source: doc/theses/andrew_beach_MMath/features.tex @ 93d0ed3

Last change on this file since 93d0ed3 was 93d0ed3, checked in by Peter A. Buhr <pabuhr@…>, 2 years ago

fix problem in virtual type examples and figure

  • Property mode set to 100644
File size: 38.6 KB
1\chapter{Exception Features}
4This chapter covers the design and user interface of the \CFA EHM
5and begins with a general overview of EHMs. It is not a strict
6definition of all EHMs nor an exhaustive list of all possible features.
7However it does cover the most common structure and features found in them.
9\section{Overview of EHMs}
10% We should cover what is an exception handling mechanism and what is an
11% exception before this. Probably in the introduction. Some of this could
12% move there.
13\subsection{Raise / Handle}
14An exception operation has two main parts: raise and handle.
15These terms are sometimes known as throw and catch but this work uses
16throw/catch as a particular kind of raise/handle.
17These are the two parts that the user writes and may
18be the only two pieces of the EHM that have any syntax in a language.
21The raise is the starting point for exception handling
22by raising an exception, which passes it to
23the EHM.
25Some well known examples include the @throw@ statements of \Cpp and Java and
26the \code{Python}{raise} statement of Python. In real systems, a raise may
27perform some other work (such as memory management) but for the
28purposes of this overview that can be ignored.
31The primary purpose of an EHM is to run some user code to handle a raised
32exception. This code is given, with some other information, in a handler.
34A handler has three common features: the previously mentioned user code, a
35region of code it guards, and an exception label/condition that matches
36the raised exception.
37Only raises inside the guarded region and raising exceptions that match the
38label can be handled by a given handler.
39If multiple handlers could can handle an exception,
40EHMs define a rule to pick one, such as ``best match" or ``first found".
42The @try@ statements of \Cpp, Java and Python are common examples. All three
43show the common features of guarded region, raise, matching and handler.
45try {                           // guarded region
46        ...     
47        throw exception;        // raise
48        ...     
49} catch( exception ) {  // matching condition, with exception label
50        ...                             // handler code
55After an exception is raised comes what is usually the biggest step for the
56EHM: finding and setting up the handler for execution. The propagation from raise to
57handler can be broken up into three different tasks: searching for a handler,
58matching against the handler and installing the handler.
61The EHM begins by searching for handlers that might be used to handle
62the exception. The search is restricted to
63handlers that have the raise site in their guarded
65The search includes handlers in the current function, as well as any in
66callers on the stack that have the function call in their guarded region.
69Each handler found is matched with the raised exception. The exception
70label defines a condition that is used with the exception and decides if
71there is a match or not.
72In languages where the first match is used, this step is intertwined with
73searching; a match check is performed immediately after the search finds
74a handler.
77After a handler is chosen, it must be made ready to run.
78The implementation can vary widely to fit with the rest of the
79design of the EHM. The installation step might be trivial or it could be
80the most expensive step in handling an exception. The latter tends to be the
81case when stack unwinding is involved.
83If a matching handler is not guaranteed to be found, the EHM needs a
84different course of action for this case.
85This situation only occurs with unchecked exceptions as checked exceptions
86(such as in Java) are guaranteed to find a matching handler.
87The unhandled action is usually very general, such as aborting the program.
90A common way to organize exceptions is in a hierarchical structure.
91This pattern comes from object-orientated languages where the
92exception hierarchy is a natural extension of the object hierarchy.
94Consider the following exception hierarchy:
98A handler labeled with any given exception can handle exceptions of that
99type or any child type of that exception. The root of the exception hierarchy
100(here \code{C}{exception}) acts as a catch-all, leaf types catch single types,
101and the exceptions in the middle can be used to catch different groups of
102related exceptions.
104This system has some notable advantages, such as multiple levels of grouping,
105the ability for libraries to add new exception types, and the isolation
106between different sub-hierarchies.
107This design is used in \CFA even though it is not a object-orientated
108language; so different tools are used to create the hierarchy.
110% Could I cite the rational for the Python IO exception rework?
