A library for dealing with human-readable time offsets and timestamps
tslib is a timestamp processing toolkit and library, written in Python. It is very helpful when dealing with the millisecond-precision timestamps that we encounter all the time in computer systems.
tslib depends on pytz, which you can install easily using the provided requirements.txt file:
$ pip install -r requirements.txt
You’re now ready to use ts. For easier access, put its directory into your $PATH:
$ export PATH=`pwd`:$PATH $ echo "export PATH=`pwd`:\$PATH" >> ~/.bashrc # or ~/.zshrc
$ ts -h usage: ts [-h] [-t TZ] [-n] [-i] Human readable timestamps optional arguments: -h, --help show this help message and exit -t TZ, --timezone TZ Use TZ as the local timezone -n, --timestamp-only Only output the resulting millisecond timestamp(s) -i, --inline Apply timestamp replacements inline of the incoming text
ts transforms milliseconds timestamps and human-readable deltas into fully qualified date and time printouts, containing:
The output is tab-delimited and can easily be re-segmented by piping into column -t -s "\t".
Arguments can be passed as command-line arguments, or piped into ts via STDIN. If no input comes from either of these sources, ts will simply output the current time:
$ ts 1423599084206 2015-02-10 20:11:24.206 UTC+0000 2015-02-10 12:11:24.206 PST-0800 0
ts can operate in inline-replacement mode, for timestamps only. In this mode, you can pipe in any text and ts will output it back to you with all timestamps replaced with a human-readable date and time and human-readable delta representation.
Given the inherent difficulty of accurately matching timestamps, the matching is limited to 13-digits millisecond precision timestamps (that may be immediately followed by a L). Numbers smaller, or larger than 13 digits are not matched, and are left alone; meaning no 13-digit section of them is matched and replaced.
$ echo B_b-3PAAYAA B_cJXlvAYAA B_cJXl6AcAA \ | xargs sfc mb g -p TSVH -f sf_checkpointTimestampMs -f sf_updatedOnMs \ | ts -i \ | column -t -s " " sf_id sf_checkpointTimestampMs sf_updatedOnMs B_b-3PAAYAA 2015-03-06 13:38:58.851 PST-0800 (-2w5h18m6s210) 2015-03-06 13:38:58.852 PST-0800 (-2w5h18m6s209) B_cJXlvAYAA 2015-03-08 09:04:36.218 PDT-0700 (-1w5d10h52m28s843) 2015-03-08 09:04:36.218 PDT-0700 (-1w5d10h52m28s844) B_cJXl6AcAA 2015-03-08 09:04:34.252 PDT-0700 (-1w5d10h52m30s810) 2015-03-08 09:04:34.342 PDT-0700 (-1w5d10h52m30s720)
$ cat << EOF | ts pipe heredoc> 1404424797009L pipe heredoc> 1415917836779L pipe heredoc> EOF 1404424797009 2014-07-03 21:59:57.009 UTC+0000 2014-07-03 14:59:57.009 PDT-0700 -31w5d1h14m26s38 1415917836779 2014-11-13 22:30:36.779 UTC+0000 2014-11-13 14:30:36.779 PST-0800 -12w5d43m46s268
Using the column output (assuming the same input in a /tmp/ts.txt file):
$ cat /tmp/ts.txt | ts | cut -f4 -31w5d1h16m17s703 -12w5d45m37s933
Sorting on a column (here, by descending timestamps in first column):
$ cat /tmp/ts.txt | ts | sort -k1 -t $'\t' -r 1415917836779 2014-11-13 22:30:36.779 UTC+0000 2014-11-13 14:30:36.779 PST-0800 -12w5d46m6s924 1404424797009 2014-07-03 21:59:57.009 UTC+0000 2014-07-03 14:59:57.009 PDT-0700 -31w5d1h16m46s694
Human-readable deltas can be expressed in weeks (w), days (d), hours (h), minutes (m), and seconds (s). The remainder, without a unit, is assumed to be milliseconds. Any “segment” can be omitted, the only requirement is that the segments that are specified are written in descending order of span (days before hours, hours before minutes, etc.).
Here’s an example: -12w4d6m57s257. Note that hours are missing, which simply means 12 weeks, 4 days, 6 minutes, 57 seconds and 257 milliseconds.
As you might have guessed, deltas can be both negative and positive. For positive deltas, the leading + may be omitted if units are used, otherwise the number is assumed to be an absolute timestamp:
$ ts -1 1423599850752 2015-02-10 20:24:10.752 UTC+0000 2015-02-10 12:24:10.752 PST-0800 -1 $ ts 0 1 0 1970-01-01 00:00:00.000 UTC+0000 1969-12-31 16:00:00.000 PST-0800 -2353w5d20h24m14s145 1 1970-01-01 00:00:00.001 UTC+0000 1969-12-31 16:00:00.001 PST-0800 -2353w5d20h24m14s144 $ ts +1 1423599855941 2015-02-10 20:24:15.941 UTC+0000 2015-02-10 12:24:15.941 PST-0800 1
The third column shows the timestamp’s representation in local time. It defaults to the US/Pacific timezone but this can be overridden with the -t command-line argument, passing in a timezone name that pytz understands:
$ ts -t Europe/Paris 1423600015955 2015-02-10 20:26:55.955 UTC+0000 2015-02-10 21:26:55.955 CET+0100 0
By prefixing a human-readable delta with an equal sign (=), you obtain an absolute offset from the Epoch. The side-effect of this is that it allows for converting a human-readable delta into its corresponding millisecond duration.
$ ts -n '=1h' 3600000 $ ts '=1d' 86400000 1970-01-02 00:00:00.000 UTC+0000 1970-01-01 16:00:00.000 PST-0800 -2365w6d21h56m20s98