More Flexible Software By Favoring Implicit Calls and Implicit Communication

1
More Flexible Software By Favoring Implicit
Callsand Implicit Communication

• by Karl Lieberherr
• Joint work with Bryan Chadwick, Ahmed Abdelmeged
  and Therapon Skotiniotis

2
What we want
well-accepted

• Data Abstraction Principle the implementation of
  objects can be changed without affecting clients
  provided the interface holds.
• Adaptive Programming Principle
  the interface of objects can be changed without
  affecting clients provided adaptive constraints
  hold.

3

• can abstract traversal code
• non-cyclic objects
• explicit and implicit cyclic objects cause
  problems

4
Abstraction boundaries

• M consists of (interfaceM, implementationM) so
  that interfaceM(implementationM). implementationM
  can be changed to implementationM provided
  interfaceM(implementationM).
• M2 uses M1.

5
Abstraction boundaries

• M consists of (faceM, mentM) so that
  faceM(mentM). mentM can be changed to mentM
  provided faceM(mentM).
• (faceM2,mentM2) uses (faceM1,mentM1). mentM2
  relies on faceM1 which changes to faceM1.
  Implementation of M2 defines constraints on M1.

6
Module

• Consists of Interface and Implementation.
  Implementation imports other modules.
• AdaptiveConstraints Constraints on imported
  modules.

7
Module Interface Implementation
uses
module M2
module M1
faceM1 mentM1
faceM2 mentM2
uses
faceM1 mentM1
faceM1 mentM12
faceM1 mentM1
faceM1 mentM13
change implementation
change interface
8
Module Interface Implementation
AdaptiveConstraints
uses (imports)
module M2
module M1
faceM1 mentM1
faceM2 mentM2 AdapConM1
uses (imports)
faceM1 mentM1
module M1
faceM1 mentM12
faceM1 mentM1
faceM1 mentM13
change implementation
change interface
AdapConM1(M1)
9
What Hinders Creation of Flexible Software

• Rigidly following rules like Follow the
  structure, follow the grammar.
• Actively call traversal methods (explicit
  traversal problem).
• Also leads to manual passing of arguments
  (explicit argument problem).

10
Alternative 1 AP-F

• Think of computation as moving or pushing up
  information in an object (combine methods).
• Allow transformation at each node (apply
  methods).
• Send information down as needed (update methods).
• Default behavior copy object.

11
Alternative 2 AP-P

• Think of computation as traversing the object and
  collecting information in a visitor object.
• Use both regular visitor variables as well as
  interposition variables.
• Interposition variables facilitate implicit
  communication.
• Default behavior traverse object.

12
AP-F implemented in Java by DemeterF

• Examples plan
• apply only
• apply and combine
• apply and combine and update

13
Type Unifying information flow
T combine(A a, T t)
X update(A a, X x)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
X update(B b, X x)
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d, X x)
E
T apply(E e, X x)
14
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
15
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
16
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
17
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
18
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
19
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
20
Type Unifying information flowno update
T combine(A a, T t)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d)
E
T apply(E e)
21
Type Unifying information flow
combine(a,apply(combine(b,apply(combine(c,apply(e,
x))),
apply(d,x))))
T combine(A a, T t)
X update(A a, X x)
A
object graph
T apply(T t) T combine(B b, T t1, T t2)
B
X update(B b, X x)
C
D
T apply(T t) T combine(C c, T t1)
T apply(D d, X x)
E
T apply(E e, X x)
22
Type Preserving information flow
A combine(A a, B b)
X update(A a, X x)
A
object graph
B apply(B b) B combine(B b, C c, D d)
B
X update(B b, X x)
C
D
C apply(C c) C combine(C c, E e)
D apply(D d, X x)
E
E apply(E e, X x)
23
Type Preserving information flow
combine(a,apply(combine(b,apply(combine(c,apply(e,
x))),
apply(d,x))))
A combine(A a, B b)
X update(A a, X x)
A
object graph
B apply(B b) B combine(B b, C c, D d)
B
X update(B b, X x)
C
D
C apply(C c) C combine(C c, E e)
D apply(D d, X x)
E
E apply(E e, X x)
24
semantics apply-combine expression
Tree ltngt Node ltleftgt Tree
ltrightgt Tree. Node IdentityApply
ltogt Object.

