% This file was created automatically from grppc.msk.
% DO NOT EDIT!
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%%
%A  grppc.msk                   GAP documentation            Alexander Hulpke
%A                                                               Bettina Eick
%%
%A  @(#)$Id: grppc.msk,v 1.40.2.3 2006/09/16 19:02:49 jjm Exp $
%%
%Y  (C) 1998 School Math and Comp. Sci., University of St.  Andrews, Scotland
%Y  Copyright (C) 2002 The GAP Group
%%
\Chapter{Pc Groups}

%
% Wunsch von Herrn Neubueser: PrintPcPresentation
%
PcGroups are polycyclic groups that use the polycyclic presentation for
element arithmetic. This presentation gives them a ``natural'' pcgs, the
`FamilyPcgs' (see~"FamilyPcgs") with respect to which pcgs operations as
described in chapter~"Polycyclic Groups" are particularly efficient.

Let $G$ be a polycyclic group with pcgs $P = (g_1, \ldots, g_n)$
and corresponding relative orders $(r_1, \ldots, r_n)$. Recall that the
$r_i$ are positive integers or infinity and let $I$ be the set of indices
$i$ with $r_i$ a positive integer. Then $G$ has a finite presentation
on the generators $g_1, \ldots, g_n$ with relations of the following
form.
$$
\matrix{
       g_i^{r_i} &=& g_{i+1}^{a(i,i,i+1)} \cdots g_n^{a(i,i,n)}\cr
                 & & \hbox{ for } 1 \leq i \leq n \hbox{ and } i \in I\cr
g_i^{-1} g_j g_i &=& g_{i+1}^{a(i,j,i+1)} \cdots g_n^{a(i,j,n)}\cr
                 & & \hbox{ for } 1 \leq i \< j \leq n\cr
}
$$
For infinite groups we need additionally
$$
\matrix{
g_i^{-1} g_j^{-1} g_i &=& g_{i+1}^{b(i,j,i+1)} \cdots g_n^{b(i,j,n)}\cr
                      & & \hbox{ for } 1 \leq i \< j \leq n%
                          \hbox{ and } j \not \in I\cr
     g_i g_j g_i^{-1} &=& g_{i+1}^{c(i,j,i+1)} \cdots g_n^{c(i,j,n)}\cr
                      & & \hbox{ for } 1 \leq i \< j \leq n%
                          \hbox{ and } i \not \in I\cr
g_i g_j^{-1} g_i^{-1} &=& g_{i+1}^{d(i,j,i+1)} \cdots g_n^{d(i,j,n)}\cr
                      & & \hbox{ for } 1 \leq i \< j \leq n%
                          \hbox{ and } i, j, \not \in I\cr
}
$$
Here the right hand sides are assumed to be words in normal
form; that is, for $k \in I$ we have for all exponents
$0 \leq a(i,j,k), b(i,j,k), c(i,j,k), d(i,j,k) \< r_k$.

A finite presentation of this type is called a *power-conjugate
presentation* and a *pc group* is a polycyclic group defined
by a power-conjugate presentation. Instead of conjugates we could
just as well work with commutators and then the presentation would
be called a *power-commutator* presentation. Both types of presentation
are abbreviated as *pc presentation*. Note that a pc presentation is a
rewriting system.

Clearly, whenever a group $G$ with pcgs $P$ is given, then we can
write down the corresponding pc presentation. On the other hand,
one may just write down a presentation on $n$ abstract generators
$g_1, \ldots, g_n$ with relations of the above form and define a
group $H$ by this. Then the subgroups $C_i = \langle g_i, \ldots, g_n
\rangle$ of $H$ form a subnormal series whose factors are cyclic
or trivial. In the case that all factors are non-trivial, we say
that the pc presentation of $H$ is *confluent*. Note that {\GAP}~4
can only work correctly with pc groups defined by a confluent pc
presentation.

At the current level of implementation {\GAP} can only deal with
finite pc groups. This will be extended in near future.

