The library makes extensive use of Pandas data frames representing measurements. The data frames typically have the following columns:
benchmark, fuzzer, trial_id, time, edges_covered, ...
We can differentiate three types of data frames described below.
An "experiment" represents a data frame that includes measurements on multiple benchmarks, multiple fuzzers, and multiple trials.
Functions taking experiment data have parameter name experiment_df.
A "benchmark" represents a data frame that includes measurements on single benchmark. (It can still have multiple fuzzers, and multiple trials.)
Functions taking benchmark data have parameter name benchmark_df.
To compare fuzzers, e.g., based on the coverage they reached, we need to select a point in time along the course of the (typically 24 hours long) experiment, to take the coverage measurements from.
We call this time the snapshot time.
Typically we are interested in the maximum coverage reachable by a fuzzer, in which case the best is to pick the latest possible time.
The measurements taken at that selected snapshot time will be the input of most of our analysis (e.g., for ranking, statistical tests).
Therefore a "snapshot" represents a data frame where all time values are the same for a given benchmark.
Functions taking a snapshot of a benchmark have parameter name benchmark_snapshot_df.
Functions taking the snapshot of all benchmarks in an experiment have parameter name experiment_snapshot_df.