dlx5.w

@s item int
@s node int
@s mod and
\let\Xmod=\bmod % this is CWEB magic for using "mod" instead of "%"

\datethis
@*Intro. This program is part of a series of ``exact cover solvers'' that
I'm putting together for my own education as I prepare to write Section
7.2.2.1 of {\sl The Art of Computer Programming}. My intent is to
have a variety of compatible programs on which I can run experiments,
in order to learn how different approaches work in practice.

The basic input format for all of these solvers is described at the beginning
of program {\mc DLX1}, and you should read that description now if you are
unfamiliar with it. You should in fact read the beginning of {\mc DLX2}, too,
because it adds ``color controls'' to the repertoire of {\mc DLX1}.

{\mc DLX5} extends {\mc DLX2} by allowing options to have nonnegative
{\it costs}. The goal is to find a minimum-cost solution (or, more generally,
to find the $k$ best solutions, in the sense that the sum of their
costs is minimized).

The input format is extended so that entries such as \.{{\char"7C}$n$}
can be appended to any option, to specify its cost. If several such
entries appear in the same option, the cost is their sum.

Whenever a solution is found whose cost is less than $k$th best seen so far,
that solution is output. For example, suppose the given problem has
only ten solutions, whose costs happen to be (0, 0, 1, 1, 2, 2, 3, 3, 4, 4).
We might discover them in any order, perhaps (3, 1, 4, 1, 2, 3, 2, 4, 0, 0).
If $k=1$ (the default), we'll output solutions of cost 3, 1, 0.
If $k=3$, we'll output solutions of cost 3, 1, 4, 1, 2, 0, 0.
If $k=5$, we'll output solutions of cost 3, 1, 4, 1, 2, 3, 2, 0, 0.
If $k\ge8$, we'll output all ten solutions.
Different values of $k$ might, however, affect the order of discovery.

This program internally assigns a ``tax'' to each item, and changes the cost
of each option to its {\it net cost}, which is the original cost minus
the taxes on each of its items. For example, the net cost of
option `\.a~\.b~\.c~\.{{\char"7C}7}' will be not \$7 but \$1, if
the tax on each of \.a, \.b, and \.c is \$2. This modification doesn't
change the problem in any essential way, because the net cost of each
solution is equal to the original cost of that solution minus the
total tax on all items (and that total tax is constant). Taxes are assessed
in such a way that each item belongs to at least one net-zero-cost option,
yet all options have a nonnegative net cost. The point is that options
whose net cost is large cannot be used in solutions whose net cost is small.

If the input contains no cost specifications, the behavior of {\mc DLX5}
will almost exactly match that of~{\mc DLX2}, except for needing
more time and space.

[{\it Historical note:\/} The simple cutoff rule in this program was
used in one of the first computer codes for min-cost exact cover;
see Garfinkel and Nemhauser, {\sl Operations Research\/ \bf17} (1969),
848--856.]

@ After this program finds its solutions, it normally prints their total
number on |stderr|, together with statistics about how many
nodes were in the search tree, and how many ``updates'' and
``cleansings'' were made.
The running time in ``mems'' is also reported, together with the approximate
number of bytes needed for data storage.
(An ``update'' is the removal of an option from its item list.
A ``cleansing'' is the removal of a satisfied color constraint from its option.
One ``mem'' essentially means a memory access to a 64-bit word.
The reported totals don't include the time or space needed to parse the
input or to format the output.)

Here is the overall structure:

@d o mems++ /* count one mem */
@d oo mems+=2 /* count two mems */
@d ooo mems+=3 /* count three mems */
@d O "%" /* used for percent signs in format strings */
@d mod % /* used for percent signs denoting remainder in \CEE/ */

@d max_level 5000 /* at most this many options in a solution */
@d max_cols 100000 /* at most this many items */
@d max_nodes 10000000 /* at most this many nonzero elements in the matrix */
@d bufsize (9*max_cols+3) /* a buffer big enough to hold all item names */
@d sortbufsize 32 /* for the z lookahead heuristic */

@c
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <ctype.h>
typedef unsigned int uint; /* a convenient abbreviation */
typedef unsigned long long ullng; /* ditto */
@<Type definitions@>;
@<Global variables@>;
@<Subroutines@>;
main (int argc, char *argv[]) {
  register int cc,i,j,k,p,pp,q,r,s,t,cur_node,best_itm;
  register ullng tmpcost,curcost,mincost,nextcost;
  @<Process the command line@>;
  @<Do the input phase@>;
  @<Solve the problem@>;
done:@+if (sanity_checking) sanity();
  @<Bid farewell@>;
}

@ @<Do the input phase@>=
@<Input the item names@>;
@<Input the options@>;
@<Assign taxes@>;
@<Sort the item lists@>;
if (vbose&show_basics)
  @<Report the successful completion of the input phase@>;
if (vbose&show_tots)
  @<Report the item totals@>;
imems=mems, mems=0;

@ @<Bid farewell@>=
if (vbose&show_tots)
  @<Report the item totals@>;
if (vbose&show_profile) @<Print the profile@>;
if (vbose&show_basics) {
  fprintf(stderr,"Altogether "O"llu solution"O"s, "O"llu+"O"llu mems,",
                              count,count==1?"":"s",imems,mems);
  bytes=last_itm*sizeof(item)+last_node*sizeof(node)+maxl*sizeof(int);
  fprintf(stderr," "O"llu updates, "O"llu cleansings,",
                              updates,cleansings);
  fprintf(stderr," "O"llu bytes, "O"llu nodes.\n",
                              bytes,nodes);
}
if ((vbose&show_opt_costs) && count) @<Print the |kthresh| best costs found@>;
@<Close the files@>;

@ You can control the amount of output, as well as certain properties
of the algorithm, by specifying options on the command line:
\smallskip\item{$\bullet$}
`\.v$\langle\,$integer$\,\rangle$' enables or disables various kinds of verbose
 output on |stderr|, given by binary codes such as |show_choices|;
\item{$\bullet$}
`\.m$\langle\,$integer$\,\rangle$' causes every $m$th solution
to be output (the default is \.{m0}, which merely counts them);
\item{$\bullet$}
`\.k$\langle\,$positive integer$\,\rangle$' causes the algorithm to cut off
solutions that don't improve costwise on the $k$ best seen so far
(the default is 1, and $k$ must not exceed |maxk|);
\item{$\bullet$}
`\.Z$\langle\,$string$\,\rangle$' causes a warning to be printed
if there's an option that doesn't have exactly one primary item
beginning with~$c$, for each character~$c$ of the string
(thereby allowing a special heuristic to be used for cutting off false starts);
\item{$\bullet$}
`\.z$\langle\,$positive integer$\,\rangle$' causes a warning to be printed
if there's an option that doesn't have exactly this many primary items
in addition to those specified by \.Z
(thereby allowing a special heuristic to be used for cutting off false starts);
\item{$\bullet$}
`\.h$\langle\,$positive integer$\,\rangle$' sets |lenthresh|, a heuristic that
limits the amount of lookahead when we're trying to identify the best item for
branching (default 10);
\item{$\bullet$}
`\.d$\langle\,$integer$\,\rangle$' sets |delta|, which causes periodic
state reports on |stderr| after the algorithm has performed approximately
|delta| mems since the previous report (default 10000000000);
\item{$\bullet$}
`\.c$\langle\,$positive integer$\,\rangle$' limits the levels on which
choices are shown during verbose tracing;
\item{$\bullet$}
`\.C$\langle\,$positive integer$\,\rangle$' limits the levels on which
choices are shown in the periodic state reports;
\item{$\bullet$}
`\.l$\langle\,$nonnegative integer$\,\rangle$' gives a {\it lower\/} limit,
relative to the maximum level so far achieved, to the levels on which
choices are shown during verbose tracing;
\item{$\bullet$}
`\.t$\langle\,$positive integer$\,\rangle$' causes the program to
stop after this many solutions have been found;
\item{$\bullet$}
`\.T$\langle\,$integer$\,\rangle$' sets |timeout| (which causes abrupt
termination if |mems>timeout| at the beginning of a level);
\item{$\bullet$}
`\.S$\langle\,$filename$\,\rangle$' to output a ``shape file'' that encodes
the search tree.

