source: doc/theses/jiada_liang_MMath/background.tex @ e561551

Last change on this file since e561551 was e561551, checked in by JiadaL <j82liang@…>, 4 months ago

Save current progress for pull

  • Property mode set to 100644
File size: 18.5 KB
Line 
1\chapter{Background}
2
3\vspace*{-8pt}
4
5\CFA is a backwards-compatible extension of the C programming language, therefore, it must support C-style enumerations.
6The following discussion covers C enumerations.
7
8As discussed in \VRef{s:Aliasing}, it is common for C programmers to ``believe'' there are three equivalent forms of named constants.
9\begin{clang}
10#define Mon 0
11static const int Mon = 0;
12enum { Mon };
13\end{clang}
14\begin{enumerate}[leftmargin=*]
15\item
16For @#define@, the programmer has to explicitly manage the constant name and value.
17Furthermore, these C preprocessor macro names are outside of the C type-system and can incorrectly change random text in a program.
18\item
19The same explicit management is true for the @const@ declaration, and the @const@ variable cannot appear in constant-expression locations, like @case@ labels, array dimensions,\footnote{
20C allows variable-length array-declarations (VLA), so this case does work, but it fails in \CC, which does not support VLAs, unless it is \lstinline{g++}.} immediate oper\-ands of assembler instructions, and occupy storage.
21\begin{clang}
22$\$$ nm test.o
230000000000000018 r Mon
24\end{clang}
25\item
26Only the @enum@ form is managed by the compiler, is part of the language type-system, works in all C constant-expression locations, and normally does not occupy storage.
27\end{enumerate}
28
29
30\section{C \lstinline{const}}
31\label{s:Cconst}
32
33C can simulate the aliasing @const@ declarations \see{\VRef{s:Aliasing}}, with static and dynamic initialization.
34\begin{cquote}
35\begin{tabular}{@{}l@{}l@{}}
36\multicolumn{1}{@{}c@{}}{\textbf{static initialization}} &  \multicolumn{1}{c@{}}{\textbf{dynamic intialization}} \\
37\begin{clang}
38static const int one = 0 + 1;
39static const void * NIL = NULL;
40static const double PI = 3.14159;
41static const char Plus = '+';
42static const char * Fred = "Fred";
43static const int Mon = 0, Tue = Mon + 1, Wed = Tue + 1,
44        Thu = Wed + 1, Fri = Thu + 1, Sat = Fri + 1, Sun = Sat + 1;
45\end{clang}
46&
47\begin{clang}
48void foo() {
49        // auto scope only
50        const int r = random() % 100;
51        int va[r];
52}
53
54
55\end{clang}
56\end{tabular}
57\end{cquote}
58However, statically initialized identifiers can not appear in constant-expression contexts, \eg @case@.
59Dynamically initialized identifiers may appear in initialization and array dimensions in @g++@, which allows variable-sized arrays on the stack.
60Again, this form of aliasing is not an enumeration.
61
62
63\section{C Enumeration}
64\label{s:CEnumeration}
65
66The C enumeration has the following syntax~\cite[\S~6.7.2.2]{C11}.
67\begin{clang}[identifierstyle=\linespread{0.9}\it]
68$\it enum$-specifier:
69        enum identifier$\(_{opt}\)$ { enumerator-list }
70        enum identifier$\(_{opt}\)$ { enumerator-list , }
71        enum identifier
72enumerator-list:
73        enumerator
74        enumerator-list , enumerator
75enumerator:
76        enumeration-constant
77        enumeration-constant = constant-expression
78\end{clang}
79The terms \emph{enumeration} and \emph{enumerator} used in this work \see{\VRef{s:Terminology}} come from the grammar.
80The C enumeration semantics are discussed using examples.
81
82
83\subsection{Type Name}
84\label{s:TypeName}
85
86An \emph{unnamed} enumeration is used to provide aliasing \see{\VRef{s:Aliasing}} exactly like a @const@ declaration in other languages.
87However, it is restricted to integral values.
