6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
6.001 Notes,Section 2.1
Slide 2.1.1
In the last lecture,we began looking at the programming
language,Scheme,with the intent of learning how that
language would provide a basis for describing procedures and
processes,and thus for understanding computational metaphors
for controlling complex processes,In this lecture,we look at
how to create procedural abstractions in our language,and how
to use those abstractions to describe and capture computational
processes,
Slide 2.1.2
Well -- we've got primitives (numbers and built in procedures);
we've got means of combination (ways of creating complex
expressions); and we've got our first means of abstraction (a
way of giving a name to something),But we are still stuck just
writing out arithmetic expressions,as the only procedures we
have are the built in ones,We need a way to capture our own
processes in our own procedures,So we need another kind of
abstraction -- we need a way of capturing particular processes
in our own procedures,and that's what we turn to next,
Slide 2.1.3
In Scheme,we have a particular expression for capturing a
procedure,It's called a lambda expression,It has the form
shown,an open parenthesis,the keyword lambda,followed by
some number of symbols enclosed within parentheses,followed
by one or more legal expressions,followed by a close
parenthesis,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.4
The keyword lambda identifies this expression as a particular
special form,The set of symbols immediately following the
lambda are called the formal parameters of the lambda,in this
case,there is just one parameter,x,
Slide 2.1.5
The subsequent expression we refer to as the body of the
procedure,This is the particular pattern we are going to use to
capture a process,
Slide 2.1.6
The way to think about this lambda expression is that it is going
to capture a common pattern of computation in a procedure -- it
will actually build a procedure for us,
The way to read the expression is,To process,.,
Slide 2.1.7
..,something,.,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.8
..,multiply it by itself,and return that value,
Slide 2.1.9
So in fact this particular lambda expression captures the process
of "squaring",It says,if you give me a value for x I will return
the value of multiplying that value by itself,In a second we will
see how this happens,
Notice that lambda expressions must be special forms,The
normal rules for evaluating a combination do not apply here,
Instead,the value returned by evaluating a lambda expression is
the actual procedure that it captures,Contained within that
procedure will be a set of formal parameters,and a body that
captures the common pattern of the process,as a function of
those formal parameters,
Slide 2.1.10
Now,where can we use such a procedure? Basically anywhere
in our earlier expressions that we could use a built-in
procedure,which for now means as the first element in a
combination,
For example,here is a compound expression,with two
subexpressions,What is the value or meaning associated with
it? The value of the first subexpression we just saw was a
procedure,The value of the second subexpression is just the
number 5,Now we have something similar to our earlier cases,
a procedure applied to a value,The only difference is that here
we have a procedure we built,rather than a pre-existing one,
We need to specify how such a procedure is applied to a set of
arguments,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.11
So here is a summary of our earlier rules for evaluating
expressions,The only change is to amplify what it means to
apply a procedure to a set of arguments,When the procedure
was a built-in arithmetic operator,we just did the obvious thing,
Now if the procedure is something built by evaluating a lambda
expression,we have a new rule,We take the body of the
procedure,substitute the value of the argument in place of the
corresponding formal parameter,and then use the same rules to
evaluate the resulting expression,
Slide 2.1.12
So let's go back to our example,Our rule says to replace the
value of the second expression,5,everywhere in the body that
we see the formal parameter,x,This then reduces the
application of the lambda expression to a simpler expression,.,
Slide 2.1.13
This reduces the application of a procedure to this simpler
expression,and we can apply our rules again,The symbol * is
just a name for the built-in multiplication operation,and 5 is
just self-evaluating,so this all reduces to,.,
Slide 2.1.14
..,25,
Thus we see that our rules now cover the evaluation of
compound expressions that include the application of
procedures created by lambda expressions,In particular,the
rules tell us to substitute into the body of a procedure for the
formal parameters,reducing to a new expression,and then
apply the same set of rules all over again,until we reach a final
answer,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.15
Thus,lambda gives us the ability to capture procedure
abstractions -- patterns of computation in a single procedure,
But we don't want to have to write out lambda expressions
everywhere we need to use this procedure,
Instead,we can combine this procedural abstraction with our
naming abstraction -- that is,we can use a define expression to
give a name to a procedure,In this case,the name square will