%%  The Pc groups were designed and implemented by Frank Celler and
%%  Werner Nickel, the code for Two-Cohomology and Extensions is due to
%%  Bettina Eick who also wrote the Random Isomorphism test together with Hans
%%  Ulrich Besche.

Algorithms for pc groups use the methods for polycyclic groups described in
chapter~"Polycyclic Groups".

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{The family pcgs}

Clearly, the generators of a power-conjugate presentation of
a pc group $G$ form a pcgs of the pc group. This pcgs is called
the *family pcgs*.

\>FamilyPcgs( <grp> ) A


\>IsFamilyPcgs( <pcgs> ) P


\>InducedPcgsWrtFamilyPcgs( <grp> ) A


\>IsParentPcgsFamilyPcgs( <pcgs> ) P

This property indicates that the pcgs <pcgs> is induced with respect to
a family pcgs.


In {\GAP}~3 the family pcgs had been the only pcgs allowed for a
pc group. Note that this has changed in {\GAP}~4 where a pc group
may have several independent polycyclic generating sequences.

However, the elementary operations for a non-family pcgs may
not be as efficient as the elementary operations for the family
pcgs.

This can have a significant influence on the performance of algorithms for
polycyclic groups. Many algorithms require a pcgs that corresponds to an
elementary abelian series (see~"PcgsElementaryAbelianSeries") or even a
special pcgs (see~"Special Pcgs"). If the family pcgs has the required
properties, it will be used for these purposes, if not {\GAP} has to work
with respect to a new pcgs which is *not* the family pcgs and thus takes
longer for elementary calculations like `ExponentsOfPcElement'.

Therefore, if the family pcgs chosen for arithmetic is not of importance it
might be worth to *change* to another, nicer, pcgs to speed up calculations.
This can be achieved, for example, by using the `Range' of the isomorphism
obtained by `IsomorphismSpecialPcGroup' (see~"IsomorphismSpecialPcGroup").

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Elements of pc groups}

The elements of a pc group $G$ are always represented as words
in normal form with respect to the family pcgs of $G$.
Thus it is straightforward to compare elements of pc group,
since this boils down to a mere comparison of exponent vectors
with respect to the family pcgs. In particular, the word problem
is efficiently solvable in pc groups.

\>`<pcword> = <pcword>'{equality!pcwords}
\>`<pcword> \< <pcword>'{smaller!pcwords}

However, multiplication and inversion of elements in pc groups
is not as straightforward as in arbitrary finitely presented groups
where a simple concatenation or reversion of the corresponding
words is sufficient (but one cannot solve the word problem).

To multiply to elements in a pc group, we
first concatenate the corresponding words and then use an algorithm
called `collection' to transform the new word into a word in normal
form.

\beginexample
gap> g := FamilyPcgs( SmallGroup( 24, 12 ) );
Pcgs([ f1, f2, f3, f4 ])
gap> g[4] * g[1];
f1*f3
gap> (g[2] * g[3])^-1;
f2^2*f3*f4
\endexample

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Pc groups versus fp groups}

In theory pc groups are finitely presented groups.  In practice the
arithmetic in pc groups is different from the arithmetic in fp
groups. Thus for technical reasons the pc groups in {\GAP} do not form
a subcategory of the fp groups and hence the methods for fp groups
cannot be applied to pc groups in general.

\>IsPcGroup( <G> ) C

tests whether <G> is a pc group.


\beginexample
gap> G := SmallGroup( 24, 12 );
<pc group of size 24 with 4 generators>
gap> IsPcGroup( G );
true
gap> IsFpGroup( G );
false
\endexample

Note that it is possible to convert a pc group to a fp group in {\GAP}.
The following function computes the power-commutator presentation defined
by <pcgs>. The string <str> can be used to give a name to the generators
of the fp group.

