Goto Chapter: Top 1 2 3 4 5 Bib Ind
 [Top of Book]  [Contents]   [Previous Chapter]   [Next Chapter] 

3 Logged Rewriting Systems
 3.1 Logged Knuth-Bendix Completion
 3.2 Logged reduction of a word

3 Logged Rewriting Systems

A logged rewrite system is associated with a group presentation. Each logged rewrite rule contains, in addition to the standard rewrite rule, a record or log component which expresses the rule in terms of the original relators of the group. We represent such a rule by a triple [ u, [L1,L2,..,Lk], v], where [u,v] is a rewrite rule and L_i = [n_i,w_i] where n_i is a group relator and w_i is a word. These three components obey the identity u = n_1^w_1 ... n_k^w_k v.

Rules of the form g^+g^- apply to the monoid presentation, but not to the group presentation, so are given an empty logged component.

3.1 Logged Knuth-Bendix Completion

The functions in this section are the logged versions of those in the previous chapter.

3.1-1 LoggedOnePassKB
‣ LoggedOnePassKB( loggedrules )( operation )

Given a logged rewrite system, this function finds all the rules that would be added to complete the rewrite system of OnePassKB in 2.2-3, and also the logs which relate the new rules to the originals. The result of applying this function to loggedrules is to add new logged rules to the system without changing the monoid it defines.

In the example, we first convert the presentation for q8 into an initial set of logged rules, and then apply one pass of Knuth-Bendix.

In the development version the function LoggedOnePassKB returns a two-element list of lists. The first element is a set of logged rules and the second element is an empty list of r-sequences, whatever they are! (Need more here.)


gap> l0 := ListWithIdenticalEntries( 8, 0 );;
gap> for j in [1..8] do 
>        r := r0[j];
>        if ( j<5 ) then
>           l0[j] := [ r[1], [ [j,id] ], r[2] ];
>        else
>           l0[j] := [ r[1], [ ], r[2] ];
>        fi;
>    od;
gap> l0;
[ [ q8_M1^4, [ [ 1, <identity ...>] ], <identity. ..> ], 
  [ q8_M2^4, [ [ 2, <identity ...>] ], <identity ...> ], 
  [ q8_M1*q8_M2*q8_M1*q8_M4, [ [ 3, <identity ...> ] ], <identity ...> ],   
  [ q8_M1^2*q8_M2^2, [ [ 4, <identity ...> ] ], <identity ...> ], 
  [ q8_M1*q8_M3, [ ], <identity ...> ], [ q8_M2*q8_M4, [ ], <identity ...> ], 
  [ q8_M3*q8_M1, [ ], <identity ...> ], [ q8_M4*q8_M2, [ ], <identity ...> ] ] 
gap> l1 := LoggedOnePassKB( l0 );;
gap> Length( l1[1] ); 
21

Note that the length 21 of l1 is, as expected, the same as that of r1 in 2.2-3.

3.1-2 LoggedKnuthBendix
‣ LoggedKnuthBendix( loggedrules )( operation )
‣ LoggedRewriteReduce( loggedrules )( operation )

The function LoggedRewriteReduce removes unnecessary rules from a logged rewrite system. It works on the same principle as RewriteReduce in 2.2-3.

The function LoggedKnuthBendix repeatedly applies functions LoggedOnePassKB and LoggedRewriteReduce until no new rules are added and no unnecessary ones are included. The output is a reduced complete logged rewrite system.

In order to clarify the meaning of the rules below, consider the second rule in l2 which shows how b^+a^+ reduces to a^+b^-. For this rule [u,L,v] we will verify that u = n_1^w_1n_2^w_2n_3^w_3 v, as in the introduction to this chapter. The rule is:


[ q8_M2*q8_M1, [ [3,q8_M3^-1], [-1,<identity...>], [4,q8_M1^-1] ], q8_M1*q8_M4 ]. 

As before, we write a^+, b^+, a^-, b^- for q8_M1,q8_M2,q8_M3,q8_M4 respectively. The relators 3,4 and the inverse of relator 1 (denoted by -1) are a^+b^+a^+b^-, a^+2b^+2 and a^-4, and these are conjugated by (a^-)^-1=a^+, (a^+)^-1=a^- and the identity respectively. So the second and third parts of the rule expand to:

(a^-(a^+b^+a^+b^-)a^+)(a^{-4})(a^+(a^{+2}b^{+2})a^-)a^+b^- ~=~ (a^-a^+)b^+a^+(b^-(a^+a^{-4}a^{+3})b^+)(b^+(a^-a^+)b^-)

which reduces to the first part of the rule, b^+a^+.