be paired with the value of the lambda expression,or quite
literally with the actual procedure created by evaluating that
lambda,
Then we can use the name square wherever we want the
procedure,since its value is the actual procedure,If you follow
the rules of evaluation for the last expression,you will see that we get a procedure applied to a number,and the
substitution of the argument into the body of the procedure reduces to a simpler expression,just as we saw earlier,
6.001 Notes,Section 2.2
Slide 2.2.1
The second kind of special form we saw was a lambda
expression,and lambda we said was our way of capturing
processes inside procedures,Lambda is Greek for "procedure
maker",and is our way of saying "Take this pattern,and
capture in a way that will let me reuse it multiple times",In our
two world view,if we type this expression into the computer
and ask it to be evaluated,the computer uses the key word
lambda to determine that this is a particular special form,It
will then use the rule designed for this type of expression,
Slide 2.2.2
It is important to stress that the actual value created inside the
machine is some representation for the procedure itself,This
representation includes information about the parameters of the
procedure,and the body of the procedure,that is,what pattern
of computation is being captured,and what are the variables to
be replaced in that pattern when we want to use this procedure,
But it is the actual procedure representation that is associated as
the value of the expression,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.2.3
And what gets returned? Basically some representation like
this,that identifies that this is a procedure created by evaluating
a lambda,and an indication of where it lives inside the machine
(i.e,how to get at the parameters and body of the procedure),It
is important to stress that evaluating a lambda creates an actual
procedure object inside the execution world,and that value is
actually returned as the value of the expression,
Slide 2.2.4
Now let's look at the interaction between creating procedures
and giving them names,First of all,I can create a lambda
expression,and the value returned by that evaluation is the
procedure itself,Note that the procedure is not executed or run,
there is no numeric value returned here as a square,rather the
procedure itself is returned as the value,
Slide 2.2.5
If I actually want to use that procedure I need to refer to it,and
the easiest way to do this is to give it a name,So I can define
square to be the value returned by that lambda expression,
which we just saw is a procedure,This then creates a pairing of
the name square with the procedure that captures the pattern of
multiplying a value by itself,
Slide 2.2.6
Having done that,I can now write any expression that uses the
name square anyplace that the associated lambda expression
would have been legitimate,
Note that the evaluation rules nicely cover this,This expression
is a standard combination,so we just get the values of the sub-
expressions,Square is a name,so we look up its value,getting
back the procedure,This can then be applied to the value of the
next sub-expression,using the standard rules for procedure
application,namely substitute 4 for x everywhere in the body of
the procedure,then evaluate that new expression using the same
rules,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.2.7
Of course,I could have used the full lambda expression in
place of the name,In this case,evaluation of the first sub-
expression will create the procedure,which is then applied to
the value of the second sub-expression,
Slide 2.2.8
Because this action of creating a procedure and giving it a name
is such a common activity,we have an alternative way of doing
this,This last expression is really the same as the second one,
but is simply written in a more convenient form (which we call
"syntactic sugar" meaning that the meaning is the same but the
form is sweeter),It is important to stress that in this last form,
there is a hidden lambda that is evaluated to create the
procedure,and then the define is used to give it a name,
Slide 2.2.9
Let's look a little more carefully at this lambda special form,
As we have seen,the syntax,as shown at the top,has three
pieces,
There is the key word,lambda,which identifies the type of
expression,The first operand position of this expression
identifies the formal parameters of the procedure,This is a list
(or a sequence of names enclosed in a set of parentheses),In
this case,the procedure takes two arguments,which we will
call x and y,Note that there can be an arbitrary number of
arguments for a procedure,including zero (in which case we
would use () as the parameter list),and that this component
determines how many arguments must be passed to the
procedure when it is applied,
The second operand position determines the body of the procedure,that is,the pattern of evaluation to be used,
This may be any valid Scheme expression,Note that this expression is NOT evaluated when the procedure is
created,It is simply stored away as a pattern,It is ONLY evaluated when the procedure is applied in some legal
combination,
That determines the syntax of the lambda,The semantics of this kind of expression is very important,so we place
it on a whole separate slide,which follows,.,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.2.10
..,and I feel like I should be shouting this out!!