@d show_basics 1 /* |vbose| code for basic stats; this is the default */
@d show_choices 2 /* |vbose| code for backtrack logging */
@d show_details 4 /* |vbose| code for further commentary */
@d show_taxes 8 /* |vbose| code to print all nonzero item taxes */
@d show_opt_costs 16 /* |vbose| code to show the best $k$ costs at end */
@d show_profile 128 /* |vbose| code to show the search tree profile */
@d show_full_state 256 /* |vbose| code for complete state reports */
@d show_tots 512 /* |vbose| code for reporting item totals at start and end */
@d show_warnings 1024 /* |vbose| code for reporting options without primaries */
@d maxk 15000 /* upper limit on parameter \.k */

@ @<Glob...@>=
int vbose=show_basics+show_opt_costs+show_warnings; /* level of verbosity */
int spacing; /* solution $t$ is output if $t$ is a multiple of |spacing| */
int show_choices_max=1000000; /* above this level, |show_choices| is ignored */
int show_choices_gap=1000000; /* below level |maxl-show_choices_gap|,
    |show_details| is ignored */
int show_levels_max=1000000; /* above this level, state reports stop */
int maxl=0; /* maximum level actually reached */
char buf[bufsize]; /* input buffer */
ullng sortbuf[sortbufsize]; /* short buffer for sorting */
ullng count; /* solutions found so far */
ullng options; /* options seen so far */
ullng imems,mems; /* mem counts */
ullng updates; /* update counts */
ullng cleansings; /* cleansing counts */
ullng bytes; /* memory used by main data structures */
ullng nodes; /* total number of branch nodes initiated */
ullng thresh=10000000000; /* report when |mems| exceeds this, if |delta!=0| */
ullng delta=10000000000; /* report every |delta| or so mems */
ullng maxcount=0xffffffffffffffff; /* stop after finding this many solutions */
ullng timeout=0x1fffffffffffffff; /* give up after this many mems */
FILE *shape_file; /* file for optional output of search tree shape */
char *shape_name; /* its name */
int kthresh=1; /* this many mincost solutions will be found, if possible */
int lenthresh=10; /* at most this many options checked per item */
int zgiven; /* this many primary items per option, if specified */
char Zchars[8]; /* prefix characters specified by parameter \.Z */
int ppgiven; /* desired footprint of primary items in every option */

@ If an option appears more than once on the command line, the first
appearance takes precedence.

@<Process the command line@>=
for (j=argc-1,k=0;j;j--) switch (argv[j][0]) {
case 'v': k|=(sscanf(argv[j]+1,""O"d",&vbose)-1);@+break;
case 'm': k|=(sscanf(argv[j]+1,""O"d",&spacing)-1);@+break;
case 'k': k|=(sscanf(argv[j]+1,""O"d",&kthresh)-1);
  if (kthresh<1 || kthresh>maxk) {
    fprintf(stderr,"Sorry, parameter k must be between 1 and "O"d!\n",
                               maxk);
    exit(-1);
  }
  break;
case 'Z': if (strlen(argv[j])>8) {
    fprintf(stderr,
       "Sorry, parameter Z must specify at most 7 prefix characters!\n");
    k|=1;
  }@+else sprintf(Zchars,"%s",argv[j]+1);
  break;
case 'z': k|=(sscanf(argv[j]+1,""O"d",&zgiven)-1);@+break;
case 'h': k|=(sscanf(argv[j]+1,""O"d",&lenthresh)-1);@+break;
case 'd': k|=(sscanf(argv[j]+1,""O"lld",&delta)-1),thresh=delta;@+break;
case 'c': k|=(sscanf(argv[j]+1,""O"d",&show_choices_max)-1);@+break;
case 'C': k|=(sscanf(argv[j]+1,""O"d",&show_levels_max)-1);@+break;
case 'l': k|=(sscanf(argv[j]+1,""O"d",&show_choices_gap)-1);@+break;
case 't': k|=(sscanf(argv[j]+1,""O"lld",&maxcount)-1);@+break;
case 'T': k|=(sscanf(argv[j]+1,""O"lld",&timeout)-1);@+break;
case 'S': shape_name=argv[j]+1, shape_file=fopen(shape_name,"w");
  if (!shape_file)
    fprintf(stderr,"Sorry, I can't open file `"O"s' for writing!\n",
      shape_name);
  break;
default: k=1; /* unrecognized command-line option */
}
if (k) {
  fprintf(stderr, "Usage: "O"s [v<n>] [m<n>] [k<n>] [Z<ABC>] [z<n>] [h<n>]"
       " [d<n>] [c<n>] [C<n>] [l<n>] [t<n>] [T<n>] [S<bar>] < foo.dlx\n",
                            argv[0]);
  exit(-1);
}

@ @<Close the files@>=
if (shape_file) fclose(shape_file);

@ @<Print the |kthresh| best costs found@>=
{
  fprintf(stderr,"The optimum cost"O"s",kthresh==1?" is":"s are:\n");
  @<Sort the |bestcost| heap in preparation for final printing@>;
  for (k=1,tmpcost=infcost;k<=kthresh && bestcost[k]<infcost;k++) {
    if (tmpcost==totaltax+bestcost[k]) r++;
    else {
      @<Print a line (except the first time)@>;
      tmpcost=totaltax+bestcost[k],r=0;
    }
  }
  @<Print a line (except the first time)@>;
}

@ @<Print a line (except the first time)@>=
if (tmpcost!=infcost) {
  if (r) fprintf(stderr," $"O"llu (repeated "O"d times)\n"@q$@>,tmpcost,r+1);
  else fprintf(stderr," $"O"llu\n"@q$@>,tmpcost);
}

@*Data structures.
Each item of the input matrix is represented by an \&{item} struct,
and each option is represented as a list of \&{node} structs. There's one
node for each nonzero entry in the matrix.

More precisely, the nodes of individual options appear sequentially,
with ``spacer'' nodes between them. The nodes are also
linked circularly with respect to each item, in doubly linked lists.
The item lists each include a header node, but the option lists do not.
Item header nodes are aligned with an \&{item} struct, which
contains further info about the item.