88\begin{clang}
89enum { Size = 20, Max = 10, MaxPlus10 = Max + 10, @Max10Plus1@, Fred = -7 };
90\end{clang}
91Here, the aliased constants are: 20, 10, 20, 21, and -7.
92Direct initialization is by a compile-time expression generating a constant value.
93Indirect initialization (without initialization, @Max10Plus1@) is \newterm{auto-initialized}: from left to right, starting at zero or the next explicitly initialized constant, incrementing by @1@.
94Because multiple independent enumerators can be combined, enumerators with the same values can occur.
95The enumerators are rvalues, so assignment is disallowed.
96Finally, enumerators are \newterm{unscoped}, \ie enumerators declared inside of an @enum@ are visible (projected) into the enclosing scope of the @enum@ type.
97For unnamed enumerations, this semantic is required because there is no type name for scoped qualification.
98
99As noted, this kind of aliasing declaration is not an enumeration, even though it is declared using an @enum@ in C.
100While the semantics is misleading, this enumeration form matches with aggregate types:
101\begin{cfa}
102typedef struct @/* unnamed */@  { ... } S;
103struct @/* unnamed */@  { ... } x, y, z;        $\C{// questionable}$
104struct S {
105        union @/* unnamed */@ {                                 $\C{// unscoped fields}$
106                int i;  double d ;  char ch;
107        };
108};
109\end{cfa}
110Hence, C programmers would expect this enumeration form to exist in harmony with the aggregate form.
111
112A \emph{named} enumeration is an enumeration:
113\begin{clang}
114enum @Week@ { Mon, Tue, Wed, Thu@ = 10@, Fri, Sat, Sun };
115\end{clang}
116and adopts the same semantics with respect to direct and auto intialization.
117For example, @Mon@ to @Wed@ are implicitly assigned with constants @0@--@2@, @Thu@ is explicitly set to constant @10@, and @Fri@ to @Sun@ are implicitly assigned with constants @11@--@13@.
118As well, initialization may occur in any order.
119\begin{clang}
120enum Week {
121        Thu@ = 10@, Fri, Sat, Sun,
122        Mon@ = 0@, Tue, Wed@,@                  $\C{// terminating comma}$
123};
124\end{clang}
125Note, the comma in the enumerator list can be a terminator or a separator, allowing the list to end with a dangling comma.\footnote{
126A terminating comma appears in other C syntax, \eg the initializer list.}
127This feature allow enumerator lines to be interchanged without moving a comma.
128Named enumerators are also unscoped.
129
130
131\subsection{Representation}
132
133C standard specifies enumeration \emph{variable} is an implementation-defined integral type large enough to hold all enumerator values.
134In practice, C uses @int@ as the underlying type for enumeration variables, because of the restriction to integral constants, which have type @int@ (unless qualified with a size suffix).
135
136\subsection{Usage}
137\label{s:Usage}
138
139C proves an implicit \emph{bidirectional} conversion between an enumeration and its integral type.
140\begin{clang}
141enum Week week = Mon;                           $\C{// week == 0}$
142week = Fri;                                                     $\C{// week == 11}$
143int i = Sun;                                            $\C{// implicit conversion to int, i == 13}$
144@week = 10000;@                                         $\C{// UNDEFINED! implicit conversion to Week}$
145\end{clang}
146While converting an enumerator to its underlying type is useful, the implicit conversion from the base type to an enumeration type is a common source of error.
147
148Enumerators can appear in @switch@ and looping statements.
149\begin{cfa}
150enum Week { Mon, Tue, Wed, Thu, Fri, Sat, Sun };
151switch ( week ) {
152        case Mon: case Tue: case Wed: case Thu: case Fri:
153                printf( "weekday\n" );
154        case Sat: case Sun:
155                printf( "weekend\n" );
156}
157for ( enum Week day = Mon; day <= Sun; day += 1 ) { // step of 1
158        printf( "day %d\n", day ); // 0-6
159}
160\end{cfa}
161For iterating to make sense, the enumerator values \emph{must} have a consecutive ordering with a fixed step between values.