The semantics says that the value of a lambda expression is a
procedure object,It is some internal representation of the
procedure that includes information about the formal
parameters and the body,or pattern of computation,
6.001 Notes,Section 2.3
Slide 2.3.1
We have now seen most of the basic elements of Scheme,We
will continue to add a few more special forms,and introduce
some additional built in procedures,but we now have enough
elements of the language to start reasoning about processes,and
especially to use procedures to describe computational
processes,
So let's look at some examples of describing processes in
procedures,
Slide 2.3.2
First,what does a procedure describe?
One useful way of thinking about this is as a means of
generalizing a common pattern of operations,For example,
consider the three expressions shown here,The first two are
straightforward,The third is a bit more general,since
foobar is presumably a name for some numerical value,
However,each of these is basically just a specific instantiation
of a process,the process of multiplying a value by itself,or the
process of "squaring",
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.3.3
So we can capture this by giving a name to the part of the
pattern that changes with each instantiation; identifying that
name as a formal parameter; and then capturing that pattern as
the body of a lambda expression,together with the set of
formal parameters,all within a lambda expression,
Slide 2.3.4
Now lets consider a more complex pattern,as shown here,
Slide 2.3.5
In this case,there are two things that vary,so we will need two
parameters to capture this,Otherwise we could just do the same
thing we did last time,and replicate the pattern,with the
parameters in place of the things that change,as shown,
Slide 2.3.6
But a better way to capture this pattern is to realize that there
are really two things going on,One is the sub-pattern of
squaring things,The other is the use of the results of two
different squaring operations within the larger pattern,
So we could capture each of those patterns within its own
procedure,with its own parameters and body,
Note that in doing this,we are relying on a property of a
combination,namely that a combination involving a named
procedure is equivalent to the pattern captured by the
procedure,with values substituted for the formal parameters,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.3.7
Why is this a better way of capturing the pattern?
The primary reason is that by breaking the pattern up into
smaller modules,we isolate pieces of the computation in
separate abstractions,And these modules can then be reused by
other computations,In particular,the idea of square is
likely to be of value elsewhere,so it makes sense to capture that
in its own procedure,and then use within this larger one,
As well,by doing this,we create code that is easier to read,as
we use a simple name to capture the idea of square,and
suppress the unnecessary details,By doing so,we isolate out
the use of a procedure from the details of its actual
implementation,a trick to which we will return later in the
term,
Of course,there may be many different ways of modularizing a computational pattern,
Slide 2.3.8
Here is a finer scale modularization of the pattern into
procedures,Note how each procedure uses the previous one
within its body,using that idea of abstraction to separate the use
of a procedure from its details,
Slide 2.3.9
Now,let's step away from the specifics of this example,and
talk about the process we just used,
In essence,we did several things,we identified modules or
parts of the computational process,which we could usefully
isolate; we then captured each of those within their own
procedural abstraction; and finally we created a procedure to
control the interactions between the individual modules,Of
course,we could apply this process within each of the modules,
in a recursive fashion,
Our goal now is to see how we can use this general approach to
capture computational processes in procedures,
6.001 Notes,Section 2.4
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.1
Let's take another look at this idea of capturing a procedural
description with our language for describing processes,namely
lambdas,Remember our description of the process of
finding square roots,from the first lecture,shown here,
Slide 2.4.2
Our first step in building code to compute,square roots” is to
determine some good modules,or stages,within this process,
Here,we can see several,
There is the idea of measuring whether our guess is good
enough that we can stop and return an answer,
There is the idea of creating a new guess if we are not close
enough,
And there will need to be a way of controlling the process,in
which we use our new guess as if it were the original one,and
continue the process,
So let's build each of these abstractions,using our idea of
capturing common patterns within lambda expressions,
Slide 2.4.3
Here is a rather naive way of deciding if a guess is close
enough,We take the guess,and square it,If the guess is good,
then that should give us a value close to the number whose
square root we are seeking,Here abs is a built in procedure
that returns the absolute value of its argument,and we simply
test to see if that absolute difference is small,
Note how we are already using procedural abstractions,we
have assumed that square is an abstraction for the process
of squaring numbers,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.4
For improving the guess,we can just use the same approach we
did with pythagoras,we capture the idea of average and we
use it to improve a guess as described by the process,
Slide 2.4.5
As before,we can see that average is likely to be a process
that we will want to use in other places,so creating an
abstraction allows us to avoid replicating the code in those
places,
Moreover,we stress that by building this abstraction,we seal
off the details of the implementation from the actual use of the
abstraction,For example,we could decide to change the
implementation of average,such as that shown here,
Doing so does not require us to make any changes to
procedures that use average however,since those simply
refer to the procedure,not the internal specifics,
Also note that the names of the parameters are internal to the lambda expression,we cannot refer to them
outside the scope of the lambda,Here,we have changed the names of the parameters,but this does not affect
those procedures that use average,
Slide 2.4.6
The last step is to decide how to integrate these pieces together
into a process that controls the steps of the computation,The
basic idea is already captured in the process description,given a
number and a guess,we want to use improve to derive a
new guess,
But now we have to make a decision,is the guess good enough,
in which case we can stop? Or should we continue the process?
