This is a tool that reads a log minute-by-minute of a day
and a set of rules on how to interpret it, and will create
graphs and reports out of it.
Dependencies : rvg and rmagick

The log has a fairly simple format. Each line must either
be the start of a day at the format 'MM-dd', a line starting
with time at the 'HHMM' format followed by an activity, or
a free-form line to match counters that's otherwise ignored.
It looks like this :
06-08
0835 Got up, made coffee
0840 Shower
0902 Work on XXX
...
2321 Zzz
06-09
...

An optional comment can be put after the date separated by
a space and a colon :
06-08 : Not feeling well

It will also take a rule file. The rules are essentially
regexps that will be used to match activities. The rule
file is composed of multiple sections. A section is started
by a line with the section name in square brackets. Any
section can be given any number of times and the script
will merge them. All sections are optional, although a
missing [rules] section will result in no rules, meaning
any non-empty data file will fail to parse ; and a
missing [general] section will create output files with
the name "unnamed".
All section names can be suffixed with /i, which makes
any regexps in the section case-insensitive ; though all
sections support this, not all sections have regexps so
it's not really meaningful for all sections.
Any line starting with a # is considered a comment.

A [general] section can define a name (defaults to
"unnamed") and a mode among 'calendar', 'occupation',
'stack', 'count', and 'interactive' (defaults to
'calendar'). These are given with the 'name = ' and
'mode = ' directives, the format is fairly intuitive.
The name and mode directives are special in two
respects : they have to be given in the top-level file
(they are ignored in included files) and if multiple
of either directive are encountered then the latter
directive wins.
The 'interactive' mode outputs dynamic HTML files, the
'count' mode outputs text, and the other modes output
PNG files. All image-outputting modes are deprecated
now seeing how the image format is too limiting compared
to the interactivity allowed by the HTML output. I'm not
even sure they still work. See a description of the
modes below in this README.
The [general] section also supports an
'include <filename>' directive to load the given file
at the place the include rule is encountered (it
behaves as if the 'include <filename>' directive is
replaced verbatim with the content of the designated
file). Priorities between rules can be managed by having
multiple [general] sections to have the includes in the
desired order if necessary.

The [rules] section is the most important section as it
defines the list of rules that will be considered to
interpret the log file. Each line must be at the
following format :
<regexp> = <activity>
Spaces can be included in both the regexp and the
activity name but the script will ignore spaces around
the = sign, so an activity name can't start with a space.
Any activity name is legal. The script will simply gather
the list of activities from the list of targets of the
list of rules.
For each log line the script will find a matching regexp
from the list of rules ; if multiple rules match a given
log line, then the regexp given earlier in the rules file
(including with the includes directives) will win.
For convenience, the [rules] section can be written
[rules/i], in which case all regexps will be considered
case-insensitive.
The "Zzz" activity is special, it means sleep and the
tool will merge sleep on this night with the sleep on
the morning of the next day. Sleep can't be called
anything else (at least if this processing is desired).
In counters and histograms, sleep is counted as ending
the day, not starting it, and will be reported as being
part of the day it ends in graphs.
Newer versions of the tool will also match activity
names from the rule to the regexps in the file, in a
recursive fashion. For example, with rules
Novel = Book
Manga = Book
Detective Conan = Manga
...then a log line of 'Detective Conan' will be
considered a manga, which will in turn be considered a
book. I think this is only supported in 'interactive'
mode, at least it hasn't been tested in image mode
since this capability was added after they have been
deprecated.

Because of the recursive matching, the [collapse]
section is now deprecated, since it's just a more
complicated way of doing the same thing. Dive into the
repo history if you want to know how it used to work.

The [counters] section consists of a list of counters
at the following format :
<regexp> = <contribution> = <counter>
The regexp can be anything that doesn't contain the =
sign. The contribution has a fairly baroque but useful
format where $ signs are interpreted to be replaced
with that match in the regexp, after which the entire
thing is interpreted as an int.
The counter is just a name for the counter. It can be
any string.
So for example,
   .*meditate.* = 1 = Meditate
...will match any line containing the text 'meditate'
and that will give a contribution of 1 to the
'Meditate' counter, and
   .*do (\d+) pushups.* = $1 = Pushups
...will match any line containing a number of pushups
and it will contribute for that number of pushups to
the relevant counter, and
   .*cheat (\d+) times = 1$1 = Cheat
...will count 13 cheats for a line 'cheat 3 times'.

The [markers] rules defines a list of markers at the
following format :
<regexp> = <policy> = <marker>
The regexp can be anything that doesn't contain the
'=' sign, and the marker can be any string.
The policy must be either 'First', 'Last' or 'Each'.
This defines what to do when multiple strings
matching the regexp appear during a day. Respectively
put the marker on the first, last of them, or on each
of them. I think this section is not implemented in
interactive mode.