162For example, a gap introduced by @Thu = 10@, results in iterating over the values 0--13, where values 3--9 are not @Week@ values.
163Note, it is the bidirectional conversion that allows incrementing @day@: @day@ is converted to @int@, integer @1@ is added, and the result is converted back to @Week@ for the assignment to @day@.
164For safety, \CC does not support the bidirectional conversion, and hence, an unsafe cast is necessary to increment @day@: @day = (Week)(day + 1)@.
165
166There is a C idiom to automatically compute the number of enumerators in an enumeration.
167\begin{cfa}
168enum E { A, B, C, D, @N@ };  // N == 4
169for ( enum E e = A; e < @N@; e += 1 ) ...
170\end{cfa}
171Here, the auto-incrementing counts the number of enumerators and puts the total into the last enumerator @N@.
172@N@ is often used as the dimension for an array assocated with the enumeration.
173\begin{cfa}
174E array[@N@];
175for ( enum E e = A; e < N; e += 1 ) {
176        array[e] = e;
177}
178\end{cfa}
179However, for typed enumerations, \see{\VRef{f:EumeratorTyping}}, this idiom fails.
180
181This idiom leads to another C idiom using an enumeration with matching companion information.
182For example, an enumeration is linked with a companion array of printable strings.
183\begin{cfa}
184enum Integral_Type { chr, schar, uschar, sshort, ushort, sint, usint, ..., NO_OF_ITYPES };
185char * Integral_Name[@NO_OF_ITYPES@] = {
186        "char", "signed char", "unsigned char",
187        "signed short int", "unsigned short int",
188        "signed int", "unsigned int", ...
189};
190enum Integral_Type integral_type = ...
191printf( "%s\n", Integral_Name[@integral_type@] ); // human readable type name
192\end{cfa}
193However, the companion idiom results in the \emph{harmonizing} problem because an update to the enumeration @Integral_Type@ often requires a corresponding update to the companion array \snake{Integral_Name}.
194The need to harmonize is at best indicated by a comment before the enumeration.
195This issue is exacerbated if enumeration and companion array are in different translation units.
196
197\bigskip
198While C provides a true enumeration, it is restricted, has unsafe semantics, and does provide useful enumeration features in other programming languages.
199
200\section{\CFA Polymorphism}
201\subsection{Function Overloading}
202Function overloading is programming languages feature wherein functions may share the same name, but with different function signatures. In both C++ and \CFA, function names can be overloaded
203with different entities as long as they are different in terms of the number and type of parameters.
204
205\begin{cfa}
206void f(); // (1)
207void f(int); // (2); Overloaded on the number of parameters
208void f(char); // (3); Overloaded on parameter type
209
210f('A');
211\end{cfa}
212In this case, the name f is overloaded with a nullity function and two arity functions with different parameters types. Exactly which precedures being executed
213is determined based on the passing arguments. The last expression of the preceding example calls f with one arguments, narrowing the possible candidates down to (2) and (3).
214Between those, function argument 'A' is an exact match to the parameter expected by (3), while needing an @implicit conversion@ to call (2). The compiler determines (3) is the better candidates among
215and procedure (3) is being executed.
216
217\begin{cfa}
218int f(int); // (4); Overloaded on return type
219[int, int] f(int); // (5) Overloaded on the number of return value
220\end{cfa}
221The function declarations (4) and (5) show the ability of \CFA functions overloaded with different return value, a feature that is not shared by C++.
222
223
224\subsection{Operator Overloading}
225Operators in \CFA are specialized function and are overloadable by with specially-named functions represents the syntax used to call the operator.
226% For example, @bool ?==?T(T lhs, T rhs)@ overloads equality operator for type T, where @?@ is the placeholders for operands for the operator.
227\begin{cfa}
228enum Weekday { Monday, Tuesday, Wednesday, Thursday, Friday, Saturday, Sunday };
229bool ?<?(const Weekday a, const Weekday b) {
230        return ((int)a + 1);
231}
232Monday < Sunday; // False
233?<?( Monday, Sunday ); // Equivalent syntax
234\end{cfa}
235Unary operators are functions that takes one argument and have name @operator?@ or @?operator@, where @?@ is the placeholders for operands.