\> IsomorphismFpGroupByPcgs( <pcgs>, <str> )

\beginexample
gap> p := FamilyPcgs( SmallGroup( 24, 12 ) );
Pcgs([ f1, f2, f3, f4 ])
gap> iso := IsomorphismFpGroupByPcgs( p, "g" );
[ f1, f2, f3, f4 ] -> [ g1, g2, g3, g4 ]
gap> F := Image( iso );
<fp group of size 24 on the generators [ g1, g2, g3, g4 ]>
gap> RelatorsOfFpGroup( F );
[ g1^2, g2^-1*g1^-1*g2*g1*g2^-1, g3^-1*g1^-1*g3*g1*g4^-1*g3^-1, 
  g4^-1*g1^-1*g4*g1*g4^-1*g3^-1, g2^3, g3^-1*g2^-1*g3*g2*g4^-1*g3^-1, 
  g4^-1*g2^-1*g4*g2*g3^-1, g3^2, g4^-1*g3^-1*g4*g3, g4^2 ]
\endexample

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Constructing Pc Groups}

If necessary, you can supply {\GAP} with a pc presentation
by hand. (Although this is the most tedious way to input a pc group.)
Note that the pc presentation has to be confluent in order to
work with the pc group in {\GAP}.

(If you have already a suitable pcgs in another representation, use
`PcGroupWithPcgs', see below~"PcGroupWithPcgs".)

One way is to define a finitely presented group with a
pc presentation in {\GAP} and then convert this presentation
into a pc group. Note that this does not work for arbitrary presentations
of polycyclic groups, see Chapter "Quotient Methods" for further information.

For performance reasons it is beneficial to enforce a ``syllable''
representation in the free group (see~"Representations for Associative
Words").

\>PcGroupFpGroup( <G> ) F

creates a PcGroup <P> from an FpGroup (see Chapter "Finitely Presented
Groups") <G> whose presentation is polycyclic. The resulting group <P>
has generators corresponding to the generators of <G>. They are printed
in the same way as generators of <G>, but they lie in a different
family. If the pc presentation of <G> is not confluent, an error message
occurs.


\beginexample
gap> F := FreeGroup(IsSyllableWordsFamily,"a","b","c","d");;
gap> a := F.1;; b := F.2;; c := F.3;; d := F.4;;
gap> rels := [a^2, b^3, c^2, d^2, Comm(b,a)/b, Comm(c,a)/d, Comm(d,a),
>             Comm(c,b)/(c*d), Comm(d,b)/c, Comm(d,c)];
[ a^2, b^3, c^2, d^2, b^-1*a^-1*b*a*b^-1, c^-1*a^-1*c*a*d^-1, d^-1*a^-1*d*a, 
  c^-1*b^-1*c*b*d^-1*c^-1, d^-1*b^-1*d*b*c^-1, d^-1*c^-1*d*c ]
gap> G := F / rels;
<fp group on the generators [ a, b, c, d ]>
gap> H := PcGroupFpGroup( G );
<pc group of size 24 with 4 generators>
\endexample

Equivalently to the above method one can initiate a collector of
a pc group by hand and use it to define a pc group. In {\GAP} there
are different collectors for different collecting strategies; at
the moment, there are two collectors to choose from: the single
collector for finite pc groups and the combinatorial collector
for finite $p$-groups. See \cite{Sims94} for further information
on collecting strategies.

A collector is initiated by underlying free group to the pc presented
group and the relative orders of the pc series. Then one adds the right
hand sides of the power and the commutator or conjugate relations one by
one. Note that omitted relators are assumed to be trivial.

\>SingleCollector( <fgrp>, <relorders> )
\>CombinatorialCollector( <fgrp>, <relorders> )

Then the right hand sides of the pc presentation have to be declared.
Let $f_1, \ldots, f_n$ be the generators of the underlying free group
<fgrp>.

A combinatorial collector can only be set up for a finite $p$-group.
Here, the relative orders <relorders> must all be equal and a prime.

\> SetConjugate( <coll>, <j>, <i>, <w> )

set the conjugate $f_j^{f_i}$ to equal <w> where <w> is a word
in $f_{i+1}, \ldots, f_n$ and $i\<j$.

