52
\$\begingroup\$

The challenge

The shortest code by character count to help Robot find kitten in the fewest steps possible.

Golfers, this is a time of crisis - Kitten gone missing and it's Robot's job to find it! Robot needs to reach Kitten in the shortest path possible. However, there are a lot of obstacles in Robot's way, and he needs you to program a solution for him.

Robot used to have a program do it for him, but that program was lost and Robot have no backup :(.

Robot's runtime is not the best, and the least characters Robot has to read from the source code, the least time it will spend processing, and that means Kitten will be found faster!

Robot's memory contains a map of the location he is currently in with the top representing North, bottom representing South, right representing East and left representing West. Robot is always in a rectangular room of an unknown size surrounded by walls, represented by # in his radar map. Areas Robot can walk in are represented by a space .

Robot's radar also scans for many obstacles in the room and marks them in various ASCII letters. Robot cannot walk across those obstacles. The radar will mark Kitten as the special ASCII character K, while Robot's location is marked with R.

Robot's navigation system works this way: He can understand a duo of direction and number of movement units he should travel to - for example, N 3 means 'go north 3 movement units'. Robot's radar map is made such that a movement unit is one ASCII character. Robot can only go in 4 directions and cannot travel diagonally.

Your task, brave Kitten saver, is to read Robot's radar map once, and output the least amount of directions, with the least movement unit travel distance. Robot is guaranteed to have at least one path to Kitten.

To ensure Robot does not waste time executing a malfunctioning program that will not help Robot find Kitten, I encourage you, brave Kitten saver, to use this output of Robot's past program to ensure no time is wasted on finding Kitten!

Test cases

Input:
    ######################
    #  d      3    Kj    #
    #                    #
    # R                  #
    #      q             #
    ######################
Output:
    E 13
    N 2

Input:
    ######################
    #  d  r   3    Kj    #
    #    p        p      #
    #         T        X #
    #      q   s   t     #
    #                    #
    #  R o    d     W    #
    #                    #
    #    g      t     U  #
    #                    #
    ######################
Output:
    N 1
    E 10
    N 4
    E 2

Input:
    ######################
    #  spdfmsdlwe9mw WEK3#
    #    we    hi        #
    #   rdf         fsszr#
    #     sdfg  gjkti    #
    #   fc  d g i        #
    #     dfg    sdd     #
    #    g        zfg    #
    #  df   df           #
    #             xcf   R#
    ######################
Output:
    N 1
    W 9
    N 5
    E 4
    N 1
    E 4
    N 1

Code count includes input/output (i.e full program).

\$\endgroup\$
  • 1
    \$\begingroup\$ Possibly inspired by this game? kongregate.com/games/Hamumu/robot-wants-kitty \$\endgroup\$ – Nabb Feb 6 '11 at 5:02
  • 1
    \$\begingroup\$ Nope, robotfindskitten.org \$\endgroup\$ – LiraNuna Feb 6 '11 at 5:05
  • 4
    \$\begingroup\$ The goal isn't unambigious. "output the least amount of directions, with the least movement unit travel distance." There might be a way with n directions and m steps and another one with less than n directions but more steps, or more directions and less steps. Is there an exchange rate? \$\endgroup\$ – user unknown Feb 6 '11 at 5:29
  • 2
    \$\begingroup\$ Number of steps is better than number of direction. \$\endgroup\$ – LiraNuna Feb 6 '11 at 6:14
  • 1
    \$\begingroup\$ If the idea is having the least number of steps, and break ties by the least number of directions, then the second example has a wrong solution. See my answer for a shortest path. \$\endgroup\$ – Daniel C. Sobral Feb 6 '11 at 18:35
10
\$\begingroup\$

C++ 1002 899 799chars

Note requires the use of C++0x to eliminate the space between > > in templates.

It does find the route with the minimum number of turns.

