Attributes and Methods
Overview
Teaching: 40 min
Exercises: 10 minQuestions
How to record information in GAP objects
Objectives
Declaring an attribute
Installing a method
Understanding method selection
Using debugging tools
Which function is faster?
Try to repeatedly calculate
AvgOrdOfGroup(M11)
andAvgOrdOfCollection(M11)
and compare runtimes. Do this for a new copy ofM11
and for the one for which this parameter has already been observed. What do you observe?
Of course, for any given group the average order of its elements needs to
be calculated only once, as the next time it will return the same value.
However, as we see from the runtimes below, each new call of AvgOrdOfGroup
will repeat the same computation again, with slightly varying runtime:
A:=AlternatingGroup(10);
Alt( [ 1 .. 10 ] )
AvgOrdOfCollection(A); time; AvgOrdOfCollection(A); time;
2587393/259200
8226
2587393/259200
8118
In the last example, the group in question was the same – we haven’t
constructed another copy of AlternatingGroup(10)
; however, the result
of the calculation was not stored in A
.
If you need to reuse this value, one option could be to store it in some variable, but then you should be careful about matching such variables with corresponding groups, and the code could become quite convoluted and unreadable. On the other hand, GAP has the notion of an attribute – a data structure that is used to accumulate information that an object learns about itself during its lifetime. Consider the following example:
G:=Group([ (1,2,3,4,5,6,7,8,9,10,11), (3,7,11,8)(4,10,5,6) ]);
gap> NrConjugacyClasses(G);time;NrConjugacyClasses(G);time;
Group([ (1,2,3,4,5,6,7,8,9,10,11), (3,7,11,8)(4,10,5,6) ])
10
39
10
0
In this case, the group G
has 10 conjugacy classes, and it took 39 ms to
establish that in the first call. The second call has zero cost since the
result was stored in G
, since NrConjugacyClasses
is an attribute:
NrConjugacyClasses;
<Attribute "NrConjugacyClasses">
Our goal is now to learn how to create own attributes.
Since we already have a function AvgOrdOfCollection
which
does the calculation, the simplest way to turn it into
an attribute is as follows:
AverageOrder := NewAttribute("AverageOrder", IsCollection);
InstallMethod( AverageOrder, "for a collection", [IsCollection], AvgOrdOfCollection);
In this example, first we declared an attribute AverageOrder
for
objects in the category IsCollection
, and then installed the function
AvgOrdOfCollection
as a method for this attribute. Instead of calling
the function AvgOrdOfCollection
, we may now call AverageOrder
.
Now we may check that subsequent calls of AverageOrder
with the same argument
are performed at zero cost. In this example the time is reduced from more than
16 seconds to zero:
S:=SymmetricGroup(10);; AverageOrder(S); time; AverageOrder(S); time;
39020911/3628800
16445
39020911/3628800
0
You may wonder why we have declared the operation for a collection and not only
for a group, and why we have installed the inefficient AvgOrdOfCollection
.
After all, we have already developed the much more efficient AvgOrdOfGroup
.
Imagine that you would like to be able to compute an average order both for a group and for a list which consists of objects having a multiplicative order. You may have a special function for each case, as we have. If it could happen that you don’t know in advance the type of the object in question, you may add checks into the code and dispatch to a suitable function. This could quickly become complicated if you have several different functions for various types of objects. Instead of that, attributes are bunches of functions, called methods, and GAP’s method selection will choose the most efficient method based on the type of all arguments.
To illustrate this, we will now install a method for AverageOrder
for a group:
InstallMethod( AverageOrder, [IsGroup], AvgOrdOfGroup);
If you apply it to a group whose AverageOrder
has already been computed, nothing
will happen, since GAP will use the stored value. However, for a newly created group,
this new method will be called:
S:=SymmetricGroup(10);; AverageOrder(S); time; AverageOrder(S); time;
39020911/3628800
26
39020911/3628800
0
Which method is being called
Try to call
AverageOrder
for a collection which is not a group (a list of group elements and/or a conjugacy class of group elements).Debugging tools like
TraceMethods
may help you see which method is being called.
ApplicableMethod
in combination withPageSource
may point you to the source code with all the comments.
A property is a boolean-valued attribute. It can be created using NewProperty
IsIntegerAverageOrder := NewProperty("IsIntegerAverageOrder", IsCollection);
Now we will install a method for IsIntegerAverageOrder
for a collection.
Observe that it is never necessary to create
a function first and then install it as a method. The following method installation
instead creates a new function as one of its arguments:
InstallMethod( IsIntegerAverageOrder,
"for a collection",
[IsCollection],
coll -> IsInt( AverageOrder( coll ) )
);
Note that because AverageOrder
is an attribute it will take care of the selection of
the most suitable method.
Does such a method always exist?
No. “No-method-found” is a special kind of error, and there are tools to investigate such errors: see
?ShowArguments
,?ShowDetails
,?ShowMethods
and?ShowOtherMethods
.
The following calculation shows that despite our success with calculating the average order for large permutation groups via conjugacy classes of elements, for pc groups from the Small Groups Library it could be faster to iterate over their elements than to calculate conjugacy classes:
l:=List([1..1000],i->SmallGroup(1536,i));; List(l,AvgOrdOfGroup);;time;
56231
l:=List([1..1000],i->SmallGroup(1536,i));; List(l,AvgOrdOfCollection);;time;
9141
Don’t panic!
Install a method for
IsPcGroup
that iterates over the group elements instead of calculations its conjugacy classes.Estimate practical boundaries of its feasibility. Can you find an example of a pc group where iterating is slower than calculating conjugacy classes?
Key Points
Positional objects may accumulate information about themselves during their lifetime.
This means that next time the stored information may be retrieved at zero cost.
Methods are bunches of functions; GAP’s method selection will choose the most efficient method based on the type of all arguments.
‘No-method-found’ is a special kind of error with useful debugging tools helping to understand it.