113After the handler has finished, the entire exception operation has to complete
114and continue executing somewhere else. This step is usually simple,
115both logically and in its implementation, as the installation of the handler
116is usually set up to do most of the work.
118The EHM can return control to many different places, where
119the most common are after the handler definition (termination)
120and after the raise (resumption).
123For effective exception handling, additional information is often passed
124from the raise to the handler and back again.
125So far, only communication of the exception's identity is covered.
126A common communication method for passing more information is putting fields into the exception instance
127and giving the handler access to them.
128Using reference fields pointing to data at the raise location allows data to be
129passed in both directions.
133Virtual types and casts are not part of \CFA's EHM nor are they required for
134an EHM.
135However, one of the best ways to support an exception hierarchy
136is via a virtual hierarchy and dispatch system.
137Ideally, the virtual system should have been part of \CFA before the work
138on exception handling began, but unfortunately it was not.
139Hence, only the features and framework needed for the EHM were
140designed and implemented for this thesis. Other features were considered to ensure that
141the structure could accommodate other desirable features in the future
142but are not implemented.
143The rest of this section only discusses the implemented subset of the
144virtual-system design.
146The virtual system supports multiple ``trees" of types. Each tree is
147a simple hierarchy with a single root type. Each type in a tree has exactly
148one parent -- except for the root type which has zero parents -- and any
149number of children.
150Any type that belongs to any of these trees is called a virtual type.
151For example, the following hypothetical syntax creates two virtual-type trees.
156vtype V0, V1(V0), V2(V0);
157vtype W0, W1(W0), W2(W1);
164% A type's ancestors are its parent and its parent's ancestors.
165% The root type has no ancestors.
166% A type's descendants are its children and its children's descendants.
167Every virtual type (tree node) has a pointer to a virtual table with a unique
168@Id@ and a list of virtual members (see \autoref{s:VirtualSystem} for
169details). Children inherit their parent's list of virtual members but may add
170and/or replace members.  For example,
172vtable W0 | { int ?<?( int, int ); int ?+?( int, int ); }
173vtable W1 | { int ?+?( int, int ); int w, int ?-?( int, int ); }
175creates a virtual table for @W0@ initialized with the matching @<@ and @+@
176operations visible at this declaration context.  Similarly, @W1@ is initialized
177with @<@ from inheritance with @W0@, @+@ is replaced, and @-@ is added, where
178both operations are matched at this declaration context. It is important to
179note that these are virtual members, not virtual methods of object-orientated
180programming, and can be of any type. Finally, trait names can be used to
181specify the list of virtual members.
183\PAB{Need to look at these when done.
185\CFA still supports virtual methods as a special case of virtual members.
186Function pointers that take a pointer to the virtual type are modified
187with each level of inheritance so that refers to the new type.
188This means an object can always be passed to a function in its virtual table
189as if it were a method.
190\todo{Clarify (with an example) virtual methods.}
193Up until this point the virtual system is similar to ones found in
194object-orientated languages but this is where \CFA diverges. Objects encapsulate a
195single set of methods in each type, universally across the entire program,
196and indeed all programs that use that type definition. Even if a type inherits and adds methods, it still encapsulate a
197single set of methods. In this sense,
198object-oriented types are ``closed" and cannot be altered.
200In \CFA, types do not encapsulate any code. Traits are local for each function and
201types can satisfy a local trait, stop satisfying it or, satisfy the same
202trait in a different way at any lexical location in the program where a function is call.
203In this sense, the set of functions/variables that satisfy a trait for a type is ``open" as the set can change at every call site.
204This capability means it is impossible to pick a single set of functions
205that represent a type's implementation across a program.
207\CFA side-steps this issue by not having a single virtual table for each
208type. A user can define virtual tables that are filled in at their
209declaration and given a name. Anywhere that name is visible, even if it is
210defined locally inside a function \PAB{What does this mean? (although that means it does not have a
211static lifetime)}, it can be used.
212Specifically, a virtual type is ``bound" to a virtual table that
213sets the virtual members for that object. The virtual members can be accessed
214through the object.