#include<iostream>
#include<vector>
#include<string>
#include<set>
#include<memory>
#define D make_pair
#define F first
#define S second
using namespace std;typedef pair<int,int>L;typedef vector<L>R;typedef multiset<pair<float,pair<L,R>>>B;vector<string> M;string l;int z,c,r=0;set<L> s;B b;L n;B::iterator f;R v;void A(int x,int y,int w){n=f->S.F;for(c=1;(z=M[n.S+=y][n.F+=x])==32||(z==75);++c)v.back()=D(w,c),b.insert(D(f->F+c+1./c,D(n,v)));}int main(){for(;getline(cin,l);++r){if((c=l.find(82))!=string::npos)b.insert(D(0,D(D(c,r),R())));M.push_back(l);}while(!b.empty()){f=b.begin();n=f->S.F;v=f->S.S;if(M[n.S][n.F]==75)break;if(s.find(n)==s.end()){s.insert(n);v.push_back(L());A(0,1,83);A(0,-1,78);A(1,0,69);A(-1,0,87);}b.erase(f);}for(c=v.size(),r=0;r<c;++r)n=v[r],printf("%c %d\n",n.F,n.S);}

It Dijkstra's Algorithm for solving the shortest path problem.
To distinguish between multiple equal size routes a long straight line has less weight that multiple short lines (this favors routes with less turns).

Cost of a path:  Len + 1/Len

Looking at Test Case 1:
========================
Thus Path E13 + N2 has a cost of 
      13 + 1/13 + 2 + 1/2
An alternative path E9 + N2 + E4 has a cost of
      9 + 1/9 + 2 + 1/2 + 4 + 1/4

The difference is
      Straight Path:   1/13 <   Bendy Path: (1/9 + 1/4)

In a more readable form:

#include<iostream>
#include<vector>
#include<string>
#include<set>
#include<memory>

using namespace std;
typedef pair<int,int>                   L;
typedef vector<L>                       R;
typedef multiset<pair<float,pair<L,R>>> B;
vector<string>                          M;

string      l;
int         z,c,r=0;
set<L>      s;
B           b;
L           n;
B::iterator f;
R           v;

void A(int x,int y,int w)
{
    n=f->second.first;
    for(c=1;(z=M[n.second+=y][n.first+=x])==32||(z==75);++c)
        v.back()=make_pair(w,c),
        b.insert(make_pair(f->first+c+1./c,make_pair(n,v)));
}

int main()
{
    for(;getline(cin,l);++r)
    {
        if((c=l.find(82))!=string::npos)
            b.insert(make_pair(0,make_pair(make_pair(c,r),R())));
        M.push_back(l);
    }

    while(!b.empty())
    {
        f=b.begin();
        n=f->second.first;
        v=f->second.second;

        if(M[n.second][n.first]==75)
            break;

        if(s.find(n)==s.end())
        {
            s.insert(n);
            v.push_back(L());
            A(0,1,83);
            A(0,-1,78);
            A(1,0,69);
            A(-1,0,87);
        }
        b.erase(f);
    }

    for(c=v.size(),r=0;r<c;++r)
        n=v[r],
        printf("%c %d\n",n.first,n.second);
}
\$\endgroup\$
9
\$\begingroup\$

Scala 2.8 (451 characters)

...but it doesn't solve ties in favor of least amount of directions (though it does find the least amount of steps).

val m=io.Source.stdin.getLines.map(_.toArray).toSeq
var l=m.indexWhere(_ contains'R')
var c=m(l)indexOf'R'
val q=collection.mutable.Queue(l->c->"")
def s{val((l,c),p)=q.dequeue
if("R "contains m(l)(c))for((i,j,k)<-Seq((-1,0,"N"),(0,1,"E"),(1,0,"S"),(0,-1,"W")))q.enqueue(((l+i,c+j),p+k))
m(l)(c)='X'}
def P(s:String){if(s.nonEmpty){val (h,t)=s.span(s.head==)
println(s.head+" "+h.size)
P(t)}}
def Q{s
val((l,c),p)=q.head
if (m(l)(c)=='K')P(p)else Q}
Q