In order to make such a decision,we need a new special form,
called an if expression,which has three sub-expressions,a
predicate,a consequence,and an alternative,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.7
Here is how an if expression is evaluated,First,the evaluator
uses its rules (such as those we have described) to determine
the value of the predicate expression,
If that value is true,then the evaluator uses its rules to evaluate
the consequence expression,and returns that value as the value
of the entire if expression,
On the other hand,if that value is not true,then the evaluator
uses its rules to evaluate the alternative expression,and returns
that value as the value of the entire if expression,
Slide 2.4.8
So why do we say that this is a special form,rather than just a
procedure? We'll let you think about this,but after the next
lecture you should be able to answer this question,For now,we
will simply accept that if expressions are evaluated in this
particular manner,
Slide 2.4.9
So back to our process,We know from our description that the
heart of the process should look something like this,We check
to see if we are close enough,and if so,then we just return the
value of the guess,Our if expression will handle that for us,
If we are not close enough,we want to improve our guess,
using the improve procedural abstraction we built,But
somehow we want to then use that value as a new guess,and
repeat the process,
Slide 2.4.10
How do we do that?
Well,let's call this overall process,of repeated improving a
guess until we are close enough,sqrt-loop,Then if we
are not close enough,we get an improved guess,and simply do
this again!
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.11
Finally,we can assemble our sqrt procedure,we just use
this repeated process of improving a guess,all we need to do is
get it started with some initial guess,
Slide 2.4.12
This may look a bit unusual,We have a procedure that refers to
itself within its body,in a recursive fashion,Can we be sure
that this procedure will correctly evolve,as we saw in the first
lecture?
In the next lecture,we are going to introduce a formal model
for tracing the evaluation of expressions,especially expressions
involving the application of procedures,For now,however,we
can be a bit informal and walk through the steps of the
computation,
As an example,let's suppose we try to find the sqrt of 2,
Slide 2.4.13
Basically,we can replace this expression with the body of the
procedure associated with sqrt,where the formal parameter
is replaced with the specific value,In other words,we reverse
the process of capturing a pattern,so we are going to loop
through the process,using an initial guess of 1,
Slide 2.4.14
Now what does sqrt-loop do? Well,the pattern it
captured was to see if we are close enough,here we have the
body of the sqrt-loop procedure with the specific values
in place,For now,we don't care about the other sub-expressions
of the if expression,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.15
In this case,we are not close enough,so we move to the
alternative stage of the if expression,as shown,
Slide 2.4.16
Now this is just like a normal combination,We reduce the
values of the subexpressions,so we get the value of the
improved guess..,
Slide 2.4.17
… and then repeat the process,We keep doing this until we get
a value that is close enough that we can stop,
Slide 2.4.18
So to summarize,we have seen that we can use the idea of a
procedure to capture a computational process,by finding good
components or modules of the process; capturing each within
its own procedure; and then deciding how to control the overall
process of the computation,In the next lecture,we will return
to this idea,looking at different ways to break a problem down
into these steps,
6.001 Notes,Section 2.1
Slide 2.1.1
In the last lecture,we began looking at the programming
language,Scheme,with the intent of learning how that
language would provide a basis for describing procedures and
processes,and thus for understanding computational metaphors
for controlling complex processes,In this lecture,we look at
how to create procedural abstractions in our language,and how
to use those abstractions to describe and capture computational
processes,
Slide 2.1.2
Well -- we've got primitives (numbers and built in procedures);
we've got means of combination (ways of creating complex
expressions); and we've got our first means of abstraction (a
way of giving a name to something),But we are still stuck just
writing out arithmetic expressions,as the only procedures we
have are the built in ones,We need a way to capture our own
processes in our own procedures,So we need another kind of
abstraction -- we need a way of capturing particular processes
in our own procedures,and that's what we turn to next,
Slide 2.1.3
