When numeric sorting is enabled for a dimension, all of the dimension values are assumed to consist of a numeric (double) part, followed by an optional non-numeric part. That is to say, 3 is evaluated as <3.0, "">. The non-numeric part is used as a secondary sort key when two or more numeric parts are equal. The non-numeric parts are sorted so that an empty non-numeric part comes first in the sort order.
In some cases, a set of primarily numeric dimension values may contain semi-numeric values, such as 1.3A (evaluated as <1.3, "A">, or non-numeric values, such as Other (evaluated as <0.0, "Other">. Numeric sort on such dimension values works as follows:
For semi-numeric dimension values, dimension values with non-numeric parts are sorted after matching dimension values without non-numeric parts. For example, 1.3A appears after 1.3 when sorted.
For non-numeric dimension values, the missing numeric part is treated as 0.0. In a data set containing the word Other and the number 0, the system would compare 0 and Other as <0.0, ""> and <0.0, "Other"> and sort 0 before Other.
Putting all of this together, a data set consisting of Other, 1.3A, 0, 3, and 1.3 would sort as follows:
0 Other 1.3 1.3A 3