Scala 2.8, 642 characters, solves ties correctly;

val m=io.Source.stdin.getLines.toSeq
var b=Map(0->0->0).withDefault(_=>Int.MaxValue)
var l=m.indexWhere(_ contains'R')
var c=m(l)indexOf'R'
val q=collection.mutable.PriorityQueue(l->c->List((' ',0)))(Ordering.by(t=>(-t._2.map(_._2).sum,-t._2.size)))
def s{val(p@(l,c),u@(h@(d,n))::t)=q.dequeue
if("R ".contains(m(l)(c))&&u.map(_._2).sum<=b(p)){b=b.updated(p,u.map(_._2).sum)
for((i,j,k)<-Seq((-1,0,'N'),(0,1,'E'),(1,0,'S'),(0,-1,'W'))){
val o=if(k==d)(d,n+1)::t else (k,1)::h::t
q.enqueue(((l+i,c+j),o))}}}
def P(l:List[(Char,Int)]){l.reverse.tail.foreach{t=>println(t._1+" "+t._2)}}
def Q{s;val((l,c),p)=q.head;if (m(l)(c)=='K')P(p)else Q}
Q

It discovered a shorter path for the second example than the one given in the problem:

N 1
E 10
N 4
E 2
\$\endgroup\$
4
\$\begingroup\$

Python 2.6 (504 characters)

import sys,itertools
W,Q=len,list   
M=[]
B=[]
for l in sys.stdin.readlines():
    r=Q(l.strip())
    B.append([0]*W(r))
    M.append(r)
    if "R" in r:x,y=r.index("R"),W(B)-1
def S(B,M,x,y,P):
    c=M[y][x]
    b=B[y][x]
    if b and W(P)>b:return 0
    B[y][x]=W(P)
    if c=="K":return P  
    elif c=="R" and P:return 0
    if c in "R ":
        b=[]
        for q,w,s in((1,0,"E"),(-1,0,"W"),(0,-1,"N"),(0,1,"S")):
            r=S(B,M,x+q,y+w,P+s)
            if r and(W(r)<W(b)or not b):b=r
        if b:return b
    return 0
print "\n".join(k+str(W(Q(g)))for k,g in itertools.groupby(S(B,M,x,y,"")))
\$\endgroup\$
  • \$\begingroup\$ This doesn't seem to solve ties in favor of least steps. \$\endgroup\$ – Daniel C. Sobral Feb 6 '11 at 17:12
4
\$\begingroup\$

Python 2.6 (535 characters)

exec 'eJzNlLFugzAQhneewhki2/KBworksVOkDu3QgVqVE2hBioFgCDhV371nJ1VbKR3SKYtl/jvs77MwtenafiBVqbs9WGcjLW05MB69tj05QEfqhpTNaMpeDyXDhsQORd3wLCK+YwT7u6PzFVK/Ep9Tsn6gmU50UTA2woHzc01KuqYZ20PPZSh85/iCO2etzBnTG8tcvlLxnovTPFVxzyGEC4lpiBay5xiuYMXBcRVtzxqTfP+IpqrelaRFsheoYJbBNvFj13asxd23gXHGmZU7bTaFDgiZH+MUYydtKBuZRuS0nvPmOt564Sl3CmlxcWAG6D3lXIkpeUMGB7nyfj82HW3FWvjTTVSYCXNJEUupEimannu+nl04WyM8XoB1F13E9S6Pt+ki0vDZXOdyd5su8X9cnm7DBa/tLGW4yh7yKCn1rIF+9vSTj/HiBeqCS1M3bMrnwOvbl5Ysi+eGLlkBhosjxl1fNwM5Ak7xH6CiT3SdT4U='.decode('base64').decode('zlib')

Unpacks to a severely mistreated A* implementation. Reads stdin. Searches for solutions with minimal total distance. Breaks ties by preferring minimal number of directions. Lists moves to stdout. Finds kittens.

