Records are next to lists the most important way to collect objects together. A record is a collection of components. Each component has a unique name, which is an identifier that distinguishes this component, and a value, which is an object of arbitrary type. We often abbreviate value of a component to element. We also say that a record contains its elements. You can access and change the elements of a record using its name.
Record literals are written by writing down the components in order
between rec(
and )
, and separating them by commas ,
. Each
component consists of the name, the assignment operator :=
, and the
value. The empty record, i.e., the record with no components, is
written as rec()
.
gap> rec( a := 1, b := "2" ); # a record with two components rec( a := 1, b := "2" ) gap> rec( a := 1, b := rec( c := 2 ) ); # record may contain records rec( a := 1, b := rec( c := 2 ) )
Records usually contain elements of various types, i.e., they are usually not homogeneous like lists.
The first section in this chapter tells you how you can access the elements of a record (see Accessing Record Elements).
The next sections tell you how you can change the elements of a record (see Record Assignment and Identical Records).
The next sections describe the operations that are available for records (see Comparisons of Records, Operations for Records, In for Records, and Printing of Records).
The next section describes the function that tests if an object is a record (see IsRec).
The next sections describe the functions that test whether a record has a component with a given name, and delete such a component (see IsBound and Unbind). Those functions are also applicable to lists (see chapter Lists).
The final sections describe the functions that create a copy of a record (see Copy and ShallowCopy). Again those functions are also applicable to lists (see chapter Lists).
rec.name
The above construct evaluates to the value of the record component with the name name in the record rec. Note that the name is not evaluated, i.e., it is taken literal.
gap> r := rec( a := 1, b := 2 );; gap> r.a; 1 gap> r.b; 2
rec.(name)
This construct is similar to the above construct. The difference is that the second operand name is evaluated. It must evaluate to a string or an integer otherwise an error is signalled. The construct then evaluates to the element of the record rec whose name is, as a string, equal to name.
gap> old := rec( a := 1, b := 2 );; gap> new := rec(); rec( ) gap> for i in RecFields( old ) do > new.(i) := old.(i); > od; gap> new; rec( a := 1, b := 2 )
If rec does not evaluate to a record, or if name does not evaluate to
a string, or if rec.name
is unbound, an error is signalled. As
usual you can leave the break loop (see Break Loops) with quit;
. On
the other hand you can return a result to be used in place of the record
element by return expr;
.
rec.name := obj;
The record assignment assigns the object obj, which may be an object of arbitrary type, to the record component with the name name, which must be an identifier, of the record rec. That means that accessing the element with name name of the record rec will return obj after this assignment. If the record rec has no component with the name name, the record is automatically extended to make room for the new component.
gap> r := rec( a := 1, b := 2 );; gap> r.a := 10;; r; rec( a := 10, b := 2 ) gap> r.c := 3;; r; rec( a := 10, b := 2, c := 3 )
The function IsBound
(see IsBound) can be used to test if a record
has a component with a certain name, the function Unbind
(see Unbind)
can be used to remove a component with a certain name again.
Note that assigning to a record changes the record. The ability to change an object is only available for lists and records (see Identical Records).
rec.(name) := obj;
This construct is similar to the above construct. The difference is that the second operand name is evaluated. It must evaluate to a string or an integer otherwise an error is signalled. The construct then assigns obj to the record component of the record rec whose name is, as a string, equal to name.
If rec does not evaluate to a record, name does not evaluate to a
string, or obj is a call to a function that does not return a value,
e.g., Print
(see Print), an error is signalled. As usual you can
leave the break loop (see Break Loops) with quit;
. On the other hand
you can continue the assignment by returning a record in the first case,
a string in the second, or an object to be assigned in the third, using
return expr;
.
With the record assignment (see Record Assignment) it is possible to change a record. The ability to change an object is only available for lists and records. This section describes the semantic consequences of this fact.
You may think that in the following example the second assignment changes the integer, and that therefore the above sentence, which claimed that only records and lists can be changed, is wrong.
i := 3; i := i + 1;
But in this example not the integer 3
is changed by adding one to it.