216While much of the virtual infrastructure is created, it is currently only used
217internally for exception handling. The only user-level feature is the virtual
218cast, which is the same as the \Cpp \code{C++}{dynamic_cast}.
223Note, the syntax and semantics matches a C-cast, rather than the function-like
224\Cpp syntax for special casts. Both the type of @EXPRESSION@ and @TYPE@ must be
225a pointer to a virtual type.
226The cast dynamically checks if the @EXPRESSION@ type is the same or a sub-type
227of @TYPE@, and if true, returns a pointer to the
228@EXPRESSION@ object, otherwise it returns @0p@ (null pointer).
231% Leaving until later, hopefully it can talk about actual syntax instead
232% of my many strange macros. Syntax aside I will also have to talk about the
233% features all exceptions support.
235Exceptions are defined by the trait system; there are a series of traits, and
236if a type satisfies them, then it can be used as an exception. The following
237is the base trait all exceptions need to match.
239trait is_exception(exceptT &, virtualT &) {
240        // Numerous imaginary assertions.
243The trait is defined over two types, the exception type and the virtual table
244type. Each exception type should have a single virtual table type.
245There are no actual assertions in this trait because the trait system
246cannot express them yet (adding such assertions would be part of
247completing the virtual system). The imaginary assertions would probably come
248from a trait defined by the virtual system, and state that the exception type
249is a virtual type, is a descendant of @exception_t@ (the base exception type),
250and note its virtual table type.
252% I did have a note about how it is the programmer's responsibility to make
253% sure the function is implemented correctly. But this is true of every
254% similar system I know of (except Agda's I guess) so I took it out.
256There are two more traits for exceptions defined as follows:
258trait is_termination_exception(
259                exceptT &, virtualT & | is_exception(exceptT, virtualT)) {
260        void defaultTerminationHandler(exceptT &);
263trait is_resumption_exception(
264                exceptT &, virtualT & | is_exception(exceptT, virtualT)) {
265        void defaultResumptionHandler(exceptT &);
268Both traits ensure a pair of types are an exception type, its virtual table
270and defines one of the two default handlers. The default handlers are used
271as fallbacks and are discussed in detail in \vref{s:ExceptionHandling}.
273However, all three of these traits can be tricky to use directly.
274While there is a bit of repetition required,
275the largest issue is that the virtual table type is mangled and not in a user
276facing way. So these three macros are provided to wrap these traits to
277simplify referring to the names:
280All three take one or two arguments. The first argument is the name of the
281exception type. The macro passes its unmangled and mangled form to the trait.
282The second (optional) argument is a parenthesized list of polymorphic
283arguments. This argument is only used with polymorphic exceptions and the
284list is be passed to both types.
285In the current set-up, the two types always have the same polymorphic
286arguments so these macros can be used without losing flexibility.
288For example consider a function that is polymorphic over types that have a
289defined arithmetic exception:
291forall(Num | IS_EXCEPTION(Arithmetic, (Num)))
292void some_math_function(Num & left, Num & right);
295\section{Exception Handling}
297As stated,
298\CFA provides two kinds of exception handling: termination and resumption.
299These twin operations are the core of \CFA's exception handling mechanism.
300This section covers the general patterns shared by the two operations and
301then goes on to cover the details of each individual operation.
303Both operations follow the same set of steps.
304First, a user raises an exception.
305Second, the exception propagates up the stack.
306Third, if a handler is found, the exception is caught and the handler is run.
307After that control continues at a raise-dependent location.
308Fourth, if a handler is not found, a default handler is run and, if it returns, then control
309continues after the raise.
311%This general description covers what the two kinds have in common.
312The differences in the two operations include how propagation is performed, where execution continues
313after an exception is caught and handled, and which default handler is run.
317Termination handling is the familiar EHM and used in most programming
318languages with exception handling.
319It is a dynamic, non-local goto. If the raised exception is matched and
320handled, the stack is unwound and control (usually) continues in the function
321on the call stack that defined the handler.
322Termination is commonly used when an error has occurred and recovery is
323impossible locally.
325% (usually) Control can continue in the current function but then a different
326% control flow construct should be used.