Each node contains five important fields, and one other that's unused but
might be important in extensions of this program. Two are the pointers |up|
and |down| of doubly linked lists, already mentioned.
A~third points directly to the item containing the node.
A~fourth specifies a color, or zero if no color is specified.
A~fifth specifies the cost of the option in which this node occurs.
A~sixth points to the spacer at the end of the option;
that one is currently set, but not looked at.

A ``pointer'' is an array index, not a \CEE/ reference (because the latter
would occupy 64~bits and waste cache space). The |cl| array is for
\&{item} structs, and the |nd| array is for \&{node}s. I assume that both of
those arrays are small enough to be allocated statically. (Modifications
of this program could do dynamic allocation if needed.)
The header node corresponding to |cl[c]| is |nd[c]|.

Notice that each \&{node} occupies three octabytes.
We count one mem for a simultaneous access to the |up| and |down| fields,
or for a simultaneous access to the |itm| and |color| fields.

Although the item-list pointers are called |up| and |down|, they need not
correspond to actual positions of matrix entries. The elements of
each item list can appear in any order, so that one option
needn't be consistently ``above'' or ``below'' another. Indeed, we
will sort each option list of a primary item from top to bottom in
order of nondecreasing cost.

This program doesn't change the |itm| fields after they've first been set up.
But the |up| and |down| fields will be changed frequently, although preserving
relative order.

Exception: In the node |nd[c]| that is the header for the list of
item~|c|, we use the |cost| field to hold the ``tax'' on that item---for
diagnostic purposes only, not as part of the algorithm's decision-making.
We also might use its |color| field for special purposes.
The alternative names |len| for |itm|, |aux| for |color|, and |tax| for |cost|
are used in the code so that this nonstandard semantics will be more clear.

A {\it spacer\/} node has |itm<=0|. Its |up| field points to the start
of the preceding option; its |down| field points to the end of the following option.
Thus it's easy to traverse an option circularly, in either direction.

The |color| field of a node is set to |-1| when that node has been cleansed.
In such cases its original color appears in the item header.
(The program uses this fact only for diagnostic outputs.)

@d len itm /* item list length (used in header nodes only) */
@d aux color /* an auxiliary quantity (used in header nodes only) */
@d tax cost /* item tax (used in header nodes only) */

@<Type...@>=
typedef struct node_struct {
  int up,down; /* predecessor and successor in item list */
  int itm; /* the item containing this node */
  int color; /* the color specified by this node, if any */
  ullng cost; /* the cost of the option containing this node */
} node;

@ Each \&{item} struct contains three fields:
The |name| is the user-specified identifier;
|next| and |prev| point to adjacent items, when this
item is part of a doubly linked list.

As backtracking proceeds, nodes
will be deleted from item lists when their option has been hidden by
other options in the partial solution.
But when backtracking is complete, the data structures will be
restored to their original state.

We count one mem for a simultaneous access to the |prev| and |next| fields.

@<Type...@>=
typedef struct itm_struct {
  char name[8]; /* symbolic identification of the item, for printing */
  int prev,next; /* neighbors of this item */
} item;

@ @<Glob...@>=
node* nd; /* the master list of nodes */
int last_node; /* the first node in |nd| that's not yet used */
item cl[max_cols+2]; /* the master list of items */
int second=max_cols; /* boundary between primary and secondary items */
int last_itm; /* the first item in |cl| that's not yet used */
ullng totaltax; /* the sum of all taxes assessed */

@ One |item| struct is called the root. It serves as the head of the
list of items that need to be covered, and is identifiable by the fact
that its |name| is empty.

@d root 0 /* |cl[root]| is the gateway to the unsettled items */

@ An option is identified not by name but by the names of the items it contains.
Here is a routine that prints an option, given a pointer to any of its
nodes. It also prints the position of the option in its item list,
given a cost threshold to measure the length of that list.

@<Sub...@>=
void print_option(int p,FILE *stream,ullng thresh) {
  register int c,j,k,q;
  register ullng s;
  c=nd[p].itm;
  if (p<last_itm || p>=last_node || c<=0) {
    fprintf(stderr,"Illegal option "O"d!\n",p);
    return;
  }
  for (q=p,s=0;;) {
    fprintf(stream," "O".8s",cl[nd[q].itm].name);
    if (nd[q].color)
      fprintf(stream,":"O"c",nd[q].color>0? nd[q].color: nd[nd[q].itm].color);
    s+=nd[nd[q].itm].tax;
    q++;
    if (nd[q].itm<=0) q=nd[q].up; /* |-nd[q].itm| is actually the option number */
    if (q==p) break;
  }
  for (q=nd[c].down,k=1;q!=p;k++) {
    if (q==c) {
      fprintf(stream," (?)");@+goto finish; /* option not in its item list! */
    }@+else q=nd[q].down;
  }
  for (q=nd[c].down,j=0;q>=last_itm;q=nd[q].down,j++)
    if (nd[q].cost>=thresh) break;
  fprintf(stream," ("O"d of "O"d)",k,j);
finish:@+if (s+nd[p].cost) fprintf(stream," $"O"llu ["O"llu]\n"@q$@>,
                     s+nd[p].cost,nd[p].cost);
  else fprintf(stream,"\n");
}
@#
void prow(int p) {
  print_option(p,stderr,infcost);
}

@ When I'm debugging, I might want to look at one of the current item lists.

@<Sub...@>=
void print_itm(int c) {
  register int p;
  if (c<root || c>=last_itm) {
    fprintf(stderr,"Illegal item "O"d!\n",c);
    return;
  }
  if (c<second)
    fprintf(stderr,"Item "O".8s, neighbors "O".8s and "O".8s:\n",
        cl[c].name,cl[cl[c].prev].name,cl[cl[c].next].name);
  else fprintf(stderr,"Item "O".8s:\n",cl[c].name);
  for (p=nd[c].down;p>=last_itm;p=nd[p].down) prow(p);
}

@ Speaking of debugging, here's a routine to check if redundant parts of our
data structure have gone awry.

@d sanity_checking 0 /* set this to 1 if you suspect a bug */

@<Sub...@>=
void sanity(void) {
  register int k,p,q,pp,qq,t;
  for (q=root,p=cl[q].next;;q=p,p=cl[p].next) {
    if (cl[p].prev!=q)
      fprintf(stderr,"Bad prev field at itm "O".8s!\n",cl[p].name);
    if (p==root) break;
    @<Check item |p|@>;
  }
}

@ @<Check item |p|@>=
for (qq=p,pp=nd[qq].down,k=0;;qq=pp,pp=nd[pp].down,k++) {
  if (nd[pp].up!=qq)
    fprintf(stderr,"Bad up field at node "O"d!\n",pp);
  if (pp==p) break;
  if (nd[pp].itm!=p)
    fprintf(stderr,"Bad itm field at node "O"d!\n",pp);
  if (qq>p && nd[pp].cost<nd[qq].cost)
    fprintf(stderr,"Costs out of order at node "O"d!\n",pp);
}
if (p<second && nd[p].len!=k) fprintf(stderr,"Bad len field in item "O".8s!\n",
                                                       cl[p].name);

@*Inputting the matrix. Brute force is the rule in this part of the code,
whose goal is to parse and store the input data and to check its validity.

