avoid indexing every time?

[wakamex]

is there a feasible way to do this? maybe it makes no sense, if displaying file size. but if i'm using this as an ls substitute, it feels weird for it to take 5 seconds every time. maybe let me set an ls alias to use a cached index, then other commands update this index?

[solidiquis]

Whether or not to leverage the cache is a decision made by the underlying read_dir API from libc; but rest assured the cache is indeed being utilized automatically. Checkout the following screenshot:

I color coordinated the highlights for the stats from iostat on the left with the timeed erd runs on the right. You'll notice two things:

• The subsequent erd run is approx 1.8 times faster than the first
• The amount of 4KB transfers per second (tps) is lower on the subsequent erd run

This all lends credence to the fact that we're not always making round-trips to the disk when disk data already exists in the cache.

Now as to why erd may not be 100% suitable as an ls alternative: Ultimately in order for us to get disk usage information of directories we need to do a full-depth traversal of a directory's descendants. Even if you were to limit the depth to 1 (-L1), erd accurately reports disk usage information which necessitates the traversal.

There IS an option to --suppress-size but this issue is still ongoing.

Would you want ls-like performance at the cost disk usage information? Could I also ask how you're using erd currently as an ls-alternative?