Unpacked:

The source has been manually anti-golfed in a few places in order to obtain a smaller compressed representation. For example, a for loop over the compass directions was unrolled.

import heapq,sys 
a=set() 
for v,p in enumerate(sys.stdin): 
 for u,s in enumerate(p): 
  if s in' KR':a.add((u,v)) 
  if s=='K':(q,r)=(u,v) 
  if s=='R':y=(u,v) 
o=[((abs(y[0]-q)+abs(y[1]-r),(y[0]!=q)+(y[1]!=r)),(0,0),y)] 
c=set() 
w={} 
while o: 
 _,h,x=heapq.heappop(o) 
 c.add(x) 
 s=lambda(u,v):(u,v-1) 
 y=s(x) 
 m=1 
 while y in a-c: 
  w[y]=[(h,x,(m,'N'))]+w.get(y,[]) 
  heapq.heappush(o,((abs(y[0]-q)+abs(y[1]-r)+h[0]+m,(y[0]!=q)+(y[1]!=r)+h[1]+1),(h[0]+m,h[1]+1),y)) 
  m+=1 
  y=s(y) 
 s=lambda(u,v):(u,v+1) 
 y=s(x) 
 m=1 
 while y in a-c: 
  w[y]=[(h,x,(m,'S'))]+w.get(y,[]) 
  heapq.heappush(o,((abs(y[0]-q)+abs(y[1]-r)+h[0]+m,(y[0]!=q)+(y[1]!=r)+h[1]+1),(h[0]+m,h[1]+1),y)) 
  m+=1 
  y=s(y) 
 s=lambda(u,v):(u+1,v) 
 y=s(x) 
 m=1 
 while y in a-c: 
  w[y]=[(h,x,(m,'E'))]+w.get(y,[]) 
  heapq.heappush(o,((abs(y[0]-q)+abs(y[1]-r)+h[0]+m,(y[0]!=q)+(y[1]!=r)+h[1]+1),(h[0]+m,h[1]+1),y)) 
  m+=1 
  y=s(y) 
 s=lambda(u,v):(u-1,v) 
 y=s(x) 
 m=1 
 while y in a-c: 
  w[y]=[(h,x,(m,'W'))]+w.get(y,[]) 
  heapq.heappush(o,((abs(y[0]-q)+abs(y[1]-r)+h[0]+m,(y[0]!=q)+(y[1]!=r)+h[1]+1),(h[0]+m,h[1]+1),y)) 
  m+=1 
  y=s(y) 
 if x==(q,r): 
  z='' 
  while x in w: 
   _,x,(m,d)=min(w[x]) 
   z='%s %d\n'%(d,m)+z 
  print z, 
  o=[]
\$\endgroup\$
3
\$\begingroup\$

c++ -- 681 necessary characters

#include <string>
#include <iostream>
#include <sstream>
using namespace std;
int d[]={-1,0,1,0},i,j,k,l,L=0,p=41,S;string I,m,D("ESWN");
int r(int o,int s){S=s;while((m[o]==32||m[o]==37)&&m[o+S]-35)
if(m[o+S]-p)S+=s;else return o+S;return 0;}
void w(int o){for(i=0;i<m.length();++i)for(j=0;j<4;++j)
if(k=r(o,d[j])){stringstream O;O<<D[j]<<" "<<int((k-o)/d[j])<<"\n"<<I;
I=O.str();if(p-41)--p,m[k]=37,w(k);cout<<I;exit(0);}}
int main(){while(getline(cin,I))m+=I,l=I.length();
d[1]-=d[3]=l;I="";for(i=0;i<m.length();++i)
switch(m[i]){case 82:case 75:m[i]/=2;case 32:break;default:m[i]=35;}
do{for(i=0;i<m.length();++i)if(r(i,-1)+r(i,-l)+r(i,1)+r(i,l))
{if(m[i]==37)w(i);m[i]=p+1;}}while(++p);}

It first replaces all the obstacles on the map with into #s (and changes the values of K and R, to leave headroom in the character space for very long paths. Then it scribbles on the map. An iterative process marks all the successively accessible squares until it is able to reach the kitten in one move. After that is uses the same accessibility checking routine to find a string of positions that lead back to the start in minimal instructions. These instructions are loaded into a string by pre-pending, so that they print in the proper order.