@d panic(m) {@+fprintf(stderr,""O"s!\n"O"d: "O".99s\n",m,p,buf);@+exit(-666);@+}

@<Input the item names@>=
nd=(node*)calloc(max_nodes,sizeof(node));
if (!nd) {
  fprintf(stderr,"I couldn't allocate space for "O"d nodes!\n",max_nodes);
  exit(-666);
}
if (max_nodes<=2*max_cols) {
  fprintf(stderr,"Recompile me: max_nodes must exceed twice max_cols!\n");
  exit(-999);
} /* every item will want a header node and at least one other node */
while (1) {
  if (!fgets(buf,bufsize,stdin)) break;
  if (o,buf[p=strlen(buf)-1]!='\n') panic("Input line way too long");
  for (p=0;o,isspace(buf[p]);p++) ;
  if (buf[p]=='|' || !buf[p]) continue; /* bypass comment or blank line */
  last_itm=1;
  break;
}
if (!last_itm) panic("No items");
for (;o,buf[p];) {
  for (j=0;j<8 && (o,!isspace(buf[p+j]));j++) {
    if (buf[p+j]==':' || buf[p+j]=='|')
              panic("Illegal character in item name");
    o,cl[last_itm].name[j]=buf[p+j];
  }
  if (j==8 && !isspace(buf[p+j])) panic("Item name too long");
  @<Check for duplicate item name@>;
  @<Initialize |last_itm| to a new item with an empty list@>;
  for (p+=j+1;o,isspace(buf[p]);p++) ;
  if (buf[p]=='|') {
    if (second!=max_cols) panic("Item name line contains | twice");
    second=last_itm;
    for (p++;o,isspace(buf[p]);p++) ;
  }
}
if (second==max_cols) second=last_itm;
oo,cl[last_itm].prev=last_itm-1, cl[last_itm-1].next=last_itm;
oo,cl[second].prev=last_itm,cl[last_itm].next=second;
  /* this sequence works properly whether or not |second=last_itm| */
oo,cl[root].prev=second-1, cl[second-1].next=root;
last_node=last_itm; /* reserve all the header nodes and the first spacer */
/* we have |nd[last_node].itm=0| in the first spacer */

@ @<Check for duplicate item name@>=
for (k=1;o,strncmp(cl[k].name,cl[last_itm].name,8);k++) ;
if (k<last_itm) panic("Duplicate item name");

@ @<Initialize |last_itm| to a new item with an empty list@>=
if (last_itm>max_cols) panic("Too many items");
 oo,cl[last_itm-1].next=last_itm,cl[last_itm].prev=last_itm-1;
 /* |nd[last_itm].len=0| */
o,nd[last_itm].up=nd[last_itm].down=last_itm;
last_itm++;

@ I'm putting the option number into the spacer that follows it, as a
possible debugging aid. But the program doesn't currently use that information.

@<Input the options@>=
@<Set |ppgiven| from parameters \.Z and \.z@>;
while (1) {
  if (!fgets(buf,bufsize,stdin)) break;
  if (o,buf[p=strlen(buf)-1]!='\n') panic("Option line too long");
  for (p=0;o,isspace(buf[p]);p++) ;
  if (buf[p]=='|' || !buf[p]) continue; /* bypass comment or blank line */
  i=last_node; /* remember the spacer at the left of this option */
  tmpcost=0;
  for (pp=0;buf[p];) {
    for (j=0;j<8 && (o,!isspace(buf[p+j])) && buf[p+j]!=':';j++)
      o,cl[last_itm].name[j]=buf[p+j];
    if (!j) panic("Empty item name");
    if (j==8 && !isspace(buf[p+j]) && buf[p+j]!=':')
          panic("Item name too long");
    if (j<8) o,cl[last_itm].name[j]='\0';
    @<Create a node for the item named in |buf[p]|@>;
    if (buf[p+j]!=':') o,nd[last_node].color=0;
    else if (k>=second) {
      if ((o,isspace(buf[p+j+1])) || (o,!isspace(buf[p+j+2])))
        panic("Color must be a single character");
      o,nd[last_node].color=buf[p+j+1];
      p+=2;
    }@+else panic("Primary item must be uncolored");
    @<Skip to next item, accruing cost information if any@>;
  }
  if (!pp) {
    if (vbose&show_warnings)
      fprintf(stderr,"Option ignored (no primary items): "O"s",buf);
    while (last_node>i) {
      @<Remove |last_node| from its item list@>;
      last_node--;
    }
  }@+else {
    @<Check for consistency with parameters \.Z and \.z@>;
    @<Insert the cost into each item of this option@>;
    o,nd[i].down=last_node;
    last_node++; /* create the next spacer */
    if (last_node==max_nodes) panic("Too many nodes");
    options++;
    o,nd[last_node].up=i+1;
    o,nd[last_node].itm=-options;
  }
}

@ @<Skip to next item, accruing cost information if any@>=
while (1) {
  register ullng d;
  for (p+=j+1;o,isspace(buf[p]);p++) ;
  if (buf[p]!='|') break;
  if (buf[p+1]<'0' || buf[p+1]>'9')
    panic("Option cost should be a decimal number");
  for (j=1,d=0;o,!isspace(buf[p+j]);j++) {
    if (buf[p+j]<'0' || buf[p+j]>'9')
      panic("Illegal digit in option cost");
    d=10*d+buf[p+j]-'0';
  }
  tmpcost+=d;
}

@ @<Insert the cost into each item of this option@>=
for (j=i+1;j<=last_node;j++)
  o,nd[j].cost=tmpcost;

@ @<Create a node for the item named in |buf[p]|@>=
for (k=0;o,strncmp(cl[k].name,cl[last_itm].name,8);k++) ;
if (k==last_itm) panic("Unknown item name");
if (o,nd[k].aux>=i) panic("Duplicate item name in this option");
last_node++;
if (last_node==max_nodes) panic("Too many nodes");
o,nd[last_node].itm=k;
if (k<second) @<Adjust |pp| for parameters \.Z and \.z@>;
o,t=nd[k].len+1;
@<Insert node |last_node| into the list for item |k|@>;

@ Insertion of a new node is simple. Before taxes have been computed,
we set only the |up| links of each item list.

We store the position of the new node into |nd[k].aux|, so that
the test for duplicate items above will be correct.

@<Insert node |last_node| into the list for item |k|@>=
o,nd[k].len=t; /* store the new length of the list */
nd[k].aux=last_node; /* no mem charge for |aux| after |len| */
o,r=nd[k].up; /* the ``bottom'' node of the item list */
oo,nd[k].up=last_node,nd[last_node].up=r;

@ @<Remove |last_node| from its item list@>=
o,k=nd[last_node].itm;
oo,nd[k].len--,nd[k].aux=i-1;
oo,nd[k].up=nd[last_node].up;

@ When the user has used the \.Z parameter to specify special prefix
characters, we want to check that each option conforms to that specification.

The rightmost bits of variable |pp| will indicate which prefixes have been seen
so far. The other bits of |pp| will count active items that don't have
a \.Z-specified prefix.