\> SetCommutator( <coll>, <j>, <i>, <w> )

set the commutator of $f_j$ and $f_i$ to equal <w> where <w> is a word
in $f_{i+1}, \ldots, f_n$ and $i\<j$.

\> SetPower( <coll>, <i>, <w> )

set the power $f_i^{r_i}$ to equal <w> where <w> is a word
in $f_{i+1}, \ldots, f_n$.

Finally, the collector has to be converted to a group.

\> GroupByRws( <coll> )
\> GroupByRwsNC( <coll> )

creates a group from a rewriting system. In the first version it
is checked whether the rewriting system is confluent, in the second
version this is assumed to be true.

\> IsConfluent( <G> )!{for pc groups}

checks whether the pc group <G> has been build from a collector with
a confluent power-commutator presentation.

\beginexample
gap> F := FreeGroup(IsSyllableWordsFamily, 2 );;
gap> coll1 := SingleCollector( F, [2,3] );
<<single collector, 8 Bits>>
gap> SetConjugate( coll1, 2, 1, F.2 );
gap> SetPower( coll1, 1, F.2 );
gap> G1 := GroupByRws( coll1 );
<pc group of size 6 with 2 generators>
gap> IsConfluent(G1);
true
gap> IsAbelian(G1);
true
\endexample

\beginexample
gap> coll2 := SingleCollector( F, [2,3] );
<<single collector, 8 Bits>>
gap> SetConjugate( coll2, 2, 1, F.2^2 );
gap> G2 := GroupByRws( coll2 );
<pc group of size 6 with 2 generators>
gap> IsAbelian(G2);
false
\endexample

With the above methods a pc group with arbitrary defining pcgs can be
constructed. However, for almost all applications within {\GAP} we need
to have a pc group whose defining pcgs is a prime order pcgs. Hence
the following functions are useful.

\>IsomorphismRefinedPcGroup( <G> ) A

\index{isomorphic!pc group}
returns an isomorphism from <G> onto an isomorphic PC group whose family
pcgs is a prime order pcgs.

\>RefinedPcGroup( <G> ) A

returns the range of `IsomorphismRefinedPcGroup(<G>)'.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Computing Pc Groups}

Another possibility to get a pc group in {\GAP} is to convert a
polycyclic group given by some other representation to a pc group.
For finitely presented groups there are various quotient methods
available. For all other types of groups one can use the following
functions.

\>PcGroupWithPcgs( <mpcgs> ) A

creates a new Pc group <G> whose family pcgs is isomorphic to the
(modulo) pcgs <mpcgs>.


\beginexample
gap> G := Group( (1,2,3), (3,4,1) );;
gap> PcGroupWithPcgs( Pcgs(G) );
<pc group of size 12 with 3 generators>
\endexample

If a pcgs is only given by a list of pc elements, `PcgsByPcSequence'
(see~"PcgsByPcSequence") can be used:
\beginexample
gap> G:=Group((1,2,3,4),(1,2));;
gap> p:=PcgsByPcSequence(FamilyObj(One(G)),
> [ (3,4), (2,4,3), (1,4)(2,3), (1,3)(2,4) ]);
Pcgs([ (3,4), (2,4,3), (1,4)(2,3), (1,3)(2,4) ])
gap> PcGroupWithPcgs(p);
<pc group of size 24 with 4 generators>
\endexample

\beginexample
gap> G := SymmetricGroup( 5 );
Sym( [ 1 .. 5 ] )
gap> H := Subgroup( G, [(1,2,3,4,5), (3,4,5)] );
Group([ (1,2,3,4,5), (3,4,5) ])
gap> modu := ModuloPcgs( G, H );
Pcgs([ (4,5) ])
gap> PcGroupWithPcgs(modu);
<pc group of size 2 with 1 generators>
\endexample

\>IsomorphismPcGroup( <G> ) A

\index{isomorphic!pc group}
returns an isomorphism from <G> onto an isomorphic PC group.
The series chosen for this PC representation depends on
the method chosen.
<G> must be a polycyclic group of any kind, for example a solvable
permutation group.


