I should add: this messy code reveals an approach which I use a lot: Concise, elegant J is for problems which I have a good or at least ok understanding of. (The distinction here, between a good understanding and an ok understanding, is perhaps best illustrated by the quip "If you can't explain it to a six year old, you don't understand it yourself.")
But when I do not have something working, then "whatever works" is a rough approximation of where I need to start, and baby steps are often appropriate. Here, I would normally have gone back and spent some more time on it, built up my understanding of the concepts and cleaned up the code. But... I'm still behind on day 24. (Among other things...) So, instead, I am hoping to use this as an opportunity to engage someone else in this puzzle. FYI, -- Raul On Thu, Jan 6, 2022 at 9:33 PM Raul Miller <rauldmil...@gmail.com> wrote: > > https://adventofcode.com/2021/day/17 > > Day 17's puzzle was a change of pace from previous puzzles. Instead of > a multiline chunk of text representing the puzzle, we got a single > line, which looked like this: > > target area: x=20..30, y=-10..-5 > > All that would vary would be the numbers. > > The puzzle itself was about a quasi-physics simulation process, with > integer positions, velocities and accelerations. > > The target area represents a rectangular region on a two dimensional > map. We get to "launch a probe" with some initial x velocity and y > velocity. It starts at time=0 at location x=0, y=0. > > Additionally, after each step, x velocity is reduced by 1 in magnitude > until it reaches 0. And, after each step, y velocity is reduced by 1 > and will can reach arbitrary negative values. > > Additionally, since this is an integer based simulation, our probe > would fall or fly right past the target area if probe's velocity is > sufficient to cross that target area, in one step. For the probe to > hit the target area, it has to be in the target area at the end of one > of it's steps. > > For part A, we were asked: how high can our probe go while still > landing in the target area. > > The key here, for our puzzle input (which had a negative value range > for y where ymax-ymin was significantly smaller in magnitude than > ymin) was to realize that "maximum height" corresponded to a velocity > moved the probe from position 0 (which is where it would fall to after > being launched upwards) to the bottom of our puzzle input. > > Our launch velocity would be 1 unit less than that (since each step > increases the proble's negative velocity by 1). > > In other words, given trian=: -:@* >: for our puzzle input the height > for part A would be trian >./|targetY > > My implementation for part B was nothing nearly as elegant. My problem > was visualizing the possibilities. Also, I did not document my > thinking adequately, and I mostly used a trial and error approach to > quickly simulate the possibilities. And, I included a simplification > (that Y would be significantly negative) which means that my part B > implementation does not work on the sample data (which would be > sample=: 20 30,.10 _5 ). > > Anyways... here's what I had for part B > > simy=: {{ > Y0=. {:<./m > Y1=. {:>./m > Y=.0 > k=.0 > y=.-1+y > while. Y >: Y0 do. > Y=.Y+y > if. (Y>:Y0) * Y <: Y1 do. 1 return. end. > y=.y-1 > end. > 0 > }} > > simX=: {{ > k=.0 > X=.0 > r=.i.0 > X0=: {.<./m > X1=: {.>./m > while.(X <: X1)*y>0 do. > k=.k+1 > X=. X+y > if. (X>:X0)*X<:X1 do. > r=.r,k > end. > y=.y-*y > end. > critical=: I. (X0&<: * X1&>:) trian i.X0 > if. y=0 do. ~.r,(*#r)#(#~ k&<:) critical return. end. > r > }} > > b17=:{{ > N=. 0 > 'X0 Yp0'=. <./|y > 'X1 Yp1'=. >./|y > Ypos=. I.y simy"0 i.1+Yp1 > Xref=: y simX"0 i.{.1+>./|y > V=: i.0 > for_s. i.>./critical do. > V=:V,<I.Xref+./ .= s > end. > Yneg=. 1+Ypos > N=. N+(#Ypos)*#critical > for_Y. -Yneg do. > py=. +/\(|Y)+i.20 > steps=. 1+I.(Yp0 <: py)*Yp1>:py > if. (>./critical) > <./steps do. > Vs=. i.0 > for_s.steps do. > Vs=.~.Vs,(s {::V)-.0 > end. > N=. N+#Vs NB. r=.r,>{Y;Vs > else. > N=.N+#critical > end. > end. > N > }} > > That's not a very good approach, though it did work for me... > > If someone has an elegant solution to part B, that would be great to see. > > Thanks, > > -- > Raul ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
