for historical reference -
Originally Posted by doorhinge
Left, right, left, right, left, right. Travel directly towards the exit in areas where there are no cars.
Of course you’re going to hit a row of cars, you’re walking thru a parking lot. It’s a bit like sailing a boat directly into the wind. You have to jibe port and starboard until you reach your intended destination.
The asterisks are on the optimal path, but your sub-paths aren’t all subsets of that path, but will still be a shorter distance than pure rectilinear motion - is how I read it. For instance, in that example, I’d make my path out to be one steeper diagonals and one shallower one, and then a second set of same:
Starting top left, I’d go down one row against the left edge, across between the cars, then diagonally across the aisle through the asterisk (in the centre of the aisle) to the start of the fourth-fifth inter-car space. That’s a “steep” diagonal. Across in that space, and then through the asterisk gets me to the 5-6 space. That’s a “shallow” diagonal. Across 5-6, then diagonal to the 8-9 (another “steep” one), then across and diagonal to the 9-10 (“shallow” slope) and then either down and across or across and down, to the exit.
Did that make any sense? I think there are probably other combos (shallow-steep-shallow-steep also look like it might work and be the same length, and that’s not considering doglegs while between the cars)
Oh, and septimus, that’s a lovely diagram.
How can people get out of their cars if there isn’t even enough room to walk between them?
You can walk faster if you don’t have to worry about hitting other peoples cars if you swing your arms too wide or take one false step.
Thank you! 
Chessic Sense is correct that I should have spent more effort making it more exact :o , but I was in a hurry to convey the general concept.
I should photograph that parking lot and post pictures. It’s an almost unbelievable parking lot, with cars also parked perpendicular blocking the paths. Getting out can be like a Traffic Jam puzzle. (For that reason I never park in the main lot, but park in one of the distant annexes.)
The “can’t fit between cars” problem applies after shopping when I’m returning to car with a shopping cart.
You can walk diagonal down your vertical aisle, and then diagonal across the horizontal to the entrance. But those are two rectangle diagonals. The diagonal of one square (or rectangle with almost equal side lengths) could be much shorter. But your zigzags to get past aisles of cars would add some extra travel distance. (If they are parked slanted and the parking lot is full, you would have to zigzag every other aisle of cars, since every other aisle switches direction of traffic. On the “good” aisles you can take the same diagonal passing between cars on each side. On the “bad” aisles, you need to walk away from the store on the first side, but then can turn to walk towards the store in the middle to take a diagonal between cars on the other side.) So it depends on:
Here is an example library (there are many different libraries out there, and various algorithms like Dijkstra’s algorithm, the Fast Marching Method, etc.) that can do some path planning, in case anyone wants to experiment with different layouts and obstacles on the computer.
None of you have even considered an issue some of us have: that it is not uncommon to forget exactly where we have parked our cars. All of this diagonal travel most of you recommend makes remembering the exact location more and more difficult–so any time we save on the initial trip is more than lost on the return trip. Thus Snfaulkner is correct: “Down one row until you can’t, then over to the exit.”
Maybe you can ride one of the elephants up and over? Once they hatch, I mean.
Then I believe we have a debate, because I am of the belief that if you change your “pitch,” which you’re calling steep and shallow, then you cannot be on the optimal path. All your pitches should be equal, unless you’re going to hit the center of a bumper.
Well, duh! Many (most?) cars have only a single occupant – the driver. So you can park as close as you want on the passenger side. Sometimes that makes it hard for the person parked next to you to get back into their car, but many parkers don’t seem to care.
Septimus’ superb diagram had me so confused I didn’t even want to address it. But then I realized the problem was line wrap on my phone. If I hold my phone horizontal there is more room for each line, and the diagram is shown as intended, and I agree for that parking lot.
My Tamil-speaking mother-in-law says there’s no need to curse fifty times!
A single pitch would be the *ideal *path, but the need to move rectilinearly between cars means that the most optimal *possible *paths are *not *the ideal. Unless you had square cars, maybe? Like I said, I think if you dogleged between cars, you might have a more even pitch and be closer to ideal.
Unless you can take a picture of the car park at the time you park and then spend time calculating the optimal path, you could never beat the mark1 eyeball/brain combo. This is because there will always be a few empty spaces which you can use as shortcuts.
Maybe someone could make an app for it, using an overhead CCTV feed, but it would have to constantly update as people empty and fill spaces on your path. There would also be a high risk of being mowed down by someone looking for a space while you are looking at your screen.
I won’t even mention the complication of a herringbone layout 
My thought-experiment is to first imagine you park in one corner of an empty parking lot with your destination at the far corner. The shortest path to your destination is obviously a straight line diagonally across the lot. Now imagine that 20 percent of the spots are filled with cars. It’s still readily apparent that the shortest path is the straightest diagonal, even though you would have to sidestep a few cars. Now imagine slowly adding cars until the lot is filled. At what point, and for what reason, would the most diagonal path not still be the shortest possible route?
This assumes that one can always pass unobstructed between any two parallel vehicles, which of course is far from true in the real world.
It was considered 9 posts before yours. See post #19.