In Scheme,we have a particular expression for capturing a
procedure,It's called a lambda expression,It has the form
shown,an open parenthesis,the keyword lambda,followed by
some number of symbols enclosed within parentheses,followed
by one or more legal expressions,followed by a close
parenthesis,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.4
The keyword lambda identifies this expression as a particular
special form,The set of symbols immediately following the
lambda are called the formal parameters of the lambda,in this
case,there is just one parameter,x,
Slide 2.1.5
The subsequent expression we refer to as the body of the
procedure,This is the particular pattern we are going to use to
capture a process,
Slide 2.1.6
The way to think about this lambda expression is that it is going
to capture a common pattern of computation in a procedure -- it
will actually build a procedure for us,
The way to read the expression is,To process,.,
Slide 2.1.7
..,something,.,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.8
..,multiply it by itself,and return that value,
Slide 2.1.9
So in fact this particular lambda expression captures the process
of "squaring",It says,if you give me a value for x I will return
the value of multiplying that value by itself,In a second we will
see how this happens,
Notice that lambda expressions must be special forms,The
normal rules for evaluating a combination do not apply here,
Instead,the value returned by evaluating a lambda expression is
the actual procedure that it captures,Contained within that
procedure will be a set of formal parameters,and a body that
captures the common pattern of the process,as a function of
those formal parameters,
Slide 2.1.10
Now,where can we use such a procedure? Basically anywhere
in our earlier expressions that we could use a built-in
procedure,which for now means as the first element in a
combination,
For example,here is a compound expression,with two
subexpressions,What is the value or meaning associated with
it? The value of the first subexpression we just saw was a
procedure,The value of the second subexpression is just the
number 5,Now we have something similar to our earlier cases,
a procedure applied to a value,The only difference is that here
we have a procedure we built,rather than a pre-existing one,
We need to specify how such a procedure is applied to a set of
arguments,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.11
So here is a summary of our earlier rules for evaluating
expressions,The only change is to amplify what it means to
apply a procedure to a set of arguments,When the procedure
was a built-in arithmetic operator,we just did the obvious thing,
Now if the procedure is something built by evaluating a lambda
expression,we have a new rule,We take the body of the
procedure,substitute the value of the argument in place of the
corresponding formal parameter,and then use the same rules to
evaluate the resulting expression,
Slide 2.1.12
So let's go back to our example,Our rule says to replace the
value of the second expression,5,everywhere in the body that
we see the formal parameter,x,This then reduces the
application of the lambda expression to a simpler expression,.,
Slide 2.1.13
This reduces the application of a procedure to this simpler
expression,and we can apply our rules again,The symbol * is
just a name for the built-in multiplication operation,and 5 is
just self-evaluating,so this all reduces to,.,
Slide 2.1.14
..,25,
Thus we see that our rules now cover the evaluation of
compound expressions that include the application of
procedures created by lambda expressions,In particular,the
rules tell us to substitute into the body of a procedure for the
formal parameters,reducing to a new expression,and then
apply the same set of rules all over again,until we reach a final
answer,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.1.15
Thus,lambda gives us the ability to capture procedure
abstractions -- patterns of computation in a single procedure,
But we don't want to have to write out lambda expressions
everywhere we need to use this procedure,
Instead,we can combine this procedural abstraction with our
naming abstraction -- that is,we can use a define expression to
give a name to a procedure,In this case,the name square will
be paired with the value of the lambda expression,or quite
literally with the actual procedure created by evaluating that
lambda,
Then we can use the name square wherever we want the
procedure,since its value is the actual procedure,If you follow
the rules of evaluation for the last expression,you will see that we get a procedure applied to a number,and the
substitution of the argument into the body of the procedure reduces to a simpler expression,just as we saw earlier,
6.001 Notes,Section 2.2
Slide 2.2.1
The second kind of special form we saw was a lambda