• apply(combine(this,
• apply(combine(left,)),
• apply(combine(right,))
• ))
• Translate tree object into apply-combine
  expression.

25
Traversal semantics
26
Default Transformercopy object
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
27
Parameterize Default Transformer
PathSpec apply(J j) return new
Complement(j)
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
28
Parameterize Default Transformer
PathSpec combine(J j, Boolean fn, Boolean sn)
return fn sn
PathSpec combine(Object j, Boolean fn, Boolean
sn) return fn sn
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
29
Simple applicationProgram transformation

• Old
• E Num Var Op Call
• Op Plus Equals.
• Equals .
• New
• E Bool.
• Bool True False.

class BoolTrans extends IDf static E newtrue
Call.parse(( 1 1)), static E newfalse
Call.parse(( 1 0) ) E apply(True t) return
newtrue E apply(False t) return newfalse

apply for transformation of result returned by
builder
30
de Bruijn indices
for later

• Old
• Var Sym.
• Sym Ident.
• New
• Var Sym Addr.
• Addr Integer.

class AddrTrans extends IDf Var apply(Var
var, SymList senv) return new
Addr(senv.lookup(var)) class SymExtender
extends IDa SymList update(Lambda l, SymList
senv) return senv.push(l.formals)
31
The default Builder for PathSpec
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
32
well-formed movie

• show how the default builder is modified to
  combine Boolean objects.

33
The default Builder for PathSpecwell-formed
specialization 1
t true f false
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
t
5
2
S
S
S
S
7
8
t
t
t
3
4
t
34
The default Builder for PathSpecwell-formed
specialization 2
t true f false
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
t
t
5
2
S
S
S
S
7
8
t
t
t
3
4
t
35
The default Builder for PathSpecwell-formed
specialization 3
t true f false
after blue arrow copy is built (like after)
10
t
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
t
t
t
3
4
t
36
The default Builder for PathSpecwell-formed
specialization 4
t true f false
after blue arrow copy is built (like after)
10
t
M
1
9
J
6
t
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
t
t
t
3
4
t
37
The default Builder for PathSpecwell-formed
specialization 5
t true f false
after blue arrow copy is built (like after)
10
t
M
1
9
J
6
t
t
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
t
t
t
3
4
t
38
The default Builder for PathSpecwell-formed
specialization 6
t true f false
after blue arrow copy is built (like after)
10
t
f
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
t
t
t
3
4
t
39
The default Builder for PathSpecwell-formed
specialization 7
f
t true f false
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
t
t
t
3
4
t
40
The default Builder for PathSpecwell-formed
specialization
t true f false
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
t
t
t
3
4
t
41
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
42
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
S
means a copy
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
43
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
44
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
45
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
46
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
47
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
48
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
after blue arrow copy is built (like after)
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
49
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
insert NOT
after blue arrow copy is built (like after)
N
10
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
50
The default Builder for PathSpecNOT_JOIN_MERGE
specialization
N
insert NOT
M
after blue arrow copy is built (like after)
N
10
N
M
1
9
J
6
Count only upon first visit (red) and upon final
visit (blue). For leaf nodes, count only in red.
J
5
2
S
S
S
S
7
8
3
4
51
Illustration of combinefor capacity constraint
violation
19a
19
C
(w1w2w3w4,2)
Cons
(w1w2w3w4,1)
E
(w4,0)
1
20
2
18a
3
18
3a
13a
(w1w2w3,1)
Cons
after blue arrow combine is active (like after)
17a
C
(w2w3,1)
4
12a
17
Cons
(w1,0)
5
13
14
12
16a
(w2w3,0)
Cons
15a
6
(0,0)
E
Empty
(w1,0)
11a
7a
15
16
11
E
(w3,0)
Cons
(w2,0)
both containers (C) violate capacity constraints
8
7
10a
9a
(w2,0)
E
Empty
(0,0)
9
10
52
Illustration of combinefor capacity constraint
violation
C
(w1w2w3w4,2)
Cons
(w1w2w3w4,1)
E
(w4,0)
(w1w2w3,1)
Cons
C
(w2w3,1)
Cons
(w1,0)
(w2w3,0)
Cons
(0,0)
E
Empty
(w1,0)
E
(w3,0)
Cons
(w2,0)
both containers (C) violate capacity constraints
(w2,0)
E
Empty
(0,0)
53
Theory
On left side of gt the term c() only indicates a
compound object.
f apply b combine