I don't intend to golf it further as it does not properly resolve ties and can not be easily adapted to do so.

It fails on

#####
#   #
# # #
#R#K#
#   #
#####

producing

 $ ./a.out < kitten_4.txt
N 2
E 2
S 2

More or less readable version:

#include <string>
#include <iostream>
#include <sstream>
using namespace std;

int d[]={-1,0,1,0}
  , i, j, k
  , l      /* length of a line on input */
  , L=0    /* length of the whole map */
  , p=41   /* previous count  ( starts 'R'/2 ) */
  , S      /* step accumulator for step function */
  ; 
string I/*nput line, later the instructions*/
  , m/*ap*/
  , D("ESWN"); /* Reversed sence for back tracking the path */

int r/*eachable?*/(int o/*ffset*/, int s/*step*/){
//   cerr << "\tReachable?" 
//        << endl
//        << "\t\tOffset: " << o << " (" << o/5 << ", " << o%5 << ")" 
//        << "  [" << m[o] << "]" << endl
//        << "\t\tStep: " << s 
//        << endl
//        << "\t\tSeeking: " << "[" << char(p) << "]" 
//        << endl
//        << "\t\tWall:    " << "[" << char(35) << "]" 
//        << endl;
  S=s;
  while ( ( (m[o]==32)      /* Current character is a space */ 
        || (m[o]==37) ) /* Current character is a kitten */ 
      && (m[o+S]-35) /* Target character is not a wall */ 
      )
    {
//     cerr << "\t\tTry: " << o+S << "(" << (o+S)/5 << ", " << (o+S)%5 << ")" 
//   << "  [" << m[o+S] << "]" << endl;
    if (m[o+S]-p       /* Target character is not the previous count */ 
    ) {
//       cerr << "\t\twrong " << "  [" << m[o+S] << "] !!!" << endl;
      S+=s;
    }
    else  {             /* Target character *is* the previous count */
//       cerr << "\t\tFound " << "  [" << m[o+S] << "] !!!" << endl;
      return o+S;
    } 
  /* while ends */
      }
  return 0;
}

void w/*on*/(int o/*ffset*/){
//   cerr << "\tWON" << endl
//        << "\t\tOffset: " << o << "(" << o/5 << ", " << o%5 << ")" 
//        << "  [" << m[o] << "]" 
//        << endl
//        << "\t\tSeeking: " << "[" << char(p) << "]" 
//        << endl;
  for(i=0;i<m.length();++i) /* On all map squares */
    for(j=0;j<4;++j)
      if (k=r(o,d[j])) {
//  cerr << "found path segment..." 
//       << (k-o)/d[j] << " in the " << d[j] << " direction." << endl;
    stringstream O;
    O << D[j] 
      << " " 
      << int((k-o)/d[j]) 
      << "\n" 
      << I;
    I = O.str();
//  cerr << I << endl;
    /* test for final solution */
    if (p-41) 
      --p,m[k]=37, w(k); /* recur for further steps */
    cout << I;
    exit(0);
      }
    /* inner for ends */
  /* outer for ends */
}