@<Set |ppgiven| from parameters \.Z and \.z@>=
if (o,Zchars[0]) {
  for (r=1;Zchars[r];r++) ;
  ppgiven=(1<<r)-1+(zgiven<<8);
}@+else ppgiven=zgiven<<8;

@ @<Check for consistency with parameters \.Z and \.z@>=
if (ppgiven) {
  if (zgiven && ((pp>>8)!=zgiven))
    fprintf(stderr,"Option has "O"d non-Z primary items, not "O"d: "O"s",
        pp>>8,zgiven,buf);
  if ((pp^ppgiven)&0xff) {
    for (r=0;Zchars[r];r++) if ((pp&(1<<r))==0)
      fprintf(stderr,"Option lacks a "O"c item: "O"s",Zchars[r],buf);
  }
}    

@ @<Adjust |pp| for parameters \.Z and \.z@>=
{
  for (r=0;Zchars[r];r++)
    if (Zchars[r]==cl[last_itm].name[0]) break;
  if (Zchars[r]) {
    if (pp&(1<<r))
      fprintf(stderr,"Option has two "O"c items: "O"s",Zchars[r],buf);
    else pp+=1<<r;
  }@+else pp+=1<<8;
}

@ We look at the option list for every primary item, in turn, to find
an option with smallest cost. If that cost |minc| is positive, we ``tax'' the
item by |minc|, and subtract |minc| from the cost of all options that contain
this item.

If an item has no options, its tax is infinite. (But nobody ever gets
to collect it.)

@d infcost ((ullng)-1) /* ``infinite'' cost */

@<Assign taxes@>=
for (k=1;k<second;k++) {
  register ullng minc;
  for (p=nd[k].up,minc=infcost;p>k && minc;o,p=nd[p].up)
    if (o,nd[p].cost<minc) minc=nd[p].cost;
  if (minc) {
    if (vbose&show_taxes)
      fprintf(stderr," "O".8s tax=$"O"llu\n"@q$@>,cl[k].name,minc);
    totaltax+=minc;
    for (p=nd[k].up;p>k;o,p=nd[p].up) {
      for (q=p+1;;) {
        o,cc=nd[q].itm;
        if (cc<=0) o,q=nd[q].up;
        else {
          oo,nd[q].cost-=minc;
          if (q==p) break;
          q++;
        }
      }
    }
    nd[k].tax=minc; /* for documentation only, so no mem charged */
  }
}
if (totaltax && (vbose&show_taxes))
  fprintf(stderr," (total tax is $"O"llu)\n"@q$@>,totaltax);

@ We use the ``natural list merge sort,'' namely Algorithm 5.2.4L as
modified by exercise 5.2.4--12.

@<Sort the item lists@>=
for (k=1;k<last_itm;k++) {
l1: o,p=nd[k].up,q=nd[p].up;
  for (o,t=root; q>k; o,p=q,q=nd[p].up) /* one mem charged for |nd[p].cost| */
    if (o,nd[p].cost<nd[q].cost) nd[t].up=-q, t=p;
  if (t!=root) @<Sort item list |k|@>;
  @<Make the |down| links consistent with the |up| links@>;
}

@ @<Make the |down| links consistent with the |up| links@>=
for (o,p=k,q=nd[p].up; q>k; o,p=q,q=nd[p].up) o,nd[q].down=p;
oo,nd[p].up=k,nd[k].down=p;

@ The item list is now divided into sorted sublists, separated by links
that have temporarily been negated.

The sorted sublists are merged, two by two. List |t| is ``above'' list~|s|;
hence the sorting is stable with respect to nodes of equal cost.

@<Sort item list |k|@>=
{
  oo,nd[t].up=nd[p].up=0; /* terminate the last two sublists with a null link */
l2:@+while (o,nd[root].up) { /* begin new pass */
    oo,s=k,t=root,p=nd[s].up,q=-nd[root].up; /* mem charged for |nd[p].cost| */
l3:@+if (o,nd[p].cost<nd[q].cost) goto l6;
l4:@+@<Advance |p|@>;
l6:@+@<Advance |q|@>;
l8: p=-p,q=-q;
  if (q) goto l3;
  oo,nd[s].up=-p,nd[t].up=0; /* end of pass */
  }
}

@ @<Advance |p|@>=
o,nd[s].up=(nd[s].up<=0? -p: p);
o,s=p,p=nd[p].up;
if (p>0) goto l3;
l5: o,nd[s].up=q,s=t;
for (;q>0;o,q=nd[q].up) t=q; /* move |q| to the end of its sublist */
goto l8; /* both sublists have now been merged */

@ @<Advance |q|@>=
o,nd[s].up=(nd[s].up<=0? -q: q);
o,s=q,q=nd[q].up;
if (q>0) goto l3;
l7: o,nd[s].up=p,s=t;
for (;p>0;o,p=nd[p].up) t=p; /* move |p| to the end of its sublist */
goto l8; /* both sublists have now been merged */

@ @<Report the successful completion of the input phase@>=
fprintf(stderr,
  "("O"lld options, "O"d+"O"d items, "O"d entries successfully read)\n",
                       options,second-1,last_itm-second,last_node-last_itm);

@ The item lengths after input should agree with the item lengths
after this program has finished. I print them (on request), in order to
provide some reassurance that the algorithm isn't badly screwed up.

@<Report the item totals@>=
{
  fprintf(stderr,"Item totals:");
  for (k=1;k<last_itm;k++) {
    if (k==second) fprintf(stderr," |");
    fprintf(stderr," "O"d",nd[k].len);
  }
  fprintf(stderr,"\n");
}

@*The dancing.
Our strategy for generating all exact covers will be to repeatedly
choose always the item that appears to be hardest to cover, namely the
item with shortest list, from all items that still need to be covered.
And we explore all possibilities via depth-first search.

The neat part of this algorithm is the way the lists are maintained.
Depth-first search means last-in-first-out maintenance of data structures;
and it turns out that we need no auxiliary tables to undelete elements from
lists when backing up. The nodes removed from doubly linked lists remember
their former neighbors, because we do no garbage collection.

The basic operation is ``covering an item.'' This means removing it
from the list of items needing to be covered, and ``hiding'' its
options: removing nodes from other lists whenever they belong to an option of
a node in this item's list.

@<Solve the problem@>=
@<Initialize for level 0@>;
forward: nodes++;
if (vbose&show_profile) profile[level]++;
if (sanity_checking) sanity();
@<Do special things if enough |mems| have accumulated@>;
@<If the remaining cost is clearly too high, |goto backdown|@>;
@<Set |best_itm| to the best item for branching, or |goto backdown|@>;
o,partcost[level]=curcost;
oo,cur_node=choice[level]=nd[best_itm].down;
o,nextcost=curcost+nd[cur_node].cost;
o,coverthresh0[level]=cutoffcost-nextcost; /* known to be positive */
cover(best_itm,coverthresh0[level]);
advance:@+if ((vbose&show_choices) && level<show_choices_max) {
  fprintf(stderr,"L"O"d:",level);
  print_option(cur_node,stderr,cutoffcost-curcost);
}
@<Cover all other items of |cur_node|@>;
if (o,cl[root].next==root) @<Visit a solution and |goto recover|@>;
if (++level>maxl) {
  if (level>=max_level) {
    fprintf(stderr,"Too many levels!\n");
    exit(-4);
  }
  maxl=level;
}
curcost=nextcost;
goto forward;
backup: o,uncover(best_itm,coverthresh0[level]);
backdown:@+if (level==0) goto done;
level--;
oo,cur_node=choice[level],best_itm=nd[cur_node].itm;
o,curcost=partcost[level];
recover: @<Uncover all other items of |cur_node|@>;
oo,cur_node=choice[level]=nd[cur_node].down;
if (cur_node==best_itm) goto backup;
o,nextcost=curcost+nd[cur_node].cost;
if (nextcost>=cutoffcost) goto backup;
goto advance;