Instead the variable i
is changed by assigning the value of i+1
,
which happens to be 4
, to i
. The same thing happens in the following
example
r := rec( a := 1 ); r := rec( a := 1, b := 2 );
The second assignment does not change the first record, instead it
assigns a new record to the variable r
. On the other hand, in the
following example the record is changed by the second assignment.
r := rec( a := 1 ); r.b := 2;
To understand the difference first think of a variable as a name for an
object. The important point is that a record can have several names at
the same time. An assignment var := record;
means in this
interpretation that var is a name for the object record. At the end
of the following example r2
still has the value rec( a := 1 )
as
this record has not been changed and nothing else has been assigned to
r2
.
r1 := rec( a := 1 ); r2 := r1; r1 := rec( a := 1, b := 2 );
But after the following example the record for which r2
is a name has
been changed and thus the value of r2
is now rec( a := 1, b := 2 )
.
r1 := rec( a := 1 ); r2 := r1; r1.b := 2;
We shall say that two records are identical if changing one of them by a record assignment also changes the other one. This is slightly incorrect, because if two records are identical, there are actually only two names for one record. However, the correct usage would be very awkward and would only add to the confusion. Note that two identical records must be equal, because there is only one records with two different names. Thus identity is an equivalence relation that is a refinement of equality.
Let us now consider under which circumstances two records are identical.
If you enter a record literal then the record denoted by this literal is
a new record that is not identical to any other record. Thus in the
following example r1
and r2
are not identical, though they are equal
of course.
r1 := rec( a := 1 ); r2 := rec( a := 1 );
Also in the following example, no records in the list l
are identical.
l := []; for i in [1..10] do l[i] := rec( a := 1 ); od;
If you assign a record to a variable no new record is created. Thus the
record value of the variable on the left hand side and the record on the
right hand side of the assignment are identical. So in the following
example r1
and r2
are identical records.
r1 := rec( a := 1 ); r2 := r1;
If you pass a record as argument, the old record and the argument of the
function are identical. Also if you return a record from a function, the
old record and the value of the function call are identical. So in the
following example r1
and r2
are identical record
r1 := rec( a := 1 ); f := function ( r ) return r; end; r2 := f( r1 );
The functions Copy
and ShallowCopy
(see Copy and ShallowCopy)
accept a record and return a new record that is equal to the old record
but that is not identical to the old record. The difference between
Copy
and ShallowCopy
is that in the case of ShallowCopy
the
corresponding elements of the new and the old records will be identical,
whereas in the case of Copy
they will only be equal. So in the
following example r1
and r2
are not identical records.
r1 := rec( a := 1 ); r2 := Copy( r1 );
If you change a record it keeps its identity. Thus if two records are
identical and you change one of them, you also change the other, and they
are still identical afterwards. On the other hand, two records that are
not identical will never become identical if you change one of them. So
in the following example both r1
and r2
are changed, and are still
identical.
r1 := rec( a := 1 ); r2 := r1; r1.b := 2;
rec1 = rec2
rec1 <> rec2
The equality operator =
returns true
if the record rec1 is equal to
the record rec2 and false
otherwise. The inequality operator <>
returns true
if the record rec1 is not equal to rec2 and false
otherwise.
Usually two records are considered equal, if for each component of one record the other record has a component of the same name with an equal value and vice versa. You can also compare records with other objects, they are of course different, unless the record has a special comparison function (see below) that says otherwise.
gap> rec( a := 1, b := 2 ) = rec( b := 2, a := 1 ); true gap> rec( a := 1, b := 2 ) = rec( a := 2, b := 1 ); false gap> rec( a := 1 ) = rec( a := 1, b := 2 ); false gap> rec( a := 1 ) = 1; false
However a record may contain a special operations
record that contains
a function that is called when this record is an operand of a comparison.
The precise mechanism is as follows. If the operand of the equality
operator =
is a record, and if this record has an element with the name
operations
that is a record, and if this record has an element with the
name =
that is a function, then this function is called with the
operands of =
as arguments, and the value of the operation is the
result returned by this function. In this case a record may also be
equal to an object of another type if this function says so. It is
probably not a good idea to define a comparison function in such a way
that the resulting relation is not an equivalence relation, i.e., not
reflexive, symmetric, and transitive. Note that there is no
corresponding <>
function, because left <> right
is implemented
as not left = right
.
