While debugging various issues you often get relevant data as always-increasing counters, like the Linux VM page fault stats:
$ for i in `seq 1 5` ; do grep fault /proc/vmstat ; sleep 1 ; done
pgfault 78258716080
pgmajfault 3202798
thp_fault_alloc 0
thp_fault_fallback 0
pgfault 78258733001
pgmajfault 3202799
thp_fault_alloc 0
thp_fault_fallback 0
pgfault 78258754872
pgmajfault 3202799
thp_fault_alloc 0
thp_fault_fallback 0
pgfault 78258833733
pgmajfault 3202799
thp_fault_alloc 0
thp_fault_fallback 0
pgfault 78258865018
pgmajfault 3202799
thp_fault_alloc 0
thp_fault_fallback 0
Now, quick: how many page faults per second were there? Did any major faults happen?
Let's say 20k-ish. Or let's see exactly:
$ timeout 5 delta grep fault /proc/vmstat
pgfault 78262968904
pgmajfault 3202834
thp_fault_alloc 0
thp_fault_fallback 0
--- Thu Nov 19 18:24:00 2015
pgfault +33040
pgmajfault +0
thp_fault_alloc +0
thp_fault_fallback +0
--- Thu Nov 19 18:24:01 2015
pgfault +27092
pgmajfault +0
thp_fault_alloc +0
thp_fault_fallback +0
--- Thu Nov 19 18:24:02 2015
pgfault +7653
pgmajfault +0
thp_fault_alloc +0
thp_fault_fallback +0
--- Thu Nov 19 18:24:03 2015
pgfault +37106
pgmajfault +0
thp_fault_alloc +0
thp_fault_fallback +0
delta works on all kinds of file formats with ints or floats (except scientific notation) and does its best
to autodetect the number format for pretty output.
pip install git+https://github.com/gnosek/delta
Tested on Python 2.6, 2.7 and 3.5. Requires Click and ansicolors.