@ @<Initialize for level 0@>=
if (zgiven) {
  for (r=0;Zchars[r];r++) ;
  if ((second-1) mod (zgiven+r)) {
    fprintf(stderr,
      "There are "O"d primary items, but z="O"d and Z="O"s!\n",
         second-1,zgiven,Zchars);
    goto done;
  }
}
level=0;
for (k=0;k<kthresh;k++) o,bestcost[k]=infcost;
cutoffcost=infcost;
curcost=0;

@ @<Glob...@>=
int level; /* number of choices in current partial solution */
int choice[max_level]; /* the node chosen on each level */
ullng profile[max_level]; /* number of search tree nodes on each level */
ullng partcost[max_level]; /* the net cost so far, on each level */
ullng coverthresh0[max_level],coverthresh[max_level]; /* historic thresholds */
ullng bestcost[maxk+1]; /* the best |kthresh| net costs known so far */
ullng cutoffcost; /* |bestcost[0]|, the cost we need to beat */
ullng cumcost[7]; /* accumulated costs for the \.Z prefix characters */
int solutionsize; /* the number of options per solution, if fixed and known */

@ @<Do special things if enough |mems| have accumulated@>=
if (delta && (mems>=thresh)) {
  thresh+=delta;
  if (vbose&show_full_state) print_state();
  else print_progress();
}
if (mems>=timeout) {
  fprintf(stderr,"TIMEOUT!\n");@+goto done;
}

@ When an option is hidden, it leaves all lists except the list of the
item that is being covered. Thus a node is never removed from a list
twice.

We can save time by not removing nodes from secondary items that have been
purified. (Such nodes have |color<0|. Note that |color| and |itm| are
stored in the same octabyte; hence we pay only one mem to look at
them both.)

We save even more time by not updating the |len| fields of secondary items.

It's not necessary to hide all the options of the list being covered.
Only the options whose cost is below a given threshold will ever
be relevant, since we seek only minimum-cost solutions.

@<Sub...@>=
void cover(int c,ullng thresh) {
  register int cc,l,r,rr,nn,uu,dd,t;
  o,l=cl[c].prev,r=cl[c].next;
  oo,cl[l].next=r,cl[r].prev=l;
  updates++;
  for (o,rr=nd[c].down;rr>=last_itm;o,rr=nd[rr].down) {
    if (o,nd[rr].cost>=thresh) break;
    for (nn=rr+1;nn!=rr;) {
      if (o,nd[nn].color>=0) {
        o,uu=nd[nn].up,dd=nd[nn].down;
        cc=nd[nn].itm;
        if (cc<=0) {
          nn=uu;@+continue;
        }
        oo,nd[uu].down=dd,nd[dd].up=uu;
        updates++;
        if (cc<second) oo,nd[cc].len--;
      }
      nn++;
    }
  }
}

@ I used to think that it was important to uncover an item by
processing its options from bottom to top, since covering was done
from top to bottom. But while writing this
program I realized that, amazingly, no harm is done if the
options are processed again in the same order.
It's easier to go down than up, because of the cutoff threshold;
hence that observation is good news.
Whether we go up or down, the pointers
execute an exquisitely choreo\-graphed dance that returns them almost
magically to their former state.

Of course we must be careful to use exactly the same thresholds
when uncovering as we did when covering, even though the
|cutoffcost| in this program is a moving target.

@<Subroutines@>=
void uncover(int c,ullng thresh) {
  register int cc,l,r,rr,nn,uu,dd,t;
  for (o,rr=nd[c].down;rr>=last_itm;o,rr=nd[rr].down) {
    if (o,nd[rr].cost>=thresh) break;
    for (nn=rr+1;nn!=rr;) {
      if (o,nd[nn].color>=0) {
        o,uu=nd[nn].up,dd=nd[nn].down;
        cc=nd[nn].itm;
        if (cc<=0) {
          nn=uu;@+continue;
        }
        oo,nd[uu].down=nd[dd].up=nn;
        if (cc<second) oo,nd[cc].len++;
      }
      nn++;
    }
  }
  o,l=cl[c].prev,r=cl[c].next;
  oo,cl[l].next=cl[r].prev=c;
}

@ @<Cover all other items of |cur_node|@>=
o,coverthresh[level]=cutoffcost-nextcost;
for (pp=cur_node+1;pp!=cur_node;) {
  o,cc=nd[pp].itm;
  if (cc<=0) o,pp=nd[pp].up;
  else {
    if (!nd[pp].color) cover(cc,coverthresh[level]);
    else if (nd[pp].color>0) purify(pp,coverthresh[level]);
    pp++;
  }
}

@ We must go leftward as we uncover the items, because we went
rightward when covering them.

@<Uncover all other items of |cur_node|@>=
o; /* charge one mem for putting |coverthresh[level]| in a register */
for (pp=cur_node-1;pp!=cur_node;) {
  o,cc=nd[pp].itm;
  if (cc<=0) o,pp=nd[pp].down;
  else {
    if (!nd[pp].color) uncover(cc,coverthresh[level]);
    else if (nd[pp].color>0) unpurify(pp,coverthresh[level]);
    pp--;
  }
}

@ When we choose an option that specifies colors in one or more items,
we ``purify'' those items by removing all incompatible options.
All options that want the chosen color in a purified item are temporarily
given the color code~|-1| so that they won't be purified again.

@<Sub...@>=
void purify(int p,ullng thresh) {
  register int cc,rr,nn,uu,dd,t,x;
  o,cc=nd[p].itm,x=nd[p].color;
  nd[cc].color=x; /* no mem charged, because this is for |print_option| only */
  cleansings++;
  for (o,rr=nd[cc].down;rr>=last_itm;o,rr=nd[rr].down) {
    if (o,nd[rr].cost>=thresh) break;
    if (o,nd[rr].color!=x) {
      for (nn=rr+1;nn!=rr;) {
        if (o,nd[nn].color>=0) {
          o,uu=nd[nn].up,dd=nd[nn].down;
          cc=nd[nn].itm;
          if (cc<=0) {
            nn=uu;@+continue;
          }
          oo,nd[uu].down=dd,nd[dd].up=uu;
          updates++;
          if (cc<second) oo,nd[cc].len--;
        }
        nn++;
      }
    }@+else if (rr!=p) cleansings++,o,nd[rr].color=-1;
  }
}

@ Just as |purify| is analogous to |cover|, the inverse process is
analogous to |uncover|.