328A termination raise is started with the @throw@ statement:
330throw EXPRESSION;
332The expression must return a reference to a termination exception, where the
333termination exception is any type that satisfies the trait
334@is_termination_exception@ at the call site.
335Through \CFA's trait system, the trait functions are implicitly passed into the
336throw code for use by the EHM.
337A new @defaultTerminationHandler@ can be defined in any scope to
338change the throw's behaviour when a handler is not found (see below).
340The throw copies the provided exception into managed memory to ensure
341the exception is not destroyed if the stack is unwound.
342It is the user's responsibility to ensure the original exception is cleaned
343up whether the stack is unwound or not. Allocating it on the stack is
344usually sufficient.
346% How to say propagation starts, its first sub-step is the search.
347Then propagation starts with the search. \CFA uses a ``first match" rule so
348matching is performed with the copied exception as the search key.
349It starts from the raise in the throwing function and proceeds towards the base of the stack,
350from callee to caller.
351At each stack frame, a check is made for termination handlers defined by the
352@catch@ clauses of a @try@ statement.
354try {
355        GUARDED_BLOCK
356} catch (EXCEPTION_TYPE$\(_1\)$ * [NAME$\(_1\)$]) {
357        HANDLER_BLOCK$\(_1\)$
358} catch (EXCEPTION_TYPE$\(_2\)$ * [NAME$\(_2\)$]) {
359        HANDLER_BLOCK$\(_2\)$
362When viewed on its own, a try statement simply executes the statements
363in the \snake{GUARDED_BLOCK}, and when those are finished,
364the try statement finishes.
366However, while the guarded statements are being executed, including any
367invoked functions, all the handlers in these statements are included in the
368search path.
369Hence, if a termination exception is raised, these handlers may be matched
370against the exception and may handle it.
372Exception matching checks the handler in each catch clause in the order
373they appear, top to bottom. If the representation of the raised exception type
374is the same or a descendant of @EXCEPTION_TYPE@$_i$, then @NAME@$_i$
375(if provided) is
376bound to a pointer to the exception and the statements in @HANDLER_BLOCK@$_i$
377are executed. If control reaches the end of the handler, the exception is
378freed and control continues after the try statement.
380If no termination handler is found during the search, then the default handler
381(\defaultTerminationHandler) visible at the raise statement is called.
382Through \CFA's trait system the best match at the raise statement is used.
383This function is run and is passed the copied exception.
384If the default handler finishes, control continues after the raise statement.
386There is a global @defaultTerminationHandler@ that is polymorphic over all
387termination exception types.
388The global default termination handler performs a cancellation
389(see \vref{s:Cancellation} for the justification) on the current stack with the copied exception.
390Since it is so general, a more specific handler is usually
391defined, possibly with a detailed message, and used for specific exception type, effectively overriding the default handler.
396Resumption exception handling is the less familar EHM, but is
397just as old~\cite{Goodenough75} and is simpler in many ways.
398It is a dynamic, non-local function call. If the raised exception is
399matched, a closure is taken from up the stack and executed,
400after which the raising function continues executing.
401The common uses for resumption exceptions include
402potentially repairable errors, where execution can continue in the same
403function once the error is corrected, and
404ignorable events, such as logging where nothing needs to happen and control
405should always continue from the raise point.
407A resumption raise is started with the @throwResume@ statement:
409throwResume EXPRESSION;
411\todo{Decide on a final set of keywords and use them everywhere.}
412It works much the same way as the termination throw.
413The expression must return a reference to a resumption exception,
414where the resumption exception is any type that satisfies the trait
415@is_resumption_exception@ at the call site.
416The assertions from this trait are available to
417the exception system while handling the exception.
419At run-time, no exception copy is made, since
420resumption does not unwind the stack nor otherwise remove values from the
421current scope, so there is no need to manage memory to keep the exception in scope.
423Then propagation starts with the search. It starts from the raise in the
424resuming function and proceeds towards the base of the stack,
425from callee to caller.
426At each stack frame, a check is made for resumption handlers defined by the
427@catchResume@ clauses of a @try@ statement.