In the development version the function LoggedOnePassKB returns a two-element list of lists. The first element is a set of logged rules and the second element is a long list of r-sequences, whatever they are! (Need more here.)


gap> l11 := LoggedRewriteReduce( l1[1] );
[ [ q8_M1*q8_M3, [ ], <identity ...> ], 
  [ q8_M2^2, [ [ -4, <identity ...> ], [ 2, q8_M1^-2 ] ], q8_M1^2 ], 
  [ q8_M2*q8_M4, [ ], <identity ...> ], [ q8_M3*q8_M1, [ ], <identity ...> ], 
      [ q8_M4*q8_M2, [ ], <identity ...> ], 
  [ q8_M1^3, [ [ 1, <identity. ..> ] ], q8_M3 ], 
  [ q8_M1^2*q8_M2, [ [ 4, <identity. ..> ] ], q8_M4 ], 
  [ q8_M1*q8_M2*q8_M1, [ [ 3, <identity ...> ] ], q8_M2 ], 
  [ q8_M2*q8_M1*q8_M4, [ [ 3, q8_M3^-1 ] ], q8_M3] ]
gap> Length( l11 ); 
9
gap> l2 := LoggedKnuthBendix( l11 );;
gap> l2[1];
[ [ q8_M1*q8_M3, [ ], <identity ...> ], 
  [ q8_M2*q8_M1, [ [ 3, q8_M3^-1 ], [-1, <identity. ..> ], [ 4, q8_M1^-1 ] ], 
      q8_M1*q8_M4 ], 
  [ q8_M2^2, [ [ -4, <identity ...> ], [2, q8_M1^-2] ], q8_M1^2 ], 
  [ q8_M2*q8_M3, [ [ -3, <identity ...> ] ], q8_M1*q8_M2 ], 
  [ q8_M2*q8_M4, [ ], <identity ...> ], [ q8_M3*q8_M1, [ ], <identity ...> ], 
  [ q8_M3*q8_M2, [ [ -1, <identity ...>], [4, q8_M1^-1 ] ], q8_M1*q8_M4 ],
  [ q8_M3^2, [ [ -1, <identity ...>] ], q8_M1^2 ], 
  [ q8_M3*q8_M4, [ [ -1, <identity ...>], [ -2, q8_M1^-2], 
        [ 4, <identity ...> ], [ 3, q8_M3^-1*q8_M2^-1 ], 
        [ -3, <identity. ..> ] ], q8_M1*q8_M2 ], 
  [ q8_M4*q8_M1, [ [ -4, <identity ...> ], [ 3, q8_M1^-1 ] ], q8_M1*q8_M2 ], 
  [ q8_M4*q8_M2, [ ], <identity ...> ], 
  [ q8_M4*q8_M3, [ [ -3, q8_M3^-1*q8_M4^-1] ], q8_M1*q8_M4 ], 
  [ q8_M4^2, [ [ -4, <identity. ..> ] ], q8_M1^2 ], 
  [ q8_M1^3, [ [ 1, <identity ...> ] ], q8_M3 ], 
  [ q8_M1^2*q8_M2, [ [4, <identity. ..> ] ], q8_M4 ], 
  [ q8_M1^2*q8_M4, [ [ -4, q8_M1^-2], [ 1, <identity ...> ] ], q8_M2 ] ] 
gap> Length( l2[1] );
16
gap> Length( l2[2] );
49

3.2 Logged reduction of a word

3.2-1 LoggedReduceWordKB
‣ LoggedReduceWordKB( word, loggedrules )( operation )
‣ LoggedOnePassReduceWord( word, loggedrules )( operation )
‣ ShorterLoggedRule( logrule1, logrule2 )( operation )

Given a word and a logged rewrite system, the function LoggedOnePassReduceWord makes one reduction pass of the word (possibly involving several reductions) (as does OnePassReduceWord in 2.2-2) and records this, using the log part of the rule(s) used and the position in the original word of the replaced part.

The function LoggedReduceWordKB repeatedly applies OnePassLoggedReduceWord until the word can no longer be reduced. Each step of the reduction is logged, showing how the original word can be expressed in terms of the original relators and the irreducible word. When loggedrules is complete the reduced word is a unique normal form for that group element. The log of the reduction depends on the order in which the rules are applied.

