|
If the file is modified rarely or in a limited manner, I would think an insertion procedure could maintain the sort, as one would do with a linked list. If it's modified grossly, then a heap sort might be in order. There are other methods that require more storage and an initial organization, but are amenable to more effective operations subsequently. A database is a good example of a sophisticated and complex process that yields ultimate gains. My initial impression is that you should deep-six the general-purpose utility approach and go for an application dedicated to the problem. One hundred megs is not a breathtaking amount these days, but one needs to realize the problems that can arise if improper amounts of cache can lead to thrashing (page swapping unduly) and other undesirable thangys. The fact that each item is on one line is immaterial. A "line" is the result of just another delimiter. It's possible that a mere modification of your approach is in order. Consider a silly extreme: you perform a bubble sort on 10 million items when the data has been modified only slightly or not at all, then you perform a binary search. If something like this is an invariant. then there is much room for improvement. Again, specific suggestions can't be made in the absence of specific information regarding the process as well as the mere organization of the data.
|