429try {
430        GUARDED_BLOCK
431} catchResume (EXCEPTION_TYPE$\(_1\)$ * [NAME$\(_1\)$]) {
432        HANDLER_BLOCK$\(_1\)$
433} catchResume (EXCEPTION_TYPE$\(_2\)$ * [NAME$\(_2\)$]) {
434        HANDLER_BLOCK$\(_2\)$
437% PAB, you say this above.
438% When a try statement is executed, it simply executes the statements in the
439% @GUARDED_BLOCK@ and then finishes.
441% However, while the guarded statements are being executed, including any
442% invoked functions, all the handlers in these statements are included in the
443% search path.
444% Hence, if a resumption exception is raised, these handlers may be matched
445% against the exception and may handle it.
447% Exception matching checks the handler in each catch clause in the order
448% they appear, top to bottom. If the representation of the raised exception type
449% is the same or a descendant of @EXCEPTION_TYPE@$_i$, then @NAME@$_i$
450% (if provided) is bound to a pointer to the exception and the statements in
451% @HANDLER_BLOCK@$_i$ are executed.
452% If control reaches the end of the handler, execution continues after the
453% the raise statement that raised the handled exception.
455% Like termination, if no resumption handler is found during the search,
456% then the default handler (\defaultResumptionHandler) visible at the raise
457% statement is called. It will use the best match at the raise sight according
458% to \CFA's overloading rules. The default handler is
459% passed the exception given to the raise. When the default handler finishes
460% execution continues after the raise statement.
462% There is a global @defaultResumptionHandler{} is polymorphic over all
463% resumption exceptions and performs a termination throw on the exception.
464% The \defaultTerminationHandler{} can be overridden by providing a new
465% function that is a better match.
467The @GUARDED_BLOCK@ and its associated nested guarded statements work the same
468for resumption as for termination, as does exception matching at each
469@catchResume@. Similarly, if no resumption handler is found during the search,
470then the currently visible default handler (\defaultResumptionHandler) is
471called and control continues after the raise statement if it returns. Finally,
472there is also a global @defaultResumptionHandler@, which can be overridden,
473that is polymorphic over all resumption exceptions but performs a termination
474throw on the exception rather than a cancellation.
476Throwing the exception in @defaultResumptionHandler@ has the positive effect of
477walking the stack a second time for a recovery handler. Hence, a programmer has
478two chances for help with a problem, fixup or recovery, should either kind of
479handler appear on the stack. However, this dual stack walk leads to following
480apparent anomaly:
482try {
483        throwResume E;
484} catch (E) {
485        // this handler runs
488because the @catch@ appears to handle a @throwResume@, but a @throwResume@ only
489matches with @catchResume@. The anomaly results because the unmatched
490@catchResuem@, calls @defaultResumptionHandler@, which in turn throws @E@.
492% I wonder if there would be some good central place for this.
493Note, termination and resumption handlers may be used together
494in a single try statement, intermixing @catch@ and @catchResume@ freely.
495Each type of handler only interacts with exceptions from the matching
496kind of raise.
498\subsubsection{Resumption Marking}
500A key difference between resumption and termination is that resumption does
501not unwind the stack. A side effect is that, when a handler is matched
502and run, its try block (the guarded statements) and every try statement
503searched before it are still on the stack. There presence can lead to
504the \emph{recursive resumption problem}.
506The recursive resumption problem is any situation where a resumption handler
507ends up being called while it is running.
508Consider a trivial case:
510try {
511        throwResume (E &){};
512} catchResume(E *) {
513        throwResume (E &){};
516When this code is executed, the guarded @throwResume@ starts a
517search and matches the handler in the @catchResume@ clause. This
518call is placed on the stack above the try-block. Now the second raise in the handler
519searches the same try block, matches, and puts another instance of the
520same handler on the stack leading to infinite recursion.
522While this situation is trivial and easy to avoid, much more complex cycles can
523form with multiple handlers and different exception types.  The key point is
524that the programmer's intuition expects every raise in a handler to start
525searching \emph{below} the @try@ statement, making it difficult to understand
526and fix the problem.
528To prevent all of these cases, each try statement is ``marked" from the
529time the exception search reaches it to either when a matching handler
530completes or when the search reaches the base
531of the stack.
