Figure 15 shows how changes in the standard deviation of the normal distribution affect the Insert operation for the various methods.
Figure 15: Insert times as the standard ceviation varies
Only the Gap method and the FsbG are affected and only at much higher standard deviations that one would expect in normal text editing. The Arr and FsbA methods are not shown (because they are too large) but they are unaffected by increases in the standard deviation of the normal distribution.
The ItemAt operation is unaffected by increases in the standard deviation of the normal distribution.
Figure 16 shows how the Insert operation is affected by changes in the percent of edit locations that are taken from a uniform distribution over the entire sequence (that is, where the next edit is randomly located in the sequence instead of instead of being normally distributed around the location of the previous edit).
Figure 16: Insert times as the percent of uniform jumps varies
Only the Gap method is affected.
Figure 17 shows how the ItemAt operation is affected by changes in the percent of edit locations that are taken from a uniform distribution over the entire sequence (that is, where the next edit is randomly located in the sequence).
Figure 17: ItemAt times as the percent of uniform jumps varies
Only the Piece and List methods are affected but only in ranges that one would not expect to find in normal text editing.
Figure 18 shows how the buffer size affects
Figure 18: Insert times as the size of the buffer increases
the time taken by the Insert operation in the FsbA and FsbG methods. The FsbG method is unaffected by the buffer size while the FsbA method goes up linearly (and sharply) as the buffer size increases. The increase levels off at 8000 where the buffer size is equal to the sequence size and so the entire sequence is in one buffer and the method has degenerated into the Arr method.
The following table gives the general trends of the results. The units vary from machine to machine but the ratios were reasonably steady. Some of the results have wide ranges. This means that the figure depends on one or more of: the size of the file being editing, the distribution of the position of edits in the sequence, and the size of the buffers (for the Fsb method).