The [colors] section defines a color for an activity
or marker in the output images. The format is :
<name> = <color>
...where name has to be the exact match (not a regexp
so it's case sensitive even in a [colors/i] section)
of an activity or marker name, and the color has to
be any color that ImageMagick knows how to interpret,
so typically a string at the '#FFFFFF' format or a
color name from rgb.txt like 'white' or 'red'.
Any counter or activity that doesn't have a color
defined will get one randomly assigned (from a hash
of it's name, so it's stable across invocations of
the tool).

The tool reads holidays from a 'holiday.rb' file.
This is a Ruby file with self-explanatory syntax.

The tool supports five modes, three of which are
deprecated – the 'interactive' mode is meant to show
all data image modes used to show in a nicer,
interactive way. I've made many changes since the
last time I used any other mode than 'interactive'
so it's likely they are not even working any more in
the latest version.
• Interactive mode
  This mode produces a static HTML file which can be
  opened in a browser and show a primary view similar
  to the 'calendar' mode, and a secondary view with
  a pie chart showing time use.
• Calendar mode
  This mode produces a calendar-style output. The day
  starts at 05:00 and ends at 29:00. Some counting
  might be a bit strange if not sleeping at the time,
  and the tool doesn't really make an effort to avoid
  painting above or below the graph.
  Each activity will be represented as a block of the
  associated color on the calendar. Each marker will
  be represented by a solid bar at the time it
  happens. The tool will output a legend with the
  name on the top and hours on the left, and the days
  on the bottom with their weekday kanji for ease of
  reading. Holidays are in red while worked days are
  white. Below that, a legend will show the colors
  for each activity and each marker. The tool spaces
  these evenly but doesn't particularly avoid them
  overlapping, so many activities/markers for short
  calendars will be difficult to read.
  This mode ignores counters.
• Occupation mode
  This mode produces a histogram with a bar for each
  day and each activity. It uses the same legend
  scheme as the Calendar mode to show days and
  activities, but hours start at 0 from the bottom.
  For each day, each activity gets a bar of the
  relevant color with its size scaled to how much
  time went into it, with the exact number on top of
  the bar. Judicious use of collapses will make such
  a graph more useful (a histogram with many
  activites per day is very hard to read). ; not
  sure how this even behaves since the deprecation
  of collapses.
  The total won't always add to 24h a day unless the
  log indicates the same time of getting up every
  day. This is because the tool will always count
  sleep as ending the day, so if the log indicates
  getting up at 07:00 the first day, 08:00 the
  second day and 06:00 the third day, then the first
  day lasts 25 hours and the second day 22.
  This mode ignores counters and markers.
• Stack mode
  This is a very similar mode to occupation but it
  will stack activities on top of each other instead
  of lining them up horizontally. This makes some
  comparisons of the time spent on a given activity
  across time harder and others easier. The order
  of stacking is alphabetical by activity and there
  is currently no way to change that. Not all days
  will have the same length for the same reasons the
  histogram won't add up to 24 hours a day. This is
  otherwise identical to the histogram mode.
• Count mode
  This will produce a text summary of the period in
  the log (or the restricted period). It will show
  on its standard output for each activity the total
  time spent on it, and a breakdown per
  workday/holiday, each with a total and an average.
  It will then produce a summary of the counters,
  ordered by name, by average time, by total time
  and count.
  This mode ignores the markers.

Invocation is :
graph.rb [options] file...

file... can be any number of log files. They are
searched in the local directory and under data/ if
not found. The tool will also try to append the
'.log' extension if a file with the specified name
is not found. If all fails, an error is produced.

Many options are supported to give different
rules files, allow restricting the data periods
read from the logs, help with debugging rules and
offer one-shot harvesting of specific data without
having to touch the rules files. See documentation
at the top of graph.rb for details.


An example rules file might look like :
[general]
name = my_calendar
mode = calendar

# Include my standard rules
include standard.grc

[rules/i]
Youtube channel .* = Youtube
Nap = Zzz
Meeting (with|about) .* = Meeting
Lunch with (Fred|Barney) = Rest
Dinner (with (Fred|Barney)|at .*) = Rest

[markers/i]
.* = First = Get up
Zzz.* = Last = Sleep
.*Meditat(e|ion).* = Each = Meditate

[colors]
# Activities
? = #3F3F3F
Rest = #359330
Work = #CF6D00
Meeting = #BF6DB0
Zzz = #003C71

# Markers
Get up = #00703C
Sleep = #003C71
Meditate = #CFCFCF

[collapse]
Facebook = Twitter = Social media
Chat = Social media = Youtube = Rest
WFH = Work

[general]
# File containing specific rules to collapse further,
# only for calendar rule files.
include collapse_calendar.grc