\beginexample
gap> G := Group( (1,2,3), (3,4,1) );;
gap> iso := IsomorphismPcGroup( G );
Pcgs([ (2,4,3), (1,2)(3,4), (1,3)(2,4) ]) -> [ f1, f2, f3 ]
gap> H := Image( iso );
Group([ f1, f2, f3 ])
\endexample

\>IsomorphismSpecialPcGroup( <G> ) A

returns an isomorphism from <G> onto an isomorphic PC group whose family
pcgs is a special pcgs. (This can be beneficial to the runtime of
calculations.)
<G> may be a polycyclic group of any kind, for example a solvable
permutation group.


%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Saving a Pc Group}

As printing a polycyclic group does not display the presentation,
one cannot simply print a pc group to a file to save it. For this
purpose we need the following function.

\>GapInputPcGroup( <grp>, <string> ) F



\beginexample
gap> G := SmallGroup( 24, 12 );
<pc group of size 24 with 4 generators>
gap> PrintTo( "save", GapInputPcGroup( G, "H" ) );
gap> Read( "save" );
#I A group of order 24 has been defined.
#I It is called H
gap> H = G;
false
gap> IdSmallGroup( H ) = IdSmallGroup( G );
true
\endexample

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Operations for Pc Groups}

All the operations described in Chapters "Groups" and "Polycyclic
Groups" apply to a pc group. Nearly all methods for pc groups are
methods for groups with pcgs as described in Chapter "Polycyclic
groups". The only method with is special for pc groups is a method
to compute intersections of subgroups, since here a pcgs of a parent
group is needed and this can only by guaranteed within pc groups.

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{2-Cohomology and Extensions}

One of the most interesting applications of pc groups is the possibility
to compute with extensions of these groups by elementary abelian groups;
that is, $H$ is an extension of $G$ by $M$, if there exists a normal
subgroup $N$ in $H$ which is isomorphic to $M$ such that $H/N$ is
isomorphic to $G$.

Pc groups are particularly suited for such applications, since the
2-cohomology can be computed efficiently for such groups and,
moreover, extensions of pc groups by elementary abelian groups can
be represented as pc groups again.

To define the elementary abelian group $M$ together with an action of $G$ on
$M$ we consider $M$ as a meataxe module for $G$ over a finite field
(section~"IrreducibleModules" describes functions that can be used to obtain
certain modules). For further information on meataxe modules see Chapter
"The MeatAxe".  Note that the matrices defining the module must correspond
to the pcgs of the group <G>.

\>TwoCoboundaries( <G>, <M> ) O

returns the group of 2-coboundaries of a pc group <G> by the <G>-module 
<M>. The generators of <M> must correspond to Pcgs(<G>). The group of 
coboundaries is given as vector space over the field underlying <M>.

\>TwoCocycles( <G>, <M> ) O

returns the 2-cocycles of a pc group <G> by the <G>-module <M>. 
The generators of <M> must correspond to Pcgs(<G>). The operation
returns a list of vectors over the field underlying <M> and the 
additive group generated by these vectors is the group of 2-cocyles.

\>TwoCohomology( <G>, <M> ) O

returns a record defining the second cohomology group as factor space of 
the space of cocycles by the space of coboundaries. <G> must be a pc group
and the generators of <M> must correspond to the pcgs of <G>.