236Binary operators are function with two parameters. They are overloadable with function name @?operator?@.
237
238\subsection{Constructor and Destructor}
239In \CFA, all objects are initialized by @constructors@ during its allocation, including basic types,
240which are initialized by constructors default-generated by a compiler.
241
242Constructors are overloadable functions with name @?{}@, return @void@, and have at least one parameter, which is a reference
243to the object being constructored (Colloquially referred to "this" or "self" in other language).
244
245\begin{cfa}
246struct Employee {
247        const char * name;
248        double salary;
249};
250
251void ?{}( Employee& this, const char * name, double salary ) {
252    this.name = name;
253    this.salary = salary;
254}
255
256Employee Sara { "Sara Schmidt", 20.5 };
257\end{cfa}
258Like Python, the "self" reference is implicitly passed to a constructor. The Employee constructors takes two additional arugments used in its
259field initialization.
260
261A destructor in \CFA is a function that has name @^?{}@. It returns void, and take only one arugment as its "self".
262\begin{cfa}
263void ^?{}( Employee& this ) {
264    free(this.name);
265    this.name = 0p;
266    this.salary = 0;
267}
268\end{cfa}
269Destructor can be explicitly evoked as a function call, or implicitly called at the end of the block in which the object is delcared.
270\begin{cfa}
271{
272^Sara{};
273Sara{ "Sara Craft", 20 };
274} // ^Sara{}
275\end{cfa}
276
277\subsection{Variable Overloading}
278C and C++ disallow more than one variable declared in the same scope with the same name. When a variable declare in a inner scope has the same name as
279a variable in an outer scope, the outer scope variable is "shadowed" by the inner scope variable and cannot be accessed directly.
280
281\CFA has variable overloading: multiple variables can share the same name in the same scope, as long as they have different type. Name shadowing only
282happens when the inner scope variable and the outer scope ones have the same type.
283\begin{cfa}
284double i = 6.0;
285int i = 5;
286void foo( double i ) { sout | i; } // 6.0
287\end{cfa}
288
289\subsection{Special Literals}
290Literal 0 has special meanings within different contexts: it can means "nothing" or "empty", an additive identity in arithmetic, a default value as in C (null pointer),
291or an initial state.
292Awaring of its significance, \CFA provides a special type for the 0 literal, @zero_t@, to define the logical @zero@ for custom types.
293\begin{cfa}
294struct S { int i, j; };
295void ?{}( S & this, @zero_t@ ) { this.i = 0; this.j = 0; } // zero_t, no parameter name allowed
296S s0 = @0@;
297\end{cfa}
298Overloading @zero_t@ for S provides new definition for @zero@ of type S.
299
300According to the C standard, @0@ is the @only@ false value. Any values compares equals to @0@ is false, and not euqals @0@ is true. As a consequence, control structure
301such as @if()@ and @while()@ only runs it true clause when its predicate @not equals@ to @0@.
302
303\CFA generalizes this concept and allows to logically overloads the boolean value for any type by overloading @not equal@ comparison against @zero_t@.
304\begin{cfa}
305int ?@!=@?( S this, @zero_t@ ) { return this.i != 0 && this.j != 0; }
306\end{cfa}
307
308% In C, the literal 0 represents the Boolean value false. The expression such as @if (x)@ is equivalent to @if (x != 0)@ .
309% \CFA allows user to define the logical zero for a custom type by overloading the @!=@ operation against a special type, @zero_t@,
310% so that an expression with the custom type can be used as a predicate without the need of conversion to the literal 0.
311% \begin{cfa}
312% struct S s;
313% int ?!=?(S, zero_t);
314% if (s) {}
315% \end{cfa}
316Literal 1 is also special. Particularly in C, the pre-increment operator and post-increment operator can be interpreted in terms of @+= 1@.
317The logical @1@ in \CFA is represented by special type @one_t@.