The function ShorterLoggedrule decides whether one logged rule is better than another, using the same criteria as ShorterRule in 2.2-3. In the example we perform logged reductions of w_0 = a^9b^9 corresponding to the ordinary reductions performed in the previous chapter (section 2.2-2).

In order to clarify the following output, note that, in the log below, b^9a^9 reduces to b^5a^5 in lw1 and to ba in the first lw2. These expand to the initial w0 using the given logged parts as follows:

(b^{+9}a^{+4}b^{-9})(b^{+4})b^{+5}a^{+5} ~=~ b^{+9}a^{+9} ~=~ (b^{+9}a^{+4}b^{-9})(b^{+4})(b^{+5}a^{+4}b^{-5})(b^{+4})b^+a^+\,.

The corresponding expansion of the final lw2 is too lengthy to include here.


gap> w0; 
q8_M2^9*q8_M1^9
gap> lw1 := LoggedOnePassReduceWord( w0, l0 );
[ [ [ 1, q8_M2^-9 ], [ 2, <identity ...> ] ], q8_M2^5*q8_M1^5 ]
gap> lw2 := LoggedReduceWordKB( w0, l0 ); 
[ [ [ 1, q8_M2^-9 ], [ 2, <identity ...> ], [ 1, q8_M2^-5 ],
      [ 2, <identity ...> ] ], q8_M2*q8_M1 ]
gap> lw2 := LoggedReduceWordKB( w0, l2[1] ); 
[ [ [ 3, q8_M3^-1*q8_M2^-8 ], [ -1, q8_M2^-8 ], [ 4, q8_M1^-1*q8_M2^-8 ], 
      [ -4, <identity ...> ], [ 2, q8_M1^-2 ], 
      [ -4, q8_M1^-1*q8_M2^-6*q8_M1^-2 ], [ 3, q8_M1^-2*q8_M2^-6*q8_M1^-2 ], 
      [ 1, q8_M2^-1*q8_M1^-2*q8_M2^-6*q8_M1^-2 ], [ 4, <identity ...> ], 
      [ 3, q8_M3^-1*q8_M2^-4*q8_M4^-1 ], [ -1, q8_M2^-4*q8_M4^-1 ], 
      [ 4, q8_M1^-1*q8_M2^-4*q8_M4^-1 ], [ -4, q8_M4^-1 ], 
      [ 2, q8_M1^-2*q8_M4^-1 ], 
      [ -3, q8_M1^-1*q8_M4^-1*q8_M1^-1*q8_M2^-2*q8_M1^-2*q8_M4^-1 ], 
      [ -4, <identity ...> ], [ 3, q8_M1^-1 ], 
      [ 1, q8_M2^-1*q8_M1^-2*q8_M4^-1*q8_M1^-1*q8_M2^-1*(q8_M2^-1*q8_M1^-1)^2 
         ], [ 4, q8_M4^-1*q8_M1^-1*q8_M2^-1*(q8_M2^-1*q8_M1^-1)^2 ], 
      [ 3, q8_M3^-1*q8_M1^-1 ], [ -1, q8_M1^-1 ], [ 4, q8_M1^-2 ], 
      [ -4, q8_M4^-1*q8_M1^-2 ], [ 2, q8_M1^-2*q8_M4^-1*q8_M1^-2 ], 
      [ -4, q8_M1^-2 ], [ 3, q8_M1^-3 ], [ -4, q8_M1^-2*q8_M2^-1*q8_M1^-3 ], 
      [ 1, <identity ...> ], [ 3, q8_M3^-2 ], [ -1, q8_M3^-1 ], 
      [ 4, q8_M1^-1*q8_M3^-1 ], [ -4, <identity ...> ], [ 3, q8_M1^-1 ], 
      [ 3, q8_M3^-1*q8_M1^-1 ], [ -1, q8_M1^-1 ], [ 4, q8_M1^-2 ], 
      [ -4, q8_M1^-2 ], [ 3, q8_M1^-3 ], [ 1, <identity ...> ], 
      [ -1, <identity ...> ], [ 4, q8_M1^-1 ] ], q8_M1*q8_M4 ]

3.2-2 LoggedRewritingSystemFpGroup
‣ LoggedRewritingSystemFpGroup( loggedrules )( attribute )