\beginexample
gap> G := SmallGroup( 4, 2 );
<pc group of size 4 with 2 generators>
gap> mats := List( Pcgs( G ), x -> IdentityMat( 1, GF(2) ) );
[ [ <a GF2 vector of length 1> ], [ <a GF2 vector of length 1> ] ]
gap> M := GModuleByMats( mats, GF(2) );
rec( field := GF(2), isMTXModule := true, dimension := 1, 
  generators := [ <an immutable 1x1 matrix over GF2>, 
      <an immutable 1x1 matrix over GF2> ] )
gap> TwoCoboundaries( G, M );
[  ]
gap> TwoCocycles( G, M );
[ [ Z(2)^0, 0*Z(2), 0*Z(2) ], [ 0*Z(2), Z(2)^0, 0*Z(2) ], 
  [ 0*Z(2), 0*Z(2), Z(2)^0 ] ]
gap> cc := TwoCohomology( G, M );;
gap> cc.cohom;
<linear mapping by matrix, <vector space of dimension 3 over GF(2)> -> ( GF(
2)^3 )>
\endexample

\>Extensions( <G>, <M> ) O

returns all extensions of <G> by the <G>-module <M> up to equivalence
as pc groups.

\>Extension( <G>, <M>, <c> ) O
\>ExtensionNC( <G>, <M>, <c> ) O

returns the extension of <G> by the <G>-module <M> via the cocycle <c>
as pc groups. The `NC' version does not check the resulting group for
consistence.

\>SplitExtension( <G>, <M> )

returns the split extension of <G> by the <G>-module <M>.

\> ModuleOfExtension( <E> ) 

returns the module of an extension <E> of <G> by <M>. This is the normal
subgroup of <E> which corresponds to <M>.

\beginexample
gap> G := SmallGroup( 4, 2 );;
gap> mats := List( Pcgs( G ), x -> IdentityMat( 1, GF(2) ) );;
gap> M := GModuleByMats( mats, GF(2) );;
gap> co := TwoCocycles( G, M );;
gap> Extension( G, M, co[2] );
<pc group of size 8 with 3 generators>
gap> SplitExtension( G, M );
<pc group of size 8 with 3 generators>
gap> Extensions( G, M );
[ <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators> ]
gap> List(last, IdSmallGroup);
[ [ 8, 5 ], [ 8, 2 ], [ 8, 3 ], [ 8, 3 ], [ 8, 2 ], [ 8, 2 ], [ 8, 3 ], 
  [ 8, 4 ] ]
\endexample

Note that the extensions returned by `Extensions' are computed up to
equivalence, but not up to isomorphism.

There exists an action of the subgroup of *compatible pairs* in
$Aut(G) \times Aut(M)$ which acts on the second cohomology group.
2-cocycles which lie in the same orbit under this action define
isomorphic extensions of $G$. However, there may be isomorphic
extensions of $G$ corresponding to cocycles in different orbits.

\>CompatiblePairs( <G>, <M> [, <D>] ) F

returns the group of compatible pairs of the group <G> with the 
<G>-module <M> as subgroup of the direct product of <Aut(G)> x <Aut(M)>.
Here <Aut(M)> is considered as subgroup of a general linear group. The 
optional argument <D> should be a subgroup of <Aut(G)> x <Aut(M)>. If it
is given, then only the compatible pairs in <D> are computed.

\>ExtensionRepresentatives( <G>, <M>, <P> ) O

returns all extensions of <G> by the <G>-module <M> up to equivalence 
under action of <P> where <P> has to be a subgroup of the group of 
compatible pairs of <G> with <M>.


\beginexample
gap> G := SmallGroup( 4, 2 );;
gap> mats := List( Pcgs( G ), x -> IdentityMat( 1, GF(2) ) );;
gap> M := GModuleByMats( mats, GF(2) );;
gap> A := AutomorphismGroup( G );;
gap> B := GL( 1, 2 );;
gap> D := DirectProduct( A, B );
<group of size 6 with 4 generators>
gap> P := CompatiblePairs( G, M, D );
<group of size 6 with 2 generators>
gap> ExtensionRepresentatives( G, M, P );
[ <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators> ]
gap> Extensions( G, M );
[ <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators>, 
  <pc group of size 8 with 3 generators> ]
\endexample

See also the forthcoming {\GAP} package on `Group Construction Methods'.

Finally we note that for the computation of split extensions it is not
necessary that <M> must correspond to an elementary abelian group. Here
it is possible to construct split extensions of arbitrary pc groups.

