Rustics

A minimal statistics library for performance monitoring

Rustics implements a simple set of statistics objects intended primarily for measuring performance parameters. It provides statistics objects that collect data and compute various statistical characterizations of the data, such as the mean and variance.

At this time, statistics can be gathered for i64 values and for time values. They can be collected as a running total of all samples recorded, or as a window covering the last N events, for a programmable N. The interface does support a "clear" function to discard all gather statistics. See the time.rs file for information on timers and to see some sample implementations.

This library also implements sets, which contain statistics that are printed and manipulated together. It provides sets contained in Arc structures and Rc structures. The two set modules are otherwise identical. The implementations are simple and will not scale to handle a single set with a very large number of elements, especially if these elements are added or deleted frequently.

Sets are recursive: a set can have a set as a member. A set hierarchy can be printed with a single procedure call, and the contained statistics object can be cleared in the same way. The set implementation automatically creates hierarchical names for objects. Each set or statistic typically will have a single element name given by the user, and the string "==>" will be used to create a hierarchical name in the manner of a Unix filesystem by concatenating the names as the print process descends the tree. For example, if a tree has one set named "Everything" and that set contains one statistics named "Lonely", the title printed will be "Everything ==> Lonely".

Statistics and sets can be deleted, but the code uses simple algorithms that are linear in the number of elements of a set to do searching for items to be deleted, so sets with a very high number of subsets or statistics can require a long time to process a deletion.