A
SizeSequence object efficiently maintains an ordered list of sizes and corresponding positions. One situation for which
SizeSequence might be appropriate is in a component that displays multiple rows of unequal size. In this case, a single
SizeSequence object could be used to track the heights and Y positions of all rows.
Another example would be a multi-column component, such as a JTable, in which the column sizes are not all equal. The JTable might use a single SizeSequence object to store the widths and X positions of all the columns. The JTable could then use the SizeSequence object to find the column corresponding to a certain position. The JTable could update the SizeSequence object whenever one or more column sizes changed.
The following figure shows the relationship between size and position data for a multi-column component.
In the figure, the first index (0) corresponds to the first column, the second index (1) to the second column, and so on. The first column's position starts at 0, and the column occupies size0 pixels, where size0 is the value returned by getSize(0). Thus, the first column ends at size0 - 1. The second column then begins at the position size0 and occupies size1 (getSize(1)) pixels.
Note that a SizeSequence object simply represents intervals along an axis. In our examples, the intervals represent height or width in pixels. However, any other unit of measure (for example, time in days) could be just as valid.
Implementation Notes
Normally when storing the size and position of entries, one would choose between storing the sizes or storing their positions instead. The two common operations that are needed during rendering are:
getIndex(position) and
setSize(index, size). Whichever choice of internal format is made one of these operations is costly when the number of entries becomes large. If sizes are stored, finding the index of the entry that encloses a particular position is linear in the number of entries. If positions are stored instead, setting the size of an entry at a particular index requires updating the positions of the affected entries, which is also a linear calculation.
Like the above techniques this class holds an array of N integers internally but uses a hybrid encoding, which is halfway between the size-based and positional-based approaches. The result is a data structure that takes the same space to store the information but can perform most operations in Log(N) time instead of O(N), where N is the number of entries in the list.
Two operations that remain O(N) in the number of entries are the insertEntries and removeEntries methods, both of which are implemented by converting the internal array to a set of integer sizes, copying it into the new array, and then reforming the hybrid representation in place.
@author Philip Milne
@since 1.3