532While a try statement is marked, its handlers are never matched, effectively
533skipping over it to the next try statement.
539There are other sets of marking rules that could be used,
540for instance, marking just the handlers that caught the exception,
541would also prevent recursive resumption.
542However, the rule selected mirrors what happens with termination,
543and hence, matches programmer intuition that a raise searches below a try.
545In detail, the marked try statements are the ones that would be removed from
546the stack for a termination exception, \ie those on the stack
547between the handler and the raise statement.
548This symmetry applies to the default handler as well, as both kinds of
549default handlers are run at the raise statement, rather than (physically
550or logically) at the bottom of the stack.
551% In early development having the default handler happen after
552% unmarking was just more useful. We assume that will continue.
554\section{Conditional Catch}
555Both termination and resumption handler clauses can be given an additional
556condition to further control which exceptions they handle:
560First, the same semantics is used to match the exception type. Second, if the
561exception matches, @CONDITION@ is executed. The condition expression may
562reference all names in scope at the beginning of the try block and @NAME@
563introduced in the handler clause. If the condition is true, then the handler
564matches. Otherwise, the exception search continues as if the exception type
565did not match.
567The condition matching allows finer matching by checking
568more kinds of information than just the exception type.
570try {
571        handle1 = open( f1, ... );
572        handle2 = open( f2, ... );
573        handle3 = open( f3, ... );
574        ...
575} catch( IOFailure * f ; fd( f ) == f1 ) {
576        // Only handle IO failure for f1.
577} catch( IOFailure * f ; fd( f ) == f3 ) {
578        // Only handle IO failure for f3.
580// Handle a failure relating to f2 further down the stack.
582In this example the file that experienced the IO error is used to decide
583which handler should be run, if any at all.
586% I know I actually haven't got rid of them yet, but I'm going to try
587% to write it as if I had and see if that makes sense:
590Within the handler block or functions called from the handler block, it is
591possible to reraise the most recently caught exception with @throw@ or
592@throwResume@, respectively.
594try {
595        ...
596} catch( ... ) {
597        ... throw;
598} catchResume( ... ) {
599        ... throwResume;
602The only difference between a raise and a reraise is that reraise does not
603create a new exception; instead it continues using the current exception, \ie
604no allocation and copy. However the default handler is still set to the one
605visible at the raise point, and hence, for termination could refer to data that
606is part of an unwound stack frame. To prevent this problem, a new default
607handler is generated that does a program-level abort.
610\subsection{Comparison with Reraising}
611Without conditional catch, the only approach to match in more detail is to reraise
612the exception after it has been caught, if it could not be handled.
616try {
617        do_work_may_throw();
618} catch(excep_t * ex; can_handle(ex)) {
620        handle(ex);
628try {
629        do_work_may_throw();
630} catch(excep_t * ex) {
631        if (can_handle(ex)) {
632                handle(ex);
633        } else {
634                throw;
635        }
640Notice catch-and-reraise increases complexity by adding additional data and
641code to the exception process. Nevertheless, catch-and-reraise can simulate
642conditional catch straightforwardly, when exceptions are disjoint, \ie no
645However, catch-and-reraise simulation becomes unusable for exception inheritance.
648exception E1;
649exception E2(E1); // inheritance
653try {
654        ... foo(); ... // raise E1/E2
655        ... bar(); ... // raise E1/E2
656} catch( E2 e; e.rtn == foo ) {
657        ...
658} catch( E1 e; e.rtn == foo ) {
659        ...
660} catch( E1 e; e.rtn == bar ) {
661        ...
667try {
668        ... foo(); ...
669        ... bar(); ...
670} catch( E2 e ) {
671        if ( e.rtn == foo ) { ...
672        } else throw; // reraise
673} catch( E1 e ) {
674        if (e.rtn == foo) { ...
675        } else if (e.rtn == bar) { ...