expression,and lambda we said was our way of capturing
processes inside procedures,Lambda is Greek for "procedure
maker",and is our way of saying "Take this pattern,and
capture in a way that will let me reuse it multiple times",In our
two world view,if we type this expression into the computer
and ask it to be evaluated,the computer uses the key word
lambda to determine that this is a particular special form,It
will then use the rule designed for this type of expression,
Slide 2.2.2
It is important to stress that the actual value created inside the
machine is some representation for the procedure itself,This
representation includes information about the parameters of the
procedure,and the body of the procedure,that is,what pattern
of computation is being captured,and what are the variables to
be replaced in that pattern when we want to use this procedure,
But it is the actual procedure representation that is associated as
the value of the expression,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.2.3
And what gets returned? Basically some representation like
this,that identifies that this is a procedure created by evaluating
a lambda,and an indication of where it lives inside the machine
(i.e,how to get at the parameters and body of the procedure),It
is important to stress that evaluating a lambda creates an actual
procedure object inside the execution world,and that value is
actually returned as the value of the expression,
Slide 2.2.4
Now let's look at the interaction between creating procedures
and giving them names,First of all,I can create a lambda
expression,and the value returned by that evaluation is the
procedure itself,Note that the procedure is not executed or run,
there is no numeric value returned here as a square,rather the
procedure itself is returned as the value,
Slide 2.2.5
If I actually want to use that procedure I need to refer to it,and
the easiest way to do this is to give it a name,So I can define
square to be the value returned by that lambda expression,
which we just saw is a procedure,This then creates a pairing of
the name square with the procedure that captures the pattern of
multiplying a value by itself,
Slide 2.2.6
Having done that,I can now write any expression that uses the
name square anyplace that the associated lambda expression
would have been legitimate,
Note that the evaluation rules nicely cover this,This expression
is a standard combination,so we just get the values of the sub-
expressions,Square is a name,so we look up its value,getting
back the procedure,This can then be applied to the value of the
next sub-expression,using the standard rules for procedure
application,namely substitute 4 for x everywhere in the body of
the procedure,then evaluate that new expression using the same
rules,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.2.7
Of course,I could have used the full lambda expression in
place of the name,In this case,evaluation of the first sub-
expression will create the procedure,which is then applied to
the value of the second sub-expression,
Slide 2.2.8
Because this action of creating a procedure and giving it a name
is such a common activity,we have an alternative way of doing
this,This last expression is really the same as the second one,
but is simply written in a more convenient form (which we call
"syntactic sugar" meaning that the meaning is the same but the
form is sweeter),It is important to stress that in this last form,
there is a hidden lambda that is evaluated to create the
procedure,and then the define is used to give it a name,
Slide 2.2.9
Let's look a little more carefully at this lambda special form,
As we have seen,the syntax,as shown at the top,has three
pieces,
There is the key word,lambda,which identifies the type of
expression,The first operand position of this expression
identifies the formal parameters of the procedure,This is a list
(or a sequence of names enclosed in a set of parentheses),In
this case,the procedure takes two arguments,which we will
call x and y,Note that there can be an arbitrary number of
arguments for a procedure,including zero (in which case we
would use () as the parameter list),and that this component
determines how many arguments must be passed to the
procedure when it is applied,
The second operand position determines the body of the procedure,that is,the pattern of evaluation to be used,
This may be any valid Scheme expression,Note that this expression is NOT evaluated when the procedure is
created,It is simply stored away as a pattern,It is ONLY evaluated when the procedure is applied in some legal
combination,
That determines the syntax of the lambda,The semantics of this kind of expression is very important,so we place
it on a whole separate slide,which follows,.,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.2.10
..,and I feel like I should be shouting this out!!