The following example shows one piece of the definition of residue
classes, using record operations. Of course this is far from a complete
implementation (see About Defining New Group Elements). Note that the
=
must be quoted, so that it is taken as an identifier (see
Identifiers).
gap> ResidueOps := rec( );; gap> ResidueOps.\= := function ( l, r ) > return (l.modulus = r.modulus) > and (l.representative-r.representative) mod l.modulus = 0; > end;; gap> Residue := function ( representative, modulus ) > return rec( > representative := representative, > modulus := modulus, > operations := ResidueOps ); > end;; gap> l := Residue( 13, 23 );; gap> r := Residue( -10, 23 );; gap> l = r; true gap> l = Residue( 10, 23 ); false
rec1 < rec2
rec1 <= rec2
rec1 > rec2
rec1 >= rec2
The operators <
, <=
, >
, and >=
evaluate to true
if the record
rec1 is less than, less than or equal to, greater than, and greater
than or equal to the record rec2, and to false
otherwise.
To compare records we imagine that the components of both records are
sorted according to their names. Then the records are compared
lexicographically with unbound elements considered smaller than anything
else. Precisely one record rec1 is considered less than another record
rec2 if rec2 has a component with name name2 and either rec1 has
no component with this name or rec1.name2 < rec2.name2
and for
each component of rec1 with name name1 < name2
rec2 has a
component with this name and rec1.name1 = rec2.name1
. Records
may also be compared with objects of other types, they are greater than
anything else, unless the record has a special comparison function (see
below) that says otherwise.
gap> rec( a := 1, b := 2 ) < rec( b := 2, a := 1 ); false # they are equal gap> rec( a := 1, b := 2 ) < rec( a := 2, b := 0 ); true # thea
elements are compared first and 1 is less than 2 gap> rec( a := 1 ) < rec( a := 1, b := 2 ); true # unbound is less than 2 gap> rec( a := 1 ) < rec( a := 0, b := 2 ); false # thea
elements are compared first and 0 is less than 1 gap> rec( b := 1 ) < rec( b := 0, a := 2 ); true # thea
-s are compared first and unbound is less than 2 gap> rec( a := 1 ) < 1; false # other objects are less than records
However a record may contain a special operations
record that contains
a function that is called when this record is an operand of a comparison.
The precise mechanism is as follows. If the operand of the equality
operator <
is a record, and if this record has an element with the
name operations
that is a record, and if this record has an element
with the name <
that is a function, then this function is called with
the operands of <
as arguments, and the value of the operation is the
result returned by this function. In this case a record may also be
smaller than an object of another type if this function says so. It is
probably not a good idea to define a comparison function in such a way
that the resulting relation is not an ordering relation, i.e., not
antisymmetric, and transitive. Note that there are no corresponding
<=
, >
, and >=
functions, since those operations are implemented as
not right < left
, right < left
, and not left < right
respectively.
The following example shows one piece of the definition of residue
classes, using record operations. Of course this is far from a complete
implementation (see About Defining New Group Elements). Note that the
<
must be quoted, so that it is taken as an identifier (see
Identifiers).
gap> ResidueOps := rec( );; gap> ResidueOps.\< := function ( l, r ) > if l.modulus <> r.modulus then > Error("<l> and <r> must have the same modulus"); > fi; > return l.representative mod l.modulus > < r.representative mod r.modulus; > end;; gap> Residue := function ( representative, modulus ) > return rec( > representative := representative, > modulus := modulus, > operations := ResidueOps ); > end;; gap> l := Residue( 13, 23 );; gap> r := Residue( -1, 23 );; gap> l < r; true # 13 is less than 22 gap> l < Residue( 10, 23 ); false # 10 is less than 13
Usually no operations are defined for record. However a record may
contain a special operations
record that contains functions that are
called when this record is an operand of a binary operation. This
mechanism is detailed below for the addition.
obj + rec
, rec + obj
If either operand is a record, and if this record contains an element
with name operations
that is a record, and if this record in turn
contains an element with the name +
that is a function, then this
function is called with the two operands as arguments, and the value of
the addition is the value returned by that function. If both operands
are records with such a function rec.operations.+
, then the function
of the right operand is called. If either operand is a record, but
neither operand has such a function rec.operations.+
, an error is
signalled.
obj - rec
, rec - obj
obj * rec
, rec * obj
obj / rec
, rec / obj
obj mod rec
, rec mod obj
obj ^ rec
, rec ^ obj
This is evaluated similar, but the functions must obviously be called
-
, *
, /
, mod
, ^
respectively.