int main(){
  while(getline(cin,I))
    m+=I,l=I.length();
//   cerr << "Read the map: '" << m << "'." << endl;
  d[1]-=d[3]=l;I="";
//   cerr << "Direction array:    " << D << endl;
//   cerr << "Displacement array: " << d[0] << d[1] << d[2] << d[3] << endl;
//   cerr << "Line length: " << l << endl;
  /* Rewrite the map so that all obstacles are '#' and the start and
     goal are '%' and ')' respectively. Now I can do pathfinding *on*
     the map. */
  for(i=0;i<m.length();++i)
    switch (m[i]) {
    case 82:         /* ASCII 82 == 'R' (41 == ')'  ) */
    case 75:m[i]/=2; /* ASCII 75 == 'K' (37 == '%' ) */
    case ' ':break;
    default: m[i]=35; /* ASCII 35 == '#' */ 
    };
//   cerr << "Re-wrote the map: '" << m << "'." << endl;
  do { /* For each needed count */
//     cerr << "Starting to mark up for step count " 
//   << p-41+1  << " '" << char(p) << "'" << endl;
    for(i=0;i<m.length();++i){ /* On all map squares */
//        cerr << "\tTrying position (" << i/l << ", " << i%l << ")" 
//      << "  [" << m[i] << "]"
//      << endl;
      if ( r(i, -1) /* west  */ +
       r(i, -l) /* north */ +
       r(i,  1) /* east  */ +
       r(i,  l) /* south */ 
       ) {
//    cerr << "Got a hit on : '" << m << "'." << endl;
//    cerr << "\twith '" << char(m[i]) <<" at position " << i << endl;
//    cerr << "target is " << char(37) << endl;
    if(m[i]==37)
      w(i); /* jump into the win routine which never returns */
    m[i]=p+1;
//  cerr << "Marked on map: '" << m << "'." << endl;
      }
    }
  } while(++p);
}
\$\endgroup\$
3
\$\begingroup\$

Ruby - 539 characters

Could do with a lot of improvement, but it does work for shortest steps as well as directions.

M=[*$<]
r=M.map{|q|q.index('R')||0}
k=M.map{|q|q.index('K')||0}
D=M.map{|q|q.split('').map{[99,[]]}} 
def c h 
h.map{|i|i.inject([[]]){|a,b|a.last[0]!=b ? a<<[b, 1]:a.last[1]+=1;a}}.sort_by{|a|a.length}[0]
end
def t x,y,s,i
z,w=D[x][y][0],D[x][y][1]
if [' ','R','K'].index(M[x][y, 1])&&(z>s||z==s&&c(w).length>=c([i]).length)
D[x][y]=[s,z==s ? w<<i:[i]]
s+=1
t x+1,y,s,i+['S']
t x-1,y,s,i+['N']
t x,y+1,s,i+['E']
t x,y-1,s,i+['W']
end
end
t r.index(r.max), r.max, 0, []
puts c(D[k.index(k.max)][k.max][1]).map{|a|a*''}
\$\endgroup\$
1
\$\begingroup\$

Ruby - 648 characters

Another one that fails the least number of directions test as I can't think of any easy way to embed it in A*.

m=$<.read.gsub /[^RK\n ]/,'#'
l=m.index($/)+1
s=m.index'R'
g=m.index'K'
h=->y{(g%l-y%l).abs+(g/l-y/l).abs}
n=->y{[y+1,y-1,y+l,y-l].reject{|i|m[i]=='#'}}
a=o=[s]
d=Array.new(m.length,9e9)
c,k,f=[],[],[]
d[s]=0
f[s]=h[s]
r=->y,u{u<<y;(y=k[y])?redo:u}
(x=o.min_by{|y|f[y]}
x==g ? (a=r[x,[]].reverse;break):0
o-=[x];c<<x
n[x].map{|y|c&[y]!=[]?0:(t=d[x]+1
o&[y]==[]?(o<<y;b=true):b=t<d[y]
b ? (k[y]=x;d[y]=t;f[y]=t+h[y]):0)})until o==[]
k=a.inject([[],nil]){|k,u|(c=k[1]) ? (k[0]<<(c==u-1?'E':c==u+1?'W':c==u+l ? 'N':'S')) : 0;[k[0],u]}[0].inject(["","",0]){|k,v|k[1]==v ? k[2]+=1 : (k[0]+=k[1]+" #{k[2]}\n";k[1]=v;k[2]=1);k}
puts k[0][3,9e9]+k[1]+" #{k[2]}\n"
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.