The semantics says that the value of a lambda expression is a
procedure object,It is some internal representation of the
procedure that includes information about the formal
parameters and the body,or pattern of computation,
6.001 Notes,Section 2.3
Slide 2.3.1
We have now seen most of the basic elements of Scheme,We
will continue to add a few more special forms,and introduce
some additional built in procedures,but we now have enough
elements of the language to start reasoning about processes,and
especially to use procedures to describe computational
processes,
So let's look at some examples of describing processes in
procedures,
Slide 2.3.2
First,what does a procedure describe?
One useful way of thinking about this is as a means of
generalizing a common pattern of operations,For example,
consider the three expressions shown here,The first two are
straightforward,The third is a bit more general,since
foobar is presumably a name for some numerical value,
However,each of these is basically just a specific instantiation
of a process,the process of multiplying a value by itself,or the
process of "squaring",
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.3.3
So we can capture this by giving a name to the part of the
pattern that changes with each instantiation; identifying that
name as a formal parameter; and then capturing that pattern as
the body of a lambda expression,together with the set of
formal parameters,all within a lambda expression,
Slide 2.3.4
Now lets consider a more complex pattern,as shown here,
Slide 2.3.5
In this case,there are two things that vary,so we will need two
parameters to capture this,Otherwise we could just do the same
thing we did last time,and replicate the pattern,with the
parameters in place of the things that change,as shown,
Slide 2.3.6
But a better way to capture this pattern is to realize that there
are really two things going on,One is the sub-pattern of
squaring things,The other is the use of the results of two
different squaring operations within the larger pattern,
So we could capture each of those patterns within its own
procedure,with its own parameters and body,
Note that in doing this,we are relying on a property of a
combination,namely that a combination involving a named
procedure is equivalent to the pattern captured by the
procedure,with values substituted for the formal parameters,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.3.7
Why is this a better way of capturing the pattern?
The primary reason is that by breaking the pattern up into
smaller modules,we isolate pieces of the computation in
separate abstractions,And these modules can then be reused by
other computations,In particular,the idea of square is
likely to be of value elsewhere,so it makes sense to capture that
in its own procedure,and then use within this larger one,
As well,by doing this,we create code that is easier to read,as
we use a simple name to capture the idea of square,and
suppress the unnecessary details,By doing so,we isolate out
the use of a procedure from the details of its actual
implementation,a trick to which we will return later in the
term,
Of course,there may be many different ways of modularizing a computational pattern,
Slide 2.3.8
Here is a finer scale modularization of the pattern into
procedures,Note how each procedure uses the previous one
within its body,using that idea of abstraction to separate the use
of a procedure from its details,
Slide 2.3.9
Now,let's step away from the specifics of this example,and
talk about the process we just used,
In essence,we did several things,we identified modules or
parts of the computational process,which we could usefully
isolate; we then captured each of those within their own
procedural abstraction; and finally we created a procedure to
control the interactions between the individual modules,Of
course,we could apply this process within each of the modules,
in a recursive fashion,
Our goal now is to see how we can use this general approach to
capture computational processes in procedures,
6.001 Notes,Section 2.4
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.1
Let's take another look at this idea of capturing a procedural
description with our language for describing processes,namely
lambdas,Remember our description of the process of
finding square roots,from the first lecture,shown here,
Slide 2.4.2
Our first step in building code to compute,square roots” is to
determine some good modules,or stages,within this process,
Here,we can see several,
There is the idea of measuring whether our guess is good
enough that we can stop and return an answer,
There is the idea of creating a new guess if we are not close
enough,
And there will need to be a way of controlling the process,in
which we use our new guess as if it were the original one,and
continue the process,
So let's build each of these abstractions,using our idea of
capturing common patterns within lambda expressions,
Slide 2.4.3
Here is a rather naive way of deciding if a guess is close
enough,We take the guess,and square it,If the guess is good,
then that should give us a value close to the number whose
square root we are seeking,Here abs is a built in procedure
that returns the absolute value of its argument,and we simply
test to see if that absolute difference is small,
Note how we are already using procedural abstractions,we
have assumed that square is an abstraction for the process
of squaring numbers,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.4
For improving the guess,we can just use the same approach we
did with pythagoras,we capture the idea of average and we
use it to improve a guess as described by the process,
Slide 2.4.5
As before,we can see that average is likely to be a process
that we will want to use in other places,so creating an
abstraction allows us to avoid replicating the code in those
places,
Moreover,we stress that by building this abstraction,we seal
off the details of the implementation from the actual use of the
abstraction,For example,we could decide to change the
implementation of average,such as that shown here,
Doing so does not require us to make any changes to
procedures that use average however,since those simply
refer to the procedure,not the internal specifics,
Also note that the names of the parameters are internal to the lambda expression,we cannot refer to them
outside the scope of the lambda,Here,we have changed the names of the parameters,but this does not affect
those procedures that use average,
Slide 2.4.6
The last step is to decide how to integrate these pieces together
into a process that controls the steps of the computation,The
basic idea is already captured in the process description,given a
number and a guess,we want to use improve to derive a
new guess,
But now we have to make a decision,is the guess good enough,
in which case we can stop? Or should we continue the process?
