Formatting Reports with Template Toolkit

Randal L. Schwartz

Recently, a Usenet posting (yes, I still read that) discussed using Perl to determine the disk block usage for various users on the system. Apparently, the poster had noticed a discrepancy between the number of blocks reported by du and the total blocks returned by summing Perl's -s function applied to each file. I replied that you can't use -s, because that's merely the length of the file, and thanks to sparse Unix files, the value wasn't necessarily the number of blocks used. Also, a large file consumes indirect blocks to locate the blocks as needed.

Because of sparse files and indirect blocks, the accurate way to determine the actual cost of a file requires calling stat and using the blocks value (element index 12). As I was thinking about this, I thought it'd be interesting to profile my own disk usage organized according to logarithmic buckets. Once I had that working, I started tinkering some more (as I often do) and wanted a nice HTML table output, organized by user, of the disk blocks and file count for a given directory hierarchy. And the result is shown in Listing 1 .

Line 5 of the listing provides a constant to help me scale a given block count into the relevant bucket. If I take the natural log of the block count and divide it by the natural log of 2, then truncate that downward to an integer, I'll get buckets for 1 block, 2-3 blocks, 4-7 blocks, 8-15 blocks, and so on, in nice powers of two. Rather than keep dividing by the natural log of 2, I'll use the reciprocal and multiply, which in hindsight is a completely unnoticed micro-optimization.