676        else throw; // reraise
681The derived exception @E2@ must be ordered first in the catch list, otherwise
682the base exception @E1@ catches both exceptions. In the catch-and-reraise code
683(right), the @E2@ handler catches exceptions from both @foo@ and
684@bar@. However, the reraise misses the following catch clause. To fix this
685problem, an enclosing @try@ statement is need to catch @E2@ for @bar@ from the
686reraise, and its handler must duplicate the inner handler code for @bar@. To
687generalize, this fix for any amount of inheritance and complexity of try
688statement requires a technique called \emph{try-block
689splitting}~\cite{Krischer02}, which is not discussed in this thesis. It is
690sufficient to state that conditional catch is more expressive than
691catch-and-reraise in terms of complexity.
694That is, they have the same behaviour in isolation.
695Two things can expose differences between these cases.
697One is the existence of multiple handlers on a single try statement.
698A reraise skips all later handlers for a try statement but a conditional
699catch does not.
700% Hence, if an earlier handler contains a reraise later handlers are
701% implicitly skipped, with a conditional catch they are not.
702Still, they are equivalently powerful,
703both can be used two mimic the behaviour of the other,
704as reraise can pack arbitrary code in the handler and conditional catches
705can put arbitrary code in the predicate.
706% I was struggling with a long explanation about some simple solutions,
707% like repeating a condition on later handlers, and the general solution of
708% merging everything together. I don't think it is useful though unless its
709% for a proof.
712The question then becomes ``Which is a better default?"
713We believe that not skipping possibly useful handlers is a better default.
714If a handler can handle an exception it should and if the handler can not
715handle the exception then it is probably safer to have that explicitly
716described in the handler itself instead of implicitly described by its
717ordering with other handlers.
718% Or you could just alter the semantics of the throw statement. The handler
719% index is in the exception so you could use it to know where to start
720% searching from in the current try statement.
721% No place for the `goto else;` metaphor.
723The other issue is all of the discussion above assumes that the only
724way to tell apart two raises is the exception being raised and the remaining
725search path.
726This is not true generally, the current state of the stack can matter in
727a number of cases, even only for a stack trace after an program abort.
728But \CFA has a much more significant need of the rest of the stack, the
729default handlers for both termination and resumption.
731% For resumption it turns out it is possible continue a raise after the
732% exception has been caught, as if it hadn't been caught in the first place.
733This becomes a problem combined with the stack unwinding used in termination
734exception handling.
735The stack is unwound before the handler is installed, and hence before any
736reraises can run. So if a reraise happens the previous stack is gone,
737the place on the stack where the default handler was supposed to run is gone,
738if the default handler was a local function it may have been unwound too.
739There is no reasonable way to restore that information, so the reraise has
740to be considered as a new raise.
741This is the strongest advantage conditional catches have over reraising,
742they happen before stack unwinding and avoid this problem.
744% The one possible disadvantage of conditional catch is that it runs user
745% code during the exception search. While this is a new place that user code
746% can be run destructors and finally clauses are already run during the stack
747% unwinding.
750%   `exception_ptr current_exception() noexcept;`
754\section{Finally Clauses}
756Finally clauses are used to preform unconditional clean-up when leaving a
757scope and are placed at the end of a try statement after any handler clauses:
759try {
760        GUARDED_BLOCK
761} ... // any number or kind of handler clauses
762... finally {
763        FINALLY_BLOCK
766The @FINALLY_BLOCK@ is executed when the try statement is removed from the
767stack, including when the @GUARDED_BLOCK@ finishes, any termination handler
768finishes, or during an unwind.
769The only time the block is not executed is if the program is exited before
770the stack is unwound.
772Execution of the finally block should always finish, meaning control runs off
773the end of the block. This requirement ensures control always continues as if
774the finally clause is not present, \ie finally is for cleanup not changing
775control flow.
776Because of this requirement, local control flow out of the finally block
777is forbidden. The compiler precludes any @break@, @continue@, @fallthru@ or
778@return@ that causes control to leave the finally block. Other ways to leave
779the finally block, such as a long jump or termination are much harder to check,
780and at best requiring additional run-time overhead, and so are only
783Not all languages with unwinding have finally clauses. Notably \Cpp does
784without it as destructors, and the RAII design pattern, serve a similar role.