In order to make such a decision,we need a new special form,
called an if expression,which has three sub-expressions,a
predicate,a consequence,and an alternative,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.7
Here is how an if expression is evaluated,First,the evaluator
uses its rules (such as those we have described) to determine
the value of the predicate expression,
If that value is true,then the evaluator uses its rules to evaluate
the consequence expression,and returns that value as the value
of the entire if expression,
On the other hand,if that value is not true,then the evaluator
uses its rules to evaluate the alternative expression,and returns
that value as the value of the entire if expression,
Slide 2.4.8
So why do we say that this is a special form,rather than just a
procedure? We'll let you think about this,but after the next
lecture you should be able to answer this question,For now,we
will simply accept that if expressions are evaluated in this
particular manner,
Slide 2.4.9
So back to our process,We know from our description that the
heart of the process should look something like this,We check
to see if we are close enough,and if so,then we just return the
value of the guess,Our if expression will handle that for us,
If we are not close enough,we want to improve our guess,
using the improve procedural abstraction we built,But
somehow we want to then use that value as a new guess,and
repeat the process,
Slide 2.4.10
How do we do that?
Well,let's call this overall process,of repeated improving a
guess until we are close enough,sqrt-loop,Then if we
are not close enough,we get an improved guess,and simply do
this again!
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.11
Finally,we can assemble our sqrt procedure,we just use
this repeated process of improving a guess,all we need to do is
get it started with some initial guess,
Slide 2.4.12
This may look a bit unusual,We have a procedure that refers to
itself within its body,in a recursive fashion,Can we be sure
that this procedure will correctly evolve,as we saw in the first
lecture?
In the next lecture,we are going to introduce a formal model
for tracing the evaluation of expressions,especially expressions
involving the application of procedures,For now,however,we
can be a bit informal and walk through the steps of the
computation,
As an example,let's suppose we try to find the sqrt of 2,
Slide 2.4.13
Basically,we can replace this expression with the body of the
procedure associated with sqrt,where the formal parameter
is replaced with the specific value,In other words,we reverse
the process of capturing a pattern,so we are going to loop
through the process,using an initial guess of 1,
Slide 2.4.14
Now what does sqrt-loop do? Well,the pattern it
captured was to see if we are close enough,here we have the
body of the sqrt-loop procedure with the specific values
in place,For now,we don't care about the other sub-expressions
of the if expression,
6.001 Structure and Interpretation of Computer Programs,Copyright? 2004 by Massachusetts Institute of Technology,
Slide 2.4.15
In this case,we are not close enough,so we move to the
alternative stage of the if expression,as shown,
Slide 2.4.16
Now this is just like a normal combination,We reduce the
values of the subexpressions,so we get the value of the
improved guess..,
Slide 2.4.17
… and then repeat the process,We keep doing this until we get
a value that is close enough that we can stop,
Slide 2.4.18
So to summarize,we have seen that we can use the idea of a
procedure to capture a computational process,by finding good
components or modules of the process; capturing each within
its own procedure; and then deciding how to control the overall
process of the computation,In the next lecture,we will return
to this idea,looking at different ways to break a problem down
into these steps,