The following example shows one piece of the definition of a residue
classes, using record operations. Of course this is far from a complete
implementation (see About Defining New Group Elements). Note that the
*
must be quoted, so that it is taken as an identifier (see
Identifiers).
gap> ResidueOps := rec( );; gap> ResidueOps.\* := function ( l, r ) > if l.modulus <> r.modulus then > Error("<l> and <r> must have the same modulus"); > fi; > return rec( > representative := (l.representative * r.representative) > mod l.modulus, > modulus := l.modulus, > operations := ResidueOps ); > end;; gap> Residue := function ( representative, modulus ) > return rec( > representative := representative, > modulus := modulus, > operations := ResidueOps ); > end;; gap> l := Residue( 13, 23 );; gap> r := Residue( -1, 23 );; gap> s := l * r; rec( representative := 10, modulus := 23, operations := rec( \* := function ( l, r ) ... end ) )
element in rec
Usually the membership test is only defined for lists. However a record
may contain a special operations
record, that contains a function that
is called when this record is the right operand of the in
operator.
The precise mechanism is as follows.
If the right operand of the in
operator is a record, and if this record
contains an element with the name operations
that is a record, and if
this record in turn contains an element with the name in
that is a
function, then this function is called with the two operands as
arguments, and the value of the membership test is the value returned by
that function. The function should of course return true
or false
.
The following example shows one piece of the definition of residue
classes, using record operations. Of course this is far from a complete
implementation (see About Defining New Group Elements). Note that the
in
must be quoted, so that it is taken as an identifier (see
Identifiers).
gap> ResidueOps := rec( );; gap> ResidueOps.\in := function ( l, r ) > if IsInt( l ) then > return (l - r.representative) mod r.modulus = 0; > else > return false; > fi; > end;; gap> Residue:= function ( representative, modulus ) > return rec( > representative := representative, > modulus := modulus, > operations := ResidueOps ); > end;; gap> l := Residue( 13, 23 );; gap> -10 in l; true gap> 10 in l; false
Print( rec )
If a record is printed by Print
(see Print, PrintTo, and
AppendTo) or by the main loop (see Main Loop), it is usually printed
as record literal, i.e., as a sequence of components, each in the format
name := value
, separated by commas and enclosed in rec(
and )
.
gap> r := rec();; r.a := 1;; r.b := 2;; gap> r; rec( a := 1, b := 2 )
But if the record has an element with the name operations
that is a
record, and if this record has an element with the name Print
that is a
function, then this function is called with the record as argument. This
function must print whatever the printed representation of the record
should look like.
The following example shows one piece of the definition of residue classes, using record operations. Of course this is far from a complete implementation (see About Defining New Group Elements). Note that it is typical for records that mimic group elements to print as a function call that, when evaluated, will create this group element record.
gap> ResidueOps := rec( );; gap> ResidueOps.Print := function ( r ) > Print( "Residue( ", > r.representative mod r.modulus, ", ", > r.modulus, " )" ); > end;; gap> Residue := function ( representative, modulus ) > return rec( > representative := representative, > modulus := modulus, > operations := ResidueOps ); > end;; gap> l := Residue( 33, 23 ); Residue( 10, 23 )
IsRec( obj )
IsRec
returns true
if the object obj, which may be an object of
arbitrary type, is a record, and false
otherwise. Will signal an error
if obj is a variable with no assigned value.
gap> IsRec( rec( a := 1, b := 2 ) ); true gap> IsRec( IsRec ); false
IsBound( rec.name )
IsBound( list[n] )
In the first form IsBound
returns true
if the record rec has a
component with the name name, which must be an ident and false
otherwise. rec must evaluate to a record, otherwise an error is
signalled.
In the second form IsBound
returns true
if the list list has a
element at the position n, and false
otherwise. list must evaluate
to a list, otherwise an error is signalled.
gap> r := rec( a := 1, b := 2 );; gap> IsBound( r.a ); true gap> IsBound( r.c ); false gap> l := [ , 2, 3, , 5, , 7, , , , 11 ];; gap> IsBound( l[7] ); true gap> IsBound( l[4] ); false gap> IsBound( l[101] ); false
Note that IsBound
is special in that it does not evaluate its argument,
otherwise it would always signal an error when it is supposed to return
false
.
Unbind( rec.name )
Unbind( list[n] )
In the first form Unbind
deletes the component with the name name in
the record rec. That is, after execution of Unbind
, rec no longer
has a record component with this name. Note that it is not an error to
unbind a nonexisting record component. rec must evaluate to a record,
otherwise an error is signalled.
