@@ -93,17 +93,17 @@ Here's a quick overview in order of simplicty:
93
93
* `Leaf ` -- A key-value pair.
94
94
* `Collision ` -- An array of key-value pairs where the keys have identical hash
95
95
values. Element order doesn't matter.
96
- * `Full ` -- An array of / 2^B/ children. Given a key you can find the child it
97
- is part of by taking / B / bits of the hash value for the key and indexing into
96
+ * `Full ` -- An array of * 2^B* children. Given a key you can find the child it
97
+ is part of by taking * B * bits of the hash value for the key and indexing into
98
98
the key. Which bits to use depends on the tree level.
99
99
* `BitmapIndexed ` -- Similar to above except that the array is implemented as a
100
100
sparse array (to avoid storing `Empty ` values). A bitmask and popcount is
101
101
used to convert from the index taken from the hash value, just like above, to
102
102
the actual index in the array. This node gets upgraded to a `Full ` node when
103
- it contains / 2 ^ B / elements.
103
+ it contains * 2 ^ B * elements.
104
104
105
- The number of bits of the hash value to use at each level of the tree, / B / , is a
106
- compiled time constant (i. e. 4 ). In general a larger / B / improves lookup
105
+ The number of bits of the hash value to use at each level of the tree, * B * , is a
106
+ compiled time constant (i. e. 4 ). In general a larger * B * improves lookup
107
107
performance (shallower tree) but hurts modification (large nodes to copy when
108
108
updating the spine of the tree).
109
109
0 commit comments