785Although destructors and finally clauses can be used for the same cases,
786they have their own strengths, similar to top-level function and lambda
787functions with closures.
788Destructors take more work for their creation, but if there is clean-up code
789that needs to be run every time a type is used, they are much easier
790to set-up.
791On the other hand finally clauses capture the local context, so is easy to
792use when the clean-up is not dependent on the type of a variable or requires
793information from multiple variables.
797Cancellation is a stack-level abort, which can be thought of as as an
798uncatchable termination. It unwinds the entire current stack, and if
799possible forwards the cancellation exception to a different stack.
801Cancellation is not an exception operation like termination or resumption.
802There is no special statement for starting a cancellation; instead the standard
803library function @cancel_stack@ is called passing an exception. Unlike a
804raise, this exception is not used in matching only to pass information about
805the cause of the cancellation.
806Finaly, since a cancellation only unwinds and forwards, there is no default handler.
808After @cancel_stack@ is called the exception is copied into the EHM's memory
809and the current stack is unwound.
810The behaviour after that depends on the kind of stack being cancelled.
812\paragraph{Main Stack}
813The main stack is the one used by the program main at the start of execution,
814and is the only stack in a sequential program.
815After the main stack is unwound there is a program-level abort.
817The reasons for this semantics in a sequential program is that there is no more code to execute.
818This semantics also applies to concurrent programs, too, even if threads are running.
819That is, if any threads starts a cancellation, it implies all threads terminate.
820Keeping the same behaviour in sequential and concurrent programs is simple.
821Also, even in concurrent programs there may not currently be any other stacks
822and even if other stacks do exist, main has no way to know where they are.
824\paragraph{Thread Stack}
825A thread stack is created for a \CFA @thread@ object or object that satisfies
826the @is_thread@ trait.
827After a thread stack is unwound, the exception is stored until another
828thread attempts to join with it. Then the exception @ThreadCancelled@,
829which stores a reference to the thread and to the exception passed to the
830cancellation, is reported from the join to the joining thread.
831There is one difference between an explicit join (with the @join@ function)
832and an implicit join (from a destructor call). The explicit join takes the
833default handler (@defaultResumptionHandler@) from its calling context while
834the implicit join provides its own; which does a program abort if the
835@ThreadCancelled@ exception cannot be handled.
837The communication and synchronization are done here because threads only have
838two structural points (not dependent on user-code) where
839communication/synchronization happens: start and join.
840Since a thread must be running to perform a cancellation (and cannot be
841cancelled from another stack), the cancellation must be after start and
842before the join, so join is used.
844% TODO: Find somewhere to discuss unwind collisions.
845The difference between the explicit and implicit join is for safety and
846debugging. It helps prevent unwinding collisions by avoiding throwing from
847a destructor and prevents cascading the error across multiple threads if
848the user is not equipped to deal with it.
849It is always possible to add an explicit join if that is the desired behaviour.
851With explicit join and a default handler that triggers a cancellation, it is
852possible to cascade an error across any number of threads, cleaning up each
853in turn, until the error is handled or the main thread is reached.
855\paragraph{Coroutine Stack}
856A coroutine stack is created for a @coroutine@ object or object that
857satisfies the @is_coroutine@ trait.
858After a coroutine stack is unwound, control returns to the @resume@ function
859that most recently resumed it. @resume@ reports a
860@CoroutineCancelled@ exception, which contains a references to the cancelled
861coroutine and the exception used to cancel it.
862The @resume@ function also takes the \defaultResumptionHandler{} from the
863caller's context and passes it to the internal report.
865A coroutine only knows of two other coroutines, its starter and its last resumer.
866The starter has a much more distant connection, while the last resumer just
867(in terms of coroutine state) called resume on this coroutine, so the message
868is passed to the latter.
870With a default handler that triggers a cancellation, it is possible to
871cascade an error across any number of coroutines, cleaning up each in turn,
872until the error is handled or a thread stack is reached.
874\PAB{Part of this I do not understand. A cancellation cannot be caught. But you
875talk about handling a cancellation in the last sentence. Which is correct?}
Note: See TracBrowser for help on using the repository browser.