In the second form Unbind
deletes the element at the position n in
the list list. That is, after execution of Unbind
, list no longer
has an assigned value at the position n. Note that it is not an error
to unbind a nonexisting list element. list must evaluate to a list,
otherwise an error is signalled.
gap> r := rec( a := 1, b := 2 );; gap> Unbind( r.a ); r; rec( b := 2 ) gap> Unbind( r.c ); r; rec( b := 2 ) gap> l := [ , 2, 3, 5, , 7, , , , 11 ];; gap> Unbind( l[3] ); l; [ , 2,, 5,, 7,,,, 11 ] gap> Unbind( l[4] ); l; [ , 2,,,, 7,,,, 11 ]
Note that Unbind
does not evaluate its argument, otherwise there would
be no way for Unbind
to tell which component to remove in which record,
because it would only receive the value of this component.
Copy( obj )
Copy
returns a copy new of the object obj. You may apply Copy
to
objects of any type, but for objects that are not lists or records Copy
simply returns the object itself.
For lists and records the result is a new list or record that is not identical to any other list or record (see Identical Lists and Identical Records). This means that you may modify this copy new by assignments (see List Assignment and Record Assignment) or by adding elements to it (see Add and Append), without modifying the original object obj.
gap> list1 := [ 1, 2, 3 ];; gap> list2 := Copy( list1 ); [ 1, 2, 3 ] gap> list2[1] := 0;; list2; [ 0, 2, 3 ] gap> list1; [ 1, 2, 3 ]
That Copy
returns the object itself if it is not a list or a record is
consistent with this definition, since there is no way to change the
original object obj by modifying new, because in fact there is no way
to change the object new.
Copy
basically executes the following code for lists, and similar code
for records.
new := []; for i in [1..Length(obj)] do if IsBound(obj[i]) then new[i] := Copy( obj[i] ); fi; od;
Note that Copy
recursively copies all elements of the object obj. If
you only want to copy the top level use ShallowCopy
(see
ShallowCopy).
gap> list1 := [ [ 1, 2 ], [ 3, 4 ] ];; gap> list2 := Copy( list1 ); [ [ 1, 2 ], [ 3, 4 ] ] gap> list2[1][1] := 0;; list2; [ [ 0, 2 ], [ 3, 4 ] ] gap> list1; [ [ 1, 2 ], [ 3, 4 ] ]
The above code is not entirely correct. If the object obj contains a list or record twice this list or record is not copied twice, as would happen with the above definition, but only once. This means that the copy new and the object obj have exactly the same structure when view as a general graph.
gap> sub := [ 1, 2 ];; list1 := [ sub, sub ];; gap> list2 := Copy( list1 ); [ [ 1, 2 ], [ 1, 2 ] ] gap> list2[1][1] := 0;; list2; [ [ 0, 2 ], [ 0, 2 ] ] gap> list1; [ [ 1, 2 ], [ 1, 2 ] ]
ShallowCopy( obj )
ShallowCopy
returns a copy of the object obj. You may apply
ShallowCopy
to objects of any type, but for objects that are not lists
or records ShallowCopy
simply returns the object itself.
For lists and records the result is a new list or record that is not identical to any other list or record (see Identical Lists and Identical Records). This means that you may modify this copy new by assignments (see List Assignment and Record Assignment) or by adding elements to it (see Add and Append), without modifying the original object obj.
gap> list1 := [ 1, 2, 3 ];; gap> list2 := ShallowCopy( list1 ); [ 1, 2, 3 ] gap> list2[1] := 0;; list2; [ 0, 2, 3 ] gap> list1; [ 1, 2, 3 ]
That ShallowCopy
returns the object itself if it is not a list or a
record is consistent with this definition, since there is no way to
change the original object obj by modifying new, because in fact
there is no way to change the object new.
ShallowCopy
basically executes the following code for lists, and
similar code for records.
new := []; for i in [1..Length(obj)] do if IsBound(obj[i]) then new[i] := obj[i]; fi; od;
Note that ShallowCopy
only copies the top level. The subobjects of the
new object new are identical to the corresponding subobjects of the
object obj. If you want to copy recursively use Copy
(see Copy).
RecFields( rec )
RecFields
returns a list of strings corresponding to the names of the
record components of the record rec.
gap> r := rec( a := 1, b := 2 );; gap> RecFields( r ); [ "a", "b" ]
Note that you cannot use the string result in the ordinary way to access
or change a record component. You must use the rec.(name)
construct (see Accessing Record Elements and Record Assignment).
gap3-jm