#136 File/Directory Change Detection

[Thomas Navara]

When you start cleaning your workspace, you might be moving files and directories, maybe rename aswell. In that case, dar does almost full backup, instead of incremental. Because dar stores file checksums, it know that two files are same.

Thus, when finding known (during incremental archiving against former archive) file in different part of filesystem tree, only it's movement shall be noted, not storing whole file again.

[sampablokuper]

Is anybody working to implement this feature request?

[Denis Corbin]

Hi,

detecting moved files would cost a lot of CPU and/or memory: Assuming you have N files under backup, for each missing file, it would require N comparison, this a complexity in O(n2). If just one file is found to be missing during a differential backup (which target is to be quick), dar would have to read the whole filesystem again, doing hash on each file (to save time if a second file is missing too) and store this possibly large set of hash file in memory (dar policy is to not use temporary files: Adding temporary files raises other concerns like disk space, permission, and partition location, with an additional drawback on performance due to disk I/O generated).

Saving a moved file from time to time is more efficient than having a backup process much slower and much fond of memory to possibily save a few bytes (worse case scenario is moving many small files).

Cheers,
Denis