• tf,b(d) gt d, where df(d), d is atomic
• tf,b(c(d0, ,dn)) gt
• f(b(c(d0, ,dn), d0, ,dn)), where di
  tf,b(di)
• Default functions
• idf(d) gt d
• idb(c(d0, ,dn), d0, ,dn) gt c(d0, ,dn)
• bc(c(d0, ,dn), d0, ,dn) gt c(d0, ,dn)

54
Theory

• f is a polymorphic function that takes a single
  argument and returns a result.
• b is a function object that is responsible for
  reconstruction of data types.

55
Traversal Component Approach

• implemented in DemeterF
• Traversal with 3 components Builder (combine),
  Transformer (apply), Augmentor (update)

56
Augmentors / update methods

• so far we covered combine and apply methods
• combine to combine information up the object
• apply to transform before the information is
  sent up.
• up refers to the traversal when a traversal has
  finished visiting an object, it goes up.
• add update, to send information down the object
• if it's not used/needed, it does not need to be
  mentioned, since the traversal will do the
  passing around.

57
motivating exampleAddress computation

• Var Sym Addr.
• Addr Integer.
• class AddrTrans extends IDf
• Var apply(Var var, SymList senv)
• return new Addr(senv.lookup(var))
• class SymExtender extends IDa
• SymList update(Lambda la, Symlist senv)
• return senv.push(la.formals)

58
Optional argument

• The argument passed into the traversal (by the
  programmer) is available everywhere (in
  everyapply/combine/update function) but it may
  be needed only in some objects.
• In the typechecker case is only needed when
  looking up a variable use, or modifying the
  type-environment with a binding.  

59
Like a let for subtraversals

• An update argument can be viewed (almost) as a
  'let' for sub-traversals before traversing
  sub-terms we do a recalculation of the traversal
  argument. Update is called before traversing sub
  terms, and themodified value is only available
  for sub terms.

60
Structure-shy passing of argument

• If you look at the Scheme code you would write to
  traverse the same structure you would pass
  alongan argument to the functions, all the way
  through recursive calls until it is needed. At a
  lambdathe recursive call on the body would look
  something like  (type-Lambda l tenv)   
  (cases Exp l       (Lambda (arg body)       
   (type-Exp body (tenv-extend tenv arg)))     
   (Call (op arglist)          (let ((ret
  (type-Exp op tenv)) ...)       (... other
  cases...)))

61
Default passing it alongno code needed with
DemeterF

• Note that we don't change the traversal argument
  in most cases, only when binding an argument to a
  type. Usually we just keep passing it along
  throughout the traversal functions. The
  augmentor/update methods encapsulate (only) the
  changes of this argument, so we simply write
• TEnv update(Lambda l, TEnv te)
• return te.extend(l.formal)
• The traversal takes care of the default case when
  we don't need to change the argument.

62
All arguments are optional

• All arguments are now optional... thus the most
  general method is    Object combine() return
  ... Since it is applicable in all cases (all
  arguments optional, including the 'traversal
  argument').(the traversal argument is the one
  updated by update methods. There is only one
  traversal argument)

63
AST illustration

• Call    --- Lambda         --- Arg     
          --- Type                  ---
  int                            ---
  Sym                    --- 'a'       
            --- Call               ---
  Plus               --- Sym               
      --- 'a'                            
   --- Num                     --- 2    ---
   Num           --- 4

Update is called at each label and the Object
returned is then available to all terms
'connected'and to the right of that label.
64
Technical Details

• Methods you want to call determine from where you
  inherit ID (all), IDa (update), IDb (combine),
  IDba (combine, update), IDf (apply), IDfa (apply,
  update), IDfb(apply, combine).

65
Motivation

• Showing a full scheme type-checking function, and
  highlighting the points wherethe
  type-environment is passed, used, and extended.
   The number of cases where it is just passed
  along motivates the want/need to put the
  modifications in one place, and being able to
  ignore theargument when it's not really needed.

66
Traversal

• a function

67
What is shy code?

• Shy code shouldnt reveal too much of itself and
  shouldnt be too nosy into others affairs.
• Law of Demeter only talk to your friends