This is probably a crazy idea (and I don't currently have a use case), but could we make it easy to apply a dict to map values? For example, to make the following easier and more reliable (i.e., not rely on inlining the dict):

In [10]: v = gb.Vector.from_values(np.arange(5), np.arange(5))

In [11]: v
Out[11]:
"v_2"      nvals  size  dtype  format
gb.Vector      5     5  INT64    full
-------------------------------------
index  0  1  2  3  4
value  0  1  2  3  4

In [12]: v.apply(lambda x: {0: 0, 1: 2, 2: 4, 3: 6, 4: 8}[x]).new()
Out[12]:
"v_4"      nvals  size  dtype  format
gb.Vector      5     5  INT64    full
-------------------------------------
index  0  1  2  3  4
value  0  2  4  6  8

This requires Numba to know the types of the dictionary. Also, I don't know how easy it would be to have a default value. This may require its own method.

This is inspired by the Pandas functionality to be able to use x.map(d) to apply a dict or Series mapping.