\>SplitExtensions( <G>, <aut>, <N> )

returns the split extensions of the pc group <G> by the pc group <N>.
<aut> should be a homomorphism from <G> into $Aut(N)$.

In the following example we construct the holomorph of $Q_8$ as split
extension of $Q_8$ by $S_4$.

\beginexample
gap> N := SmallGroup( 8, 4 );
<pc group of size 8 with 3 generators>
gap> IsAbelian( N );
false
gap> A := AutomorphismGroup( N );
<group of size 24 with 4 generators>
gap> iso := IsomorphismPcGroup( A );
CompositionMapping( Pcgs([ (2,6,5,3), (1,3,5)(2,4,6), (2,5)(3,6), (1,4)(3,6) 
 ]) -> [ f1, f2, f3, f4 ], <action isomorphism> )
gap> H := Image( iso );
Group([ f1, f2, f3, f4 ])
gap> G := Subgroup( H, Pcgs(H){[1,2]} );
Group([ f1, f2 ])
gap> inv := InverseGeneralMapping( iso );
[ f1*f2, f2^2*f3, f4, f3 ] -> [ Pcgs([ f1, f2, f3 ]) -> [ f1*f2, f2, f3 ], 
  Pcgs([ f1, f2, f3 ]) -> [ f2, f1*f2, f3 ], 
  Pcgs([ f1, f2, f3 ]) -> [ f1*f3, f2, f3 ], 
  Pcgs([ f1, f2, f3 ]) -> [ f1, f2*f3, f3 ] ]
gap> K := SplitExtension( G, inv, N );
<pc group of size 192 with 7 generators>
\endexample

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Coding a Pc Presentation}

If one wants to store a large number of pc groups, then it can be useful
to store them in a compressed format, since pc presentations can be
space consuming. Here we introduce a method to code and decode pc
presentations by integers. To decode a given code the size of the
underlying pc group is needed as well. For the full definition and
the coding and decoding
procedures see \cite{BescheEick98}. This method is used with the small
groups library, see Section "Small Groups".

\>CodePcgs( <pcgs> ) F

returns the code corresponding to <pcgs>.

\>CodePcGroup( <G> ) F

returns the code for a pcgs of <G>.

\>PcGroupCode( <code>, <size> ) F

returns a pc group of size <size> corresponding to <code>.
The argument <code> must be a valid code for a pcgs, otherwise anything
may happen. Valid codes are usually obtained by one of the functions
`CodePcgs' or `CodePcGroup'.

\>PcGroupCodeRec( <rec> ) F

Here <rec> needs to have entries .code and .order. Then PcGroupCode 
returns a pc group of size .order corresponding to .code.


\beginexample
gap> G := SmallGroup( 24, 12 );;
gap> p := Pcgs( G );;
gap> code := CodePcgs( p );
5790338948
gap> H := PcGroupCode( code, 24 );
<pc group of size 24 with 4 generators>
gap> map := GroupHomomorphismByImages( G, H, p, FamilyPcgs(H) );
Pcgs([ f1, f2, f3, f4 ]) -> Pcgs([ f1, f2, f3, f4 ])
gap> IsBijective(map);
true
\endexample

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\Section{Random Isomorphism Testing}

The generic isomorphism test for groups may be applied to pc groups
as well. However, this test is often quite time consuming. Here we
describe another method to test isomorphism by a probabilistic approach.

This method takes a list of groups and a non-negative integer as
input. The output is a sublist of the input list where only isomorphic
copies have been removed. The integer gives a certain amount of control
over the probability to detect all isomorphisms. If it is 0, then nothing
will be done at all. The larger the integer is, the larger is the probability
of finding all isomorphisms. However, due to the underlying method we can
not guarantee that the algorithm finds all isomorphisms, no matter how
large $n$ is.

\>RandomIsomorphismTest( <list>, <n> ) F

<list> must be a list of code records of pc groups and <n> a non-negative 
integer. Returns a sublist of <list> where isomorphic copies detected by 
the probabilistic test have been removed.