Re: du: cache directory size

[Pádraig Brady]

Mildred wrote:
> Hi,
> 
> I noticed that du takes a long time to scan directories and measure
> disk usage.
> 
> I was thinking that perhaps caching the size of directories could bu
> useful. Perhaps, after du computes the size of a directory, it could
> write its size in its extended attributes (if the filesystem support
> it). Next time, du would only compare the directory atime (or mtime
> perhaps) with the time of the scan. That could possibly save some time
> during the scan.

Those values would be invalidated though once a change
has been written anywhere in the tree under a directory.

> Perhaps this is not something to be included in du itself, but I
> thought it could be a good idea to give the idea anyway.

The idea might be appropriate on a per file basis.

For example, md5sum which needs to do significant processing
on all of a file's data, could cache the computed value in
an extended attribute. You would need support in the filesystem
though, to invalidate certain extended attributes on write.

Pádraig.