@<Sub...@>=
void unpurify(int p,ullng thresh) {
  register int cc,rr,nn,uu,dd,t,x;
  o,cc=nd[p].itm,x=nd[p].color; /* there's no need to clear |nd[cc].color| */
  for (o,rr=nd[cc].down;rr>=last_itm;o,rr=nd[rr].down) {
    if (o,nd[rr].cost>=thresh) break;
    if (o,nd[rr].color<0) o,nd[rr].color=x;
    else if (rr!=p) {
      for (nn=rr+1;nn!=rr;) {
        if (o,nd[nn].color>=0) {
          o,uu=nd[nn].up,dd=nd[nn].down;
          cc=nd[nn].itm;
          if (cc<=0) {
            nn=uu;@+continue;
          }
          oo,nd[uu].down=nd[dd].up=nn;
          if (cc<second) oo,nd[cc].len++;
        }
        nn++;
      }
    }
  }
}

@ Here's where we use the \.Z and \.z heuristics to provide
lower bounds that don't apply in general.

@<If the remaining cost is clearly too high, |goto backdown|@>=
if (ppgiven && cutoffcost!=infcost) {
  if (zgiven>1) {
    if (second-level*zgiven<=sortbufsize+1) pp=zgiven;
    else if (ppgiven&0xff) pp=0;
    else pp=-1;
  }@+else pp=zgiven;
  if (pp>=0) @<Go to |backdown| if the remaining min costs are too high@>@;
}

@ @<Go to |backdown| if the remaining min costs are too high@>=
{
  register ullng newcost,oldcost,acccost;
  acccost=curcost;
  for (r=0;Zchars[r];r++) o,cumcost[r]=curcost;
  for (o,k=cl[root].next,t=0;k!=root;o,k=cl[k].next) {
    o,p=nd[k].down;
    if (p<last_itm) {
      if (explaining) fprintf(stderr,
         "(Level "O"d, "O".8s's list is empty)\n",level,cl[k].name);
      goto backdown;
    }
    oo,cc=cl[k].name[0], tmpcost=nd[p].cost;
    for (r=0;Zchars[r];r++) if (Zchars[r]==cc) break;
    if (Zchars[r]) @<Include |tmpcost| in |cumcost[r]|@>@;
    else if (pp) @<Include |tmpcost| in |acccost|@>;
  }
}

@ @<Include |tmpcost| in |cumcost[r]|@>=
{
  if (o,cumcost[r]+tmpcost>=cutoffcost) {
    if (explaining) fprintf(stderr,
      "(Level "O"d, "O".8s's cost overflowed)\n",level,cl[k].name);
    goto backdown;
  }
  o,cumcost[r]+=tmpcost;
}

@ At this point |pp=zgiven| is a positive number $z$, and |cl[k]| is one
of the |pp| active items that doesn't begin with a \.Z-specified prefix.
We also know that exactly $kz={}$|second-1-level*z| primary
items are active, and that exactly $k$ more levels must be completed
before we have a solution.

The situation is simple when $z=1$. But when $z>1$, suppose the minimum
net costs of active items are $c_1\le c_2\le\cdots\le c_{kz}$.
Then we'll spend at least $c_z+c_{2z}+\cdots+c_{kz}$ while
covering them. A cute little online algorithm computes this lower bound nicely.

@<Include |tmpcost| in |acccost|@>=
{
  if (pp==1) {
    if (acccost+tmpcost>=cutoffcost) {
      if (explaining) fprintf(stderr,
        "(Level "O"d, "O".8s's cost overflowed)\n",level,cl[k].name);
      goto backdown;
    }
    acccost+=tmpcost;
  }@+else {
    /* we'll sort |tmpcost| into |sortbuf|, which has |t| costs already */
    for (p=t,oldcost=0;p;p--,oldcost=newcost) {
      o,newcost=sortbuf[sortbufsize-p];
      if (tmpcost<=newcost) break;
      if ((p mod pp)==0) {
        acccost+=newcost-oldcost;
        if (acccost>=cutoffcost) {
        if (explaining) fprintf(stderr,
          "(Level "O"d, "O".8s's cost overflowed)\n",level,cl[k].name);
          goto backdown;
        }
      }
      o,sortbuf[sortbufsize-p-1]=newcost; /* it had been |oldcost| */
    }
    if ((p mod pp)==0) {
      acccost+=tmpcost-oldcost;
      if (acccost>=cutoffcost) {
        if (explaining)
          fprintf(stderr,"("O".8s's cost caused overflow)\n",cl[k].name);
        goto backdown;
      }
    }
    o,sortbuf[sortbufsize-p-1]=tmpcost; /* it had been |oldcost| */
    t++;
  }
}  

@ The ``best item'' is considered to be an item that minimizes the
number of remaining choices. If there are several candidates, we
choose the leftmost one that has maximum minimum net cost (because
that cost must be paid somehow).

(This part of the program, whose logic is justified by the sorting that was
done during the input phase, represents the most significant changes between
{\mc DLX5} and {\mc DLX2}. I imagine that the heuristics used here might be
significantly improvable, especially for certain classes of problems.
For example, it may be better to do a 5-way branch on expensive choices
than a 2-way branch on cheap ones, because the expensive choices might quickly
peter out. And more elaborate ways to derive
lower bounds on the cost of covering the remaining primary items might be
based on the minimum cost per item in the remaining options. For example,
we could give each node a new field |optref|, which points to the
spacer following its option. Then the length of this option would readily
be obtained from that spacer, |nd[nd[p].optref]|.
One could use the currently dormant |cost| and |optref| fields of each spacer to
maintain a doubly linked list of options in order of their cost/item.
But I don't have time to investigate such ideas myself.)

@d explaining ((vbose&show_details) &&
    level<show_choices_max && level>=maxl-show_choices_gap)

@<Set |best_itm| to the best item for branching...@>=
t=max_nodes, tmpcost=0;
if (explaining) fprintf(stderr,"Level "O"d:",level);
for (o,k=cl[root].next;k!=root;o,k=cl[k].next) {
  o,p=nd[k].down;
  if (p==k) { /* the item list is empty, we must backtrack */
    if (explaining) fprintf(stderr," "O".8s(0)",cl[k].name);
    t=0,best_itm=k;
    break;
  }
  o,mincost=nd[p].cost;
  if (mincost>=cutoffcost-curcost) { /* no usable items, we must backtrack */
    if (explaining) fprintf(stderr," "O".8s(0$"O"llu)"@q$@>,
                                   cl[k].name,mincost);
    t=0,best_itm=k;
    break;
  }
  @<Look at the least-cost options for item |k|, possibly updating |best_itm|@>;
}
if (explaining)
  fprintf(stderr," branching on "O".8s("O"d)\n",cl[best_itm].name,t);
if (shape_file) {
  fprintf(shape_file,""O"d "O".8s\n",t,cl[best_itm].name);
  fflush(shape_file);
}
if (t==0) goto backdown;

@ At this point we know that |t>=1|, |p=nd[k].down!=k|, and
|mincost=nd[p].cost<cutoffcost-curcost|. Therefore |k| might turn out
to be the new |best_itm|.

