Method and system for defining transactions from a database log

A method of defining transactions from a log of a database based upon a plurality of transaction templates, the log comprising data relating to changes to the database stored in a plurality of records each including a filename field representing a database file affected by a change, job identifier data, and data values from data fields in the database, each template being associated with a key value comprising one or more data fields in the database, each template comprising a plurality of filenames representing files changed during one or more given transactions, the method comprising:

• retrieving a log record from the database log;
• identifying one of the plurality of templates by comparing the data fields whose data values are stored in the retrieved record to the key values contained in the templates;
• searching for an index file containing job identifier data which matches the job identifier data stored in the retrieved record;
• if no index file is found having matching job identifier data, creating an index file and inserting information relating to the record into the index file including the job identifier data, key value of the identified template, and filename from the retrieved record;
• if an index file is found having matching job identifier data, comparing the key value contained in the found index file with the data fields whose data values are stored in the retrieved record;
• and
• if the key value contained in the found index file matches one or more of the data fields whose data values are stored in the retrieved record, inserting information relating to the record into the index file including the filename from the retrieved record.

A method of providing transaction protection for a database, wherein actions performed on data within the database are recorded in a log having a plurality of entries each storing at least job identifier data representing a source of a job being performed, a filename representing a file in the database affected by an action, an action performed on data in the database file, and one or more data fields from the database file which comprise a key value, and wherein transactions comprise a plurality of actions performed on a plurality of files, the method comprising:

• retrieving entries from the log;
• storing at least one template listing filenames which are affected by one or more transactions;
• for each retrieved log entry, comparing the filename in the entry to the filenames in the at least one template to determine whether the entry represents part of a transaction;
• for each entry which represents part of a transaction:
• determining based at least in part on the job identifier data whether a transaction is in progress for the job identified by the job identifier data;
• if no transaction is in progress for the job, marking the entry as part of a new transaction;
• if a transaction is in progress for the job, determining whether the transaction in progress is for the key value stored in the entry and if so, marking the entry as part of the transaction in progress if the filename in the entry has not already been marked in the transaction in progress;
• if the transaction in progress is for a key value different than the key value in the entry, or if the filename in the entry has already been marked in the transaction in progress, closing the transaction in progress and marking the entry as part of a new transaction;
• and
• in the event of loss or damage to data in the database, undoing any actions represented by entries marked as part of a transaction in progress which has not been closed.

A method of creating a transaction template from a log of a database, the log comprising a plurality of records of changes made to the database, the records containing data values associated with data fields in the database, the method comprising:

• retrieving the data values from the records and storing the data values;
• comparing the data values from a given record with data values from other records;
• when one or more data values from the given record match one or more data values from one or more other records, identifying the one or more data fields from the database files associated with the matching one or more data values;
• and
• creating a transaction template associated with a key value comprising the identified one or more data fields.

A method of providing transaction protection for a database comprising:

• creating a plurality of transaction templates each comprising a plurality of filenames identifying files in the database affected during a transaction and a key value representing one or more data fields in the database included in every action performed during the transaction;
• for each of a plurality of entries in a log of the database, determining whether the entry belongs to a transaction based on the data in the entry and the key values associated with the templates;
• maintaining index files representing transactions in progress by creating a new index file for each new transaction, adding data to a given index file from a log entry which belongs to the transaction associated with the given index file, and deleting index files when transactions with which they are associated are complete;
• and
• in the event the database is damaged, using existing index files to determine at least which transactions did not complete before the database was damaged.

For use with a database manager program in which a log of database operations is maintained having a plurality of records each representing an action performed on the database, a system for providing externally defined transaction protection to a database, the system comprising:

• means for retrieving records from the log;
• a plurality of transaction templates each containing information representing one or more types of transactions;
• first comparison means for determining based on the information contained in the templates whether each log record is part of a type of transaction;
• second comparison means for indicating which records are part of transactions in progress;
• means for, in the event of computer system failure or damage, undoing any actions performed on the database as represented by records which are part of transactions in progress.

A computer usable medium having computer readable program code embodied in said medium for causing the computer to perform method steps for creating a transaction template from a log of a database, the log comprising a plurality of records of changes made to the database, the records containing data values associated with data fields in the database, the method steps comprising:

• retrieving the data values from the records and storing the data values;
• comparing the data values from a given record with data values from other records;
• when one or more data values from the given record match one or more data values from one or more other records, identifying the one or more data fields from the database files associated with the matching one or more data values;
• and
• creating a transaction template associated with a key value comprising the identified one or more data fields.

A computer usable medium having computer readable program code embodied in said medium for causing the computer to perform method steps for defining transactions from a log of a database, the log comprising data relating to changes to the database, the log data being stored in a plurality of records each having a plurality of fields, the method steps comprising:

• for each of a plurality of records in the log, determining whether the record relates to a transaction included in a transaction template;
• for each record that relates to a transaction included in the template, determining whether the record relates to a new transaction or an incomplete transaction;
• for each record relating to a new transaction, creating and storing an indicator for the transaction and inserting information relating to the record into the indicator;
• and
• for each record relating to an incomplete transaction, determining whether the record represents the end of the incomplete transaction and, if the record does not represent the end of the incomplete transaction, inserting information relating to the record into the previously created indicator for the incomplete transaction.