10.2 Exceptions

Exceptions in The Racket Guide introduces exceptions.

See Exceptions for information on the Racket exception model. It is based on a proposal by Friedman, Haynes, and Dybvig [Friedman95].

Whenever a primitive error occurs in Racket, an exception is raised. The value that is passed to the current exception handler for a primitive error is always an instance of the exn structure type. Every exn structure value has a message field that is a string, the primitive error message. The default exception handler recognizes exception values with the exn? predicate and passes the error message to the current error display handler (see error-display-handler).

Primitive procedures that accept a procedure argument with a particular required arity (e.g., call-with-input-file, call/cc) check the argument’s arity immediately, raising exn:fail:contract if the arity is incorrect.

10.2.1 Error Message Conventions

Racket’s error message convention is to produce error messages with the following shape:

‹srcloc›: ‹name›: ‹message›;

‹continued-message› ...

‹field›: ‹detail›

...

The message starts with an optional source location, ‹srcloc›, which is followed by a colon and space when present. The message continues with an optional ‹name› that usually identifies the complaining function, syntactic form, or other entity, but may also refer to an entity being complained about; the ‹name› is also followed by a colon and space when present.

The ‹message› should be relatively short, and it should be largely independent of specific values that triggered the error. More detailed explanation that requires multiple lines should continue with each line indented by a single space, in which case ‹message› should end in a semi-colon (but the semi-colon should be omitted if ‹continued-message› is not present). Message text should be lowercase—using semi-colons to separate sentences if needed, although long explanations may be better deferred to extra fields.

Specific values that triggered the error or other helpful information should appear in separate ‹field› lines, each of which is indented by two spaces. If a ‹detail› is especially long or takes multiple lines, it should start on its own line after the ‹field› label, and each of its lines should be indented by three spaces. Field names should be all lowercase.

A ‹field› name should end with ... if the field provides relatively detailed information that might be distracting in common cases but useful in others. For example, when a contract failure is reported for a particular argument of a function, other arguments to the function might be shown in an “other arguments...” field. The intent is that fields whose names end in ... might be hidden by default in an environment such as DrRacket.

Make ‹field› names as short as possible, relying on ‹message› or ‹continued message› text to clarify the meaning for a field. For example, prefer “given” to “given turtle” as a field name, where ‹message› is something like “given turtle is too sleepy” to clarify that “given” refers to a turtle.

10.2.2 Raising Exceptions

If barrier? is true, then the call to the exception handler is protected by a continuation barrier, so that multiple returns/escapes are impossible. All exceptions raised by racket functions effectively use raise with a #t value for barrier?.

Breaks are disabled from the time the exception is raised until the exception handler obtains control, and the handler itself is parameterize-breaked to disable breaks initially; see Breaks for more information on breaks.

• (error message-sym) creates a message string by concatenating "error: " with the string form of message-sym. Use this form sparingly.

• (error message-str v ...) creates a message string by concatenating message-str with string versions of the vs (as produced by the current error value conversion handler; see error-value->string-handler). A space is inserted before each v. Use this form sparingly, because it does not conform well to Racket’s error message conventions; consider raise-arguments-error, instead.

• (error who-sym format-str v ...) creates a message string equivalent to the string created by

  (format (string-append "~s: " format-str) who-sym v ...)

  When possible, use functions such as raise-argument-error, instead, which construct messages that follow Racket’s error message conventions.

In all cases, the constructed message string is passed to make-exn:fail, and the resulting exception is raised.

In the first form, v is the value received by the procedure that does not have the expected type.

In the second form, the bad argument is indicated by an index bad-pos (counting from 0), and all of the original arguments v are provided (in order). The resulting error message names the bad argument and also lists the other arguments. If bad-pos is not less than the number of vs, the exn:fail:contract exception is raised.

Since upper-bound is inclusive, a typical value is one less than the size of a collection—for example, (sub1 (vector-length vec)), (sub1 (length lst)), and so on.

The arity-v value must be a possible result from procedure-arity, except that it does not have to be normalized (see procedure-arity? for the details of normalized arities); raise-arity-error will normalize the arity and use the normalized form in the error message. If name is a procedure, its actual arity is ignored.

The arg-v arguments are the actual supplied arguments, which are shown in the error message (using the error value conversion handler; see error-value->string-handler); also, the number of supplied arg-vs is explicitly mentioned in the message.

Added in version 7.0.0.11 of package base.

Added in version 6.90.0.26 of package base.

The name argument is usually #f when expr is provided; it is described in more detail below. The message is used as the main body of the error message; if message contains newline characters, each new line should be suitably indented (with one space at the start), and it should not end with a newline character.

The optional expr argument is the erroneous source syntax object or S-expression (but the expression #f cannot be represented by itself; it must be wrapped as a syntax object). The optional sub-expr argument is a syntax object or S-expression (again, #f cannot represent itself) within expr that more precisely locates the error. Both may appear in the generated error-message text if error-print-source-location is #t. Source location information in the error-message text is similarly extracted from sub-expr or expr when at least one is a syntax object and error-print-source-location is #t.

If sub-expr is provided and not #f, it is used (in syntax form) for the exprs field of the generated exception record, else the expr is used if provided and not #f. In either case, the syntax object is consed onto extra-sources to produce the exprs field, or extra-sources is used directly for exprs if neither expr nor sub-expr is provided and not #f.

The form name used in the generated error message is determined through a combination of the name, expr, and sub-expr arguments:

• When name is #f, and when expr is either an identifier or a syntax pair containing an identifier as its first element, then the form name from the error message is the identifier’s symbol.

• When name is #f and when expr is not an identifier or a syntax pair containing an identifier as its first element, then the form name in the error message is "?".

• When name is a symbol, then the symbol is used as the form name