Heads-I-Lose

Not a great implementation, but it works. If no route geometry choice is
made it defaults to the default "route_geometry". If
"alternative_geometries" is specified it uses the first one of the
alternatives (because as far as I can tell there is only ever one
alternative? I supposed at some point I could extend if needed, perhaps
use a tuple argument: {"alternative_geomtries", 1}). Leaf check-in: 9c81571ac3 user: base@atomicules.co.uk tags: origin/master, trunk

Not a great implementation, but it works. If no route geometry choice is
made it defaults to the default "route_geometry". If
"alternative_geometries" is specified it uses the first one of the
alternatives (because as far as I can tell there is only ever one
alternative? I supposed at some point I could extend if needed, perhaps
use a tuple argument: {"alternative_geomtries", 1}). check-in: 6807958aad user: base@atomicules.co.uk tags: refs/stash, trunk

Which means I also need to wrap around the list if I get 16 segments.
I.e. both 0 rads and 2*Pi rads are valid.

This means I probably could have just simply added Pi (previous commit)
to reverse the journey as I can just keep wrapping around. Oh well. check-in: 598ec92c32 user: base@atomicules.co.uk tags: origin/master, trunk

Had not actually tried out my changes, or, it seems, used my brain at all
in making them.

- Can't simply just add Pi on, need to keep it within 2*Pi limit
- Also forgot to get compass direction check-in: 3022ce9541 user: base@atomicules.co.uk tags: origin/master, trunk

I don't want to have to query OSRM for th ereverse route - it's
pointless. Instead reverse the route we already have - because I don't
actually need a properly ordered reverse route, just a list of
directions and distances.

In order to do this made sense to re-jig some of the functions a bit:

- Need date and time at top level, not buried in get_weather since
reverse route or normal route is also dependent on time of day
- Don't convert to distances and compass directions in one swoop since
it is easier to reverse the heading angles rather than compass points check-in: 935d1fbeae user: base@atomicules.co.uk tags: origin/master, trunk

In other words: done for now. Those are pretty much the changes I
wanted to do. Sure they will be much room for improvements to the code,
but it works! check-in: c4d0b2bbec user: base@atomicules.co.uk tags: origin/master, trunk

- Get rid of the mixed up mess of CamelCase, underscores, prefixes,
suffixes, abbreviations, full names, etc.
- Make indenting of lists:map functions consistent
- Remove spurious whitespace check-in: 301914a6a1 user: base@atomicules.co.uk tags: origin/master, trunk

by mapping.

Need to tidy up formatting and make it consistent and sort out the
naming of variables - it's such a mess. check-in: 6912ce6208 user: base@atomicules.co.uk tags: origin/master, trunk

...for a polyline route.

I'd made silly mistakes like forgetting which order the list of tuples
were in. I.e {Distances, Wind} or {Wind, Distances}. It doesn't really
matter which order I use, but I should back sure it is consistent and
matches the variable naming.

Now need to extend this to sum up all wind directions and check for
other bugs (such as sum of distances making sense, headwinds are
actually headwinds and not tailwinds, etc). check-in: 289839a15d user: base@atomicules.co.uk tags: origin/master, trunk

Noticed the nth_wrap problem whilst developing this osrm branch as
building the list of winds would be incomplete. check-in: 915a1fb3d0 user: base@atomicules.co.uk tags: origin/master, trunk

I only noticed this just recently, which means it has been wrong for
about a year! I think it's just luck I'd never come across an issue with
it not working.

I don't know if I'd got half way through writing this and forgot to
finish it off, or whether I was just being an imbecile: If the remainder
is 0 then that should return the last item in a list, not just the
number 1. Leaf check-in: 76eac348fc user: base@atomicules.co.uk tags: trunk, v1.1.0

Hadn't realised it returns negative numbers. check-in: 0fe524de48 user: base@atomicules.co.uk tags: origin/master, trunk

Wasn't thinking straight when I first wrote that, obviously check-in: 5c494abd30 user: base@atomicules.co.uk tags: origin/master, trunk

Compiles, but doesn't yet run.

The general principle is that it will get a previously saved polyline
and convert this into a list of compass headings and distances. It will
then sum up the distances of head, side and tail winds to see which
wins.

For the time being I'm focusing on summing up headwinds only.

Adds the following functions:

- build_list_of_wind_directions

Previously this didn't exist as a standalone function, rather it was
built into the main headsilose routine. Before it was fed with a
direction of travel and based on this it would work out which 16 point
compass headings were head, side or tail winds. Now it needs to work a
bit in reverse (although I haven't made the necessary changes yet - I
don't think1) and instead feed it with the actual wind direction and
it will use this to build the list of head, side and tail winds.

- get_compass_direction_for

Convert an angle into a 16-point compass direction

- head_side_or_tail_wind

Decided whether, for a given direction of heading, it is a head, side
or tailwind. Uses the lists from build_list_of_wind_directions.

Commiting this in a back-to-front way as I actually first started with
this file before doing the polyline branch. But it was as doing this
that I realised I was going to need a polyline decoder after all because
the API didn't actually return the route instructions and so I'd have to
use the route geometry instead.

So this is just a "works enough to develop further" commit so I can work
on the changes to headsilose.erl next. check-in: a0c2266624 user: base@atomicules.co.uk tags: origin/master, trunk

As per the Encoded Polyline Algorithm Format page:
https://developers.google.com/maps/documentation/utilities/polylinealgorithm

- Previously the decoder only worked with a single point (baby steps and
all that) and was tested with the example point of `~oia@
- This commit extends it to work with the example of
_p~iF~ps|U_ulLnnqC_mqNvxq`@
- Probably still to do is figure out decoding an unsigned value and the
bit about string literals.

Basically alter functions to use a list of lists approach:

- Adds split_up_six_bits which does what it says on the tin and looks
for the absence of the 0x20 bit and creates a list of lists of each
point's bit chunks.
- five_bit_chunks is now called six_bit_chunks since the groups of
chunks have to be split up at the six bit stage.
- Similarly, five_bit_chunk is now called six_bit_chunk; it is otherwise
exactly the same except from not "un-or-ing" the 0x20 bit to go from
six bits to five bits. Still padding to six bits, but I wonder if this
is entirely unnecessary and I can use this to determine the absence of
the 0x20 bit and thus where to split up the chunks into groups?
- five_bit_chunks now does the "un-or-ing" by mapping over the list of
lists. Perhaps this could be bundled in with the rest of the big
lists:map in the top level decode function?
- most of what was in the top level decode function remains the same,
but is just mapped over a list now. These should probably be at least
pulled out as separate functions that are mapped over. Need to decide
which approach to go for: functions that map or mapping functions.
- Changes to reversing, specifically the un-reversing of the five bit
chunks. This was bundled in with the five_bit_chunks function and the
un-reversing was achieved by simply not doing lists:reverse since the
function naturally reversed it. Now, to follow the algorithm backwards
step by step, there is now a lists:reverse at the end of
six_bit_chunks which then has to be undone by including a
lists:reverse in five_bit_chunks.
- Add in algorithm step numbers to comments. Leaf check-in: 8d51987b6f user: base@atomicules.co.uk tags: polyline, trunk