------- -- Fred Stluka 2/12/2007 Original version. -- INSERT INTO itfa_product (audited_change_user ,audited_change_dt ,audited_change_type ,itf_product_id ,name ,descrip ,notes ,create_user ,create_dt ,update_user ,update_dt ,status_id ) VALUES (USER() -- Any better solution for DELETE?? -- Could have convention of doing an -- UPDATE immediately before the -- DELETE and use OLD.update_user ,SYSDATE() ,'DELETE' ,OLD.itf_product_id ,OLD.name ,OLD.descrip ,OLD.notes ,OLD.create_user ,OLD.create_dt ,OLD.update_user ,OLD.update_dt ,OLD.status_id ); END; ;; DELIMITER ; Hope this helps! --Fred
Fred Stluka -- mailto:f...@bristle.com -- http://bristle.com/~fred/ Bristle Software, Inc -- http://bristle.com -- Glad to be of service! Open Source: Without walls and fences, we need no Windows or Gates. On 1/22/17 5:52 AM, enrico baranski
wrote:
-- You received this message because you are subscribed to the Google Groups "Django users" group. To unsubscribe from this group and stop receiving emails from it, send an email to django-users+unsubscr...@googlegroups.com. To post to this group, send email to django-users@googlegroups.com. Visit this group at https://groups.google.com/group/django-users. To view this discussion on the web visit https://groups.google.com/d/msgid/django-users/58982210-c552-34ad-76b5-11ff6add0187%40bristle.com. For more options, visit https://groups.google.com/d/optout. |
- audit trail functionality in database model enrico baranski
- Re: audit trail functionality in database model Mike Dewhirst
- Re: audit trail functionality in database model enrico baranski
- Re: audit trail functionality in database model Fred Stluka
- Re: audit trail functionality in database model enrico baranski
- Re: audit trail functionality in database mod... enrico baranski
- Re: audit trail functionality in database... Fred Stluka
- Re: audit trail functionality in database mod... Fred Stluka
- Re: audit trail functionality in database model Melvyn Sopacua
- Re: audit trail functionality in database model Ryan Castner
- Re: audit trail functionality in database model enrico baranski
