Take the 2-minute tour ×
Programmers Stack Exchange is a question and answer site for professional programmers interested in conceptual questions about software development. It's 100% free, no registration required.

In my experience, trying to ensure that new versions of an application retain compatibility with data storage from previous versions can often be a painful process.

What I currently do is to save a version number for each 'unit' of data (be it a file, database row/table, or whatever) and ensure that the version number gets updated each time the data changes in some way. I also create methods to convert from v1 to v2, v2 to v3, and so on. That way, if I'm at v7 and I encounter a v3 file, I can do v3->v4->v5->v6->v7.

So far this approach seems to be working out well, but I haven't had to make use of it extensively yet so there may be unforseen problems. I'm also concerned that if the objects I'm loading change significantly, I'll either have to keep around old versions of the classes or face updating all my conversion methods to handle the new class definition.

Is my approach sound? Are there other/better approaches I could be using? Are there any design patterns applicable to this problem?

share|improve this question
add comment

3 Answers

up vote 5 down vote accepted

You're doing it right. You're stamping data with its version, which means you have a definite interpretation of it. The only open question is how to handle "old" data. Your choices are essentially between upgrading data where it lives, having your code adapt the data in realtime, or having the code handle multiple data versions. From 30+ years experience, I can tell you the former is the only sane way to go. Bite the bullet and write a conversion routine for each step along the history, and run them in sequence. If you find that a later step obviates an earlier one (e.g., why upgrade the rows in a table if a later step deletes the table?), resist the temptation to short-circuit things unless there is a large, demonstrable performance gain in the update process.

share|improve this answer
add comment

If your objects is just data holders, you could use the Google protobuff library or something similar. For each member of the class/object exists a property that indicates if the member exists or not. For example a objects version V1 has a property called DataForV1 and a property called HasDataForV1. The version V2 does not use the DataForV1 anymore so HasDataForV1 will always returns false and the DataForV1 property could contains null or a default value.

share|improve this answer
    
Thanks for your input, but while your technique can help code to be more tolerant of differently-formatted data, it doesn't address the problem of dealing with increased complexity when there are multiple old versions to deal with. I've dealt with code which just takes the solution you describe, and have found that it can quickly turn into an unmanageable/unmaintainable mess. –  Baqueta Apr 5 '12 at 9:08
add comment

that is a decent system besided the code bloat for upgrading data

my advise is to upgrade as much as possible while remaining lazy about it (you read a object from a previous version->you overwrite it with the same (now upgraded) object from the current version)

but you will run into trouble if you ever tighten up the data invariants or simply do a complete overhoal so an older version cannot be upgraded

share|improve this answer
    
This is an especially difficult technique when you have referential integrity constraints between elements that may or may not have been upgraded. –  Ross Patterson Apr 5 '12 at 20:54
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.