@<Look at the least-cost options for item |k|...@>=
for (o,s=1,p=nd[p].down;;o,p=nd[p].down,s++) {
  if (p<last_itm || (o,nd[p].cost>=cutoffcost-curcost)) {
    if (explaining) fprintf(stderr,
      " "O".8s("O"d$"O"llu)"@q$@>,cl[k].name,s,mincost);
    break; /* there are |s| usable options in |k|'s item list */
  }
  if (s==t) { /* there are more than |t| usable options */
    if (explaining) fprintf(stderr,
      " "O".8s(>"O"d)",cl[k].name,t);
    goto no_change;
  }
  if (s>=lenthresh) { /* let's not search too far down the list */
    o,s=nd[k].len; /* be content with an upper bound */
    if (explaining) fprintf(stderr,
      " "O".8s("O"d?$"O"llu)"@q$@>,cl[k].name,s,mincost);
    break;
  }
}
if (s<t || (s==t && mincost>tmpcost))
  t=s,best_itm=k,tmpcost=mincost;
no_change:

@ @<Visit a solution and |goto recover|@>=
{
  nodes++; /* a solution is a special node, see 7.2.2--(4) */
  if (level+1>maxl) {
    if (level+1>=max_level) {
      fprintf(stderr,"Too many levels!\n");
      exit(-5);
    }
    maxl=level+1;
  }
  if (vbose&show_profile) profile[level+1]++;
  if (shape_file) {
    fprintf(shape_file,"sol\n");@+fflush(shape_file);
  }
  @<Update |cutoffcost|@>;
  @<Record solution and |goto recover|@>;
}

@ We remember the |kthresh| best costs found so far in a heap, with
|bestcost[h]>=bestcost[h+h+1]| and
|bestcost[h]>=bestcost[h+h+2]|. In particular, |bestcost[0]=cutoffcost| is the
largest of these net costs, and we remove it from the heap when a new
solution has been found.

When |kthresh| is even, this code uses the fact that |bestcost[kthresh]=0|.

@<Update |cutoffcost|@>=
{
  register int h,hh; /* a hole in the heap, and its larger successor */
  tmpcost=cutoffcost;
  for (h=0,hh=2;hh<=kthresh;hh=h+h+2) {
    if (oo,bestcost[hh]>bestcost[hh-1]) {
      if (nextcost<bestcost[hh]) o,bestcost[h]=bestcost[hh],h=hh;
      else break;
    }@+else if (nextcost<bestcost[hh-1]) o,bestcost[h]=bestcost[hh-1],h=hh-1;
    else break;
  }
  o,bestcost[h]=nextcost;
  o,cutoffcost=bestcost[0];
}

@ @<Sort the |bestcost| heap in preparation for final printing@>=
for (p=kthresh;p>2;p--) {
  register int h,hh; /* a hole in the heap, and its larger successor */
  nextcost=bestcost[p-1],bestcost[p-1]=0,bestcost[p]=bestcost[0];
  for (h=0,hh=2;hh<p;hh=h+h+2) {
    if (bestcost[hh]>bestcost[hh-1]) {
      if (nextcost<bestcost[hh]) bestcost[h]=bestcost[hh],h=hh;
      else break;
    }@+else if (nextcost<bestcost[hh-1]) bestcost[h]=bestcost[hh-1],h=hh-1;
    else break;
  }
  bestcost[h]=nextcost;
}
bestcost[p]=bestcost[0]; /* at this point |p=1| or |p=2| */
  /* now $|bestcost[1]|\le|bestcost[2]|\le\cdots\le|bestcost[kthresh]|$ */

@ @<Record solution and |goto recover|@>=
{
  count++;
  if (spacing && (count mod spacing==0)) {
    printf(""O"lld: (total cost $"O"llu)\n"@q$@>,count,totaltax+nextcost);
    for (k=0;k<=level;k++) print_option(choice[k],stdout,tmpcost-partcost[k]);
    fflush(stdout);
  }
  if (count>=maxcount) goto done;
  goto recover;
}

@ @<Sub...@>=
void print_state(void) {
  register int l;
  fprintf(stderr,"Current state (level "O"d):\n",level);
  for (l=0;l<level;l++) {
    print_option(choice[l],stderr,cutoffcost-partcost[l]);
    if (l>=show_levels_max) {
      fprintf(stderr," ...\n");
      break;
    }
  }
  if (cutoffcost<infcost)
    fprintf(stderr,
   " "O"lld solutions, $"O"llu, "O"lld mems, and max level "O"d so far.\n"@q$@>,
                              count,cutoffcost+totaltax,mems,maxl);
  else fprintf(stderr,
    " "O"lld solutions, "O"lld mems, and max level "O"d so far.\n",
                              count,mems,maxl);
}

@ During a long run, it's helpful to have some way to measure progress.
The following routine prints a string that indicates roughly where we
are in the search tree. The string consists of character pairs, separated
by blanks, where each character pair represents a branch of the search
tree. When a node has $d$ descendants and we are working on the $k$th,
the two characters respectively represent $k$ and~$d$ in a simple code;
namely, the values 0, 1, \dots, 61 are denoted by
$$\.0,\ \.1,\ \dots,\ \.9,\ \.a,\ \.b,\ \dots,\ \.z,\ \.A,\ \.B,\ \dots,\.Z.$$
All values greater than 61 are shown as `\.*'. Notice that as computation
proceeds, this string will increase lexicographically.

Following that string, a fractional estimate of total progress is computed,
based on the na{\"\i}ve assumption that the search tree has a uniform
branching structure. If the tree consists
of a single node, this estimate is~.5; otherwise, if the first choice
is `$k$ of~$d$', the estimate is $(k-1)/d$ plus $1/d$ times the
recursively evaluated estimate for the $k$th subtree. (This estimate
might obviously be very misleading, in some cases, but at least it
grows monotonically.)

@<Sub...@>=
void print_progress(void) {
  register int l,k,d,c,p;
  register double f,fd;
  if (cutoffcost<infcost)
    fprintf(stderr," after "O"lld mems: "O"lld sols, $"O"llu,"@q$@>,
                   mems,count,cutoffcost+totaltax);
  else fprintf(stderr," after "O"lld mems: "O"lld sols,",
                   mems,count);
  for (f=0.0,fd=1.0,l=0;l<level;l++) {
    c=nd[choice[l]].itm;
    for (k=1,p=nd[c].down;p!=choice[l];k++,p=nd[p].down) ;
    for (d=k-1;p>=last_itm;p=nd[p].down,d++)
      if (nd[p].cost>=cutoffcost-partcost[l]) break;
    fd*=d,f+=(k-1)/fd; /* choice |l| is |k| of |d| */
    fprintf(stderr," "O"c"O"c",
      k<10? '0'+k: k<36? 'a'+k-10: k<62? 'A'+k-36: '*',
      d<10? '0'+d: d<36? 'a'+d-10: d<62? 'A'+d-36: '*');
    if (l>=show_levels_max) {
      fprintf(stderr,"...");
      break;
    }
  }
  fprintf(stderr," "O".5f\n",f+0.5/fd);
}

@ @<Print the profile@>=
{
  fprintf(stderr,"Profile:\n");
  for (level=0;level<=maxl;level++)
    fprintf(stderr,""O"3d: "O"lld\n",
                              level,profile[level]);
}

@ @<Sub...@>=
int confusioncount;
void confusion(char *id) { /* an assertion has failed */
  if (confusioncount++==0) /* can fiddle with debugger */
    fprintf(stderr,"This can't happen (%s)!\n",id);
}

@*Index.