318\begin{cfa}
319void ?{}( S & this, one_t ) { this.i = 1; this.j = 1; } // one_t, no parameter name allowed
320S & ?+=?( S & this, one_t ) { this.i += 1; this.j += 1; return op; }
321\end{cfa}
322Without explictly overloaded by a user, \CFA uses the user-defined @+=(S&, one_t)@ to interpret @?++@ and @++?@, as both are polymorphic functions in \CFA.
323
324\subsection{Polymorphics Functions}
325Parametric-Polymorphics functions are the functions that applied to all types. \CFA functions are parametric-polymorphics when
326they are written with the @forall@ clause.
327
328\begin{cfa}
329forall(T)
330T identity(T x) { return x; }
331identity(42);
332\end{cfa}
333The identity function accepts a value from any type as an arugment, and the type parameter @T@ is bounded to @int@ when the function
334is called with 42.
335
336The forall clause can takes @type assertions@ that restricts the polymorphics type.
337\begin{cfa}
338forall( T | { void foo(T); } )
339void bar(T t) { foo(t); }
340
341struct S {} s;
342void foo(struct S);
343
344bar(s);
345\end{cfa}
346The assertion on @T@ restricts the range of types for bar to only those implements foo with the matching a signature, so that bar() 
347can call @foo@ in its body with type safe.
348Calling on type with no mathcing @foo()@ implemented, such as int, causes a compile time type assertion error.
349
350A @forall@ clause can asserts on multiple types and with multiple asserting functions. A common practice in \CFA is to group
351the asserting functions in to a named @trait@ .
352
353\begin{cfa}
354trait Bird(T) {
355        int days_can_fly(T i);
356        void fly(T t);
357};
358
359forall(B | Bird(B)) {
360        void bird_fly(int days_since_born, B bird) {
361                if (days_since_born > days_can_fly(bird)) {
362                        fly(bird);
363                }
364        }
365}
366
367struct Robin {} r;
368int days_can_fly(Robin r) { return 23; }
369void fly(Robin r) {}
370
371bird_fly( r );
372\end{cfa}
373
374Grouping type assertions into named trait effectively create a reusable interface for parametrics polymorphics types.
375
376\section{Expression Resolution}
377
378The overloading feature poses a challenge in \CFA expression resolution. Overloadeded identifiers can refer multiple
379candidates, with multiples being simultaneously valid. The main task of \CFA resolver is to identity a best candidate that
380involes less implicit conversion and polymorphism.
381
382\subsection{Conversion Cost}
383In C, functions argument and parameter type does not need to be exact match, and the compiler performs an @implicit conversion@ on argument.
384\begin{cfa}
385void foo(double i);
386foo(42);
387\end{cfa}
388The implicit conversion in C is relatively simple because of the abscence of overloading, with the exception of binary operators, for which the
389compiler needs to find a common type of both operands and the result. The pattern is known as "usual arithmetic conversions".
390
391\CFA generalizes C implicit conversion to function overloading as a concept of @conversion cost@.
392Initially designed by Bilson, conversion cost is a 3-tuple, @(unsafe, poly, safe)@, where unsafe is the number of narrowing conversion,
393poly is the count of polymorphics type binding, and safe is the sum of the degree of widening conversion. Every
394basic type in \CFA has been assigned with a @distance to Byte@, or @distance@, and the degree of widening conversion is the difference between two distances.
395
396Aaron extends conversion cost to a 7-tuple,
397@@(unsafe, poly, safe, sign, vars, specialization, reference)@@. The summary of Aaron's cost model is the following:
398\begin{itemize}
399\item Unsafe is the number of argument that implicitly convert to a type with high rank.
400\item Poly accounts for number of polymorphics binding in the function declaration.
401\item Safe is sum of distance (add reference/appendix later).
402\item Sign is the number of sign/unsign variable conversion.
403\item Vars is the number of polymorphics type declared in @forall@.
404\item Specialization is opposite number of function declared in @forall@. More function declared implies more constraint on polymorphics type, and therefore has the lower cost.
405\item Reference is number of lvalue-to-rvalue conversion.
406\end{itemize}
Note: See TracBrowser for help on using the repository browser.