For several years I have been teaching a file updating algorithm which is essentially the same as that in Dwyer's admirable paper [1]. There is one unjustified objection to the algorithm that perceptive students and people with batch processing experience almost invariably raise and which is not addressed by Dwyer. The objection is based on a situation which arises in a batch processing environment with several users, where key-ordered sequential update is used.

Updating a master file—yet one more time

The Automatic Design of Batch Processing Systems

Large numbers of data processing applications require processing of master records and their associated detail records. These relationships between master and detail files may be classified as one-to-many, zero-or-one-to-many and many-to-many. All are important to information systems, but database management system implementations differ widely in their ability to represent and process such relationships. Processing of zero-or-one-to-many relationships, in particular, is difficult in most relational databases. An obscure relational algebra operator, left outer join, that provides such a capability is described and its importance to master-detail processing is illustrated. An implementation of left outer join is presented for dBase III, a database management system providing some relational-like capability.

A processing algorithm for master-detail records in a relational database

Expositions of sequential master file updating are usually concerned only with insertions of new records, alterations or deletions of existing records and with the joint processing of these transactions. In actual practice other types of file transactions may also be needed, such as printing of (parts of) records, preventing possible alterations of certain records, etc. A general framework for dealing as well with such transactions in one processing run is presented. Simplifications of this framework are possible for random-access files.

File updating revisited

In concluding we are sure to point ou t to students that the juke box is th e technological part of any on — lin e information system (e .g. stock control , airline seat reservation) or proces s control system (e .g. power station) i n miniature. The algorithm presented by D. Richards to update a sequentia l master file is not correct. Th e " deletes " have not the correc t effect because after the virtua l output of the undefined record, th e outbuf .key is equal to , s o master .key will be greater tha n outbuf .key, then the " suppressed " master will be moved again t o outbuf. .. For example, if a delete is the last transaction fo r a key, the corresponding maste r record will appear in the outpu t file. I agree with D. Richards in tha t the File Update Problem is a "simple situation " , but no t necessarily having a obviou s solution. Reference s Dwyer, B. " One More Time — How t o Update a Master File" .

Updating a master file—yet one more time

Citations

The Automatic Design of Batch Processing Systems

A processing algorithm for master-detail records in a relational database

File updating revisited

Note about: "On the file update problem"

A curriculum for a University course in advanced COBOL

References

One more time — how to update a master file

Related Papers (5)

One more time — how to update a master file

Principles of program design

Lexical file names in plan 9 or getting dot-dot right

A brief parallel I/O tutorial.

Unacceptable file operations in a relational data base