This is a WG2 proposal for condition objects. A condition object encapsulates information about an exceptional situation. Typically the rest of the system is notified about a condition using the raise and raise-continuable procedures and their relatives. Conditions are logically independent of the exception system, however: conditions may be used for any purpose, and any object may be passed to the exception system.
The design of this condition system attempts to assume as little as possible about any existing implementation-specific condition system. In particular, there is no specified relationship between conditions and records, as there is in R6RS, nor is there any notion of subtyping required by the system. There are condition types for convenience in dispatching, but they are just symbols and in general entail nothing about what information is encapsulated, as different implementations will provide different kinds of information when creating an implementation-specified condition.
Within the above constraints, I have attempted to make the names as compatible as possible with the R6RS condition system and its predecessors SRFI 35 and SRFI 36, and with SRFI 12.
A condition is said to be "belong to type sym" if (a) it was created by a call to make-condition that was passed the symbol sym as one of its condition types, or (b) it belongs to an implementation-defined set (possibly empty) of conditions of type sym. This allows implementation-dependent condition objects to participate in this condition system.
(make-condition symlist ( sym obj ) ...)
Returns a newly allocated condition which belongs to the types whose names are given in symlist. The remainder of the arguments are alternating property names (which are symbols) and values (which can be any object) that specify the information encapsulated by this condition. It is an error if the value associated with the property name message is not a string; it is also an error if the value associated with the property name irritants is not a list.
(condition? obj)
Returns #t if obj is a condition of any type, and #f otherwise.
(condition-of-type? obj sym)
Returns #t if obj is a condition belonging to type sym, and #f otherwise.
(<type>-error? obj)
Returns #t if obj is a condition belonging to type <type> from the list below, and #f otherwise.
(condition-types condition)
Returns the list of types to which condition belongs. It is an error to attempt to mutate this list.
(condition-properties condition)
Returns the list of property names associated with this condition. It is an error to attempt to mutate this list.
(condition-ref condition sym [ default ] )
Returns the property value associated with the property named sym of condition. If it has no such property, returns default. If default is not specified, returns #f.
(condition-predicate sym)
Returns a predicate which will return #t if applied to a condition belonging to type sym, and #f otherwise.
(condition-accessor condition sym [ default ] )
Returns an accessor which will return the value of sym if applied to a condition object and default otherwise. If default is not specified, the accessor will return #f.
(condition-<property-name> condition)
Returns the value of the property named <property-name> (from the standard list of property names below) of condition, or #f if it has no such property.
(error-object? obj)
Returns #t if obj is a condition belonging to type simple, and #f otherwise. Such conditions are normally created only by user code. Part of the small language, but shown here for completeness.
(file-error? obj)
Returns #t if obj is a condition belonging to type file, and #f otherwise. Such conditions may be created by the implementation if there is an error related to file operations; in particular, the inability to open a file for input. Part of the small language, but shown here for completeness.
(read-error? obj)
Returns #t if obj is a condition belonging to type read, and #f otherwise. Such conditions may be created by the implementation if there is an error related to read, such as a lexical syntax error in the input. Part of the small language, but shown here for completeness.
(syntax-error? obj)
Returns #t if obj is a condition belonging to type syntax, and #f otherwise. Such conditions may be created by the implementation if program code is syntactically ill-formed. When such a condition is raised, it may or may not be possible for the exception system to catch it. Part of the small language, but shown here for completeness.
(implementation-restriction? obj)
Returns #t if obj is a condition belonging to type implementation-restriction, and #f otherwise. Such conditions may be created by the implementation if one of its restrictions is exceeded, such as consuming too much memory or trying to compute an exact number too large to represent.
The following condition types are standardized. Conditions of each type may be created by the implementation in the specified situations as well as any analogous situations. The list is intended to be comprehensive but not complete: it draws on R6RS, Java, and other sources.
simple |
error created by error |
assert |
error created by assert |
file |
file-related error |
filename |
mangled filename |
protection |
file protection error |
input |
input error |
output |
output error |
network |
socket or network error |
closed |
I/O operation on closed port |
already-exists |
file already exists |
not-found |
file not found |
read |
textual error during reading |
lexical |
lexical syntax error |
eof |
EOF inside a lexical construct |
syntax |
Scheme syntax error |
domain |
argument has wrong type or value |
type |
wrong type |
value |
wrong value |
state |
invalid state event |
arithmetic |
arithmetic error |
divide |
division by exact zero |
fixnum |
sufficiently small exact integer expected |
conversion |
attempted impossible conversion |
improper-list |
improper lists not supported |
circular-list |
circular lists not supported |
arity |
too many or too few arguments |
immutability |
modifying immutable data |
undefined |
getting the value of a variable that has not been defined |
non-continuable |
continuing from an exception raised by raise |
match |
unsatisfied pattern match |
uncaught |
uncaught exception |
nonexistent |
reference to something that does not exist |
range |
violation of start-end conditions |
scheduler |
task scheduler error |
deadlock |
scheduler deadlock |
timeout |
operation timed out |
termination |
thread termination |
concurrency |
invalid concurrent modification |
invalid-position |
invalid file position |
encoding |
encoding or decoding error |
os |
operating system reported error |
implementation-restriction |
implementation restriction |
memory |
out of memory |
unsupported |
unsupported operation |
no-infinities |
implementation does not support infinities |
no-nans |
implementation does not support NaNs |
continuation |
escape procedure invoked when not supported |
bignum |
exact number too large to represent |
version-skew |
mismatched versions of code |
security |
security violation |
implementation-error |
low-level failure |
message |
human-readable description string |
irritants |
list of problematic arguments |
who |
an object reporting a problem with another object |
what |
an object which has a problem |
position |
the position in what at which the problem occurred |