The proposal suggests to use readDir to generate the attributes.
This will reduce performance compared to a file that lists the directories.
Note that the readDircan not be lazy, despite the sharding, because // is strict in both arguments.
The file tree flattening alone is already a worthwhile change, and the readDir solution can be applied later, when we can actually measure it.
The proposal suggests to use
readDir
to generate the attributes. This will reduce performance compared to a file that lists the directories. Note that thereadDir
can not be lazy, despite the sharding, because//
is strict in both arguments.The file tree flattening alone is already a worthwhile change, and the
readDir
solution can be applied later, when we can actually measure it.