Given a group presentation, the function LoggedRewritingSystemFpGroup determines a logged rewrite system based on the relators. The initial logged rewrite system associated with a group presentation consists of two types of rule. These are logged versions of the two types of rule in the monoid presentation. For each relator rel of the group there is a logged rule [rel,[[1,rel]],id]. For each inverse relator there is a logged rule [ gen*inv, [], id ]. The function then attempts a completion of the logged rewrite system. The rules in the final system are partially ordered by the function ShorterLoggedRule.


gap> lwrs := LoggedRewritingSystemFpGroup( q8 );
[ [ q8_M4*q8_M2, [ ], <identity ...> ], [ q8_M3*q8_Ml, [ ], <identity ...> ], 
    [ q8_M2*q8_M4, [ ], <identity ...> ], 
  [ q8_Ml*q8_M3, [ ], <identity ...> ], 
  [ q8_Ml^2*q8_M4, [ [ -8, q8_Ml^-2 ], [ 5, <identity ...> ] ], q8_M2 ], 
  [ q8_Ml^2*q8_M2, [ [ 8, <identity ...> ] ], q8_M4 ], 
  [ q8_Ml^3, [ [ 5, <identity ...> ] ], q8_M3 ], 
  [ q8_M4^2, [ [ -8, <identity ...> ] ], q8_Ml^2 ], 
  [ q8_M4*q8_M3, [ [ -7, q8_M3^-1*q8_M4^-1 ] ], q8_Ml*q8_M4 ], 
  [ q8_M4*q8_Ml, [ [ -8, <identity. ..> ], [ 7, q8_Ml^-1 ] ], q8_Ml*q8_M2 ], 
  [ q8_M3*q8_M4, 
      [ [ -5, <identity ...>], [ -6, q8_Ml^-2 ], [ 8, <identity ...> ], 
          [ 7, q8_M3^-1*q8_M2^-1 ], [ -7, <identity. ..> ] ], q8_Ml*q8_M2 ], 
  [ q8_M3^2, [ [ -5, <identity ...>] ], q8_Ml^2 ], 
  [ q8_M3*q8_M2, [ [ -5, <identity. ..> ], [ 8, q8_Ml^-1 ] ], q8_Ml*q8_M4 ], 
  [ q8_M2*q8_M3, [ [ -7, <identity ...> ] ], q8_M1*q8_M2 ], 
  [ q8_M2^2, [ [ -a, <identity ...> ], [ 6, q8_M1^-2 ] ], q8_M1^2 ], 
  [ q8_M2*q8_M1, [ [ 7, q8_M3^-1 ], [ -5, <identity ...> ], [ a, q8_M1^-1 ] ], 
      q8_M1*q8_M4 ] ] 
gap> Length( lwrs );
16

Consider now the two-generator abelian group T considered in the previous chapter (2.2-1). Using the alternative ordering on the monoid generators, [ T_M1=a^+, T_M2=a^-, T_M3=b^+, T_M4=b^- ], we obtain the following set of 8 logged rules. The last of these may be checked as follows:

(b^+a^+(b^-a^-b^+a^+)a^-b^-)a^+b^+ ~=~ b^+a^+(b^-(a^-(b^+(a^+a^-)b^-)a^+)b^+)

and is a logged version of the rule b^+a^+ -> a^+b^+.


gap> lrwsT := LoggedRewritingSystemFpGroup( T );
[ [ T_M4*T_M3, [  ], <identity ...> ], [ T_M3*T_M4, [  ], <identity ...> ], 
  [ T_M2*T_M1, [  ], <identity ...> ], [ T_M1*T_M2, [  ], <identity ...> ], 
  [ T_M4*T_M2, [ [ -5, <identity ...> ] ], T_M2*T_M4 ], 
  [ T_M4*T_M1, [ [ 5, T_M1^-1 ] ], T_M1*T_M4 ], 
  [ T_M3*T_M2, [ [ 5, T_M3^-1 ] ], T_M2*T_M3 ], 
  [ T_M3*T_M1, [ [ -5, T_M1^-1*T_M3^-1 ] ], T_M1*T_M3 ] ]

 [Top of Book]  [Contents]   [Previous Chapter]   [Next Chapter] 
Goto Chapter: Top 1 2 3 4 5 Bib Ind

generated by GAPDoc2HTML