Save column pointers as in the hashing code, for the output and for all the inputs.
Once we have only array pointers we can assign in parallel.
List-cols involve the R API and so must be assigned separately.
df-cols could in principle be assigned in parallel too, but this would be a bit more complex.
Should make a big improvement when recombining data frames with vec_unchop().
Technical challenge: Figure out recursive proxying and restoration. Will require keeping a tree of restoration targets. Also need to think through interaction with method composition of proxy and restore.
Save column pointers as in the hashing code, for the output and for all the inputs.
Once we have only array pointers we can assign in parallel.
List-cols involve the R API and so must be assigned separately.
df-cols could in principle be assigned in parallel too, but this would be a bit more complex.
Should make a big improvement when recombining data frames with
vec_unchop()
.Technical challenge: Figure out recursive proxying and restoration. Will require keeping a tree of restoration targets. Also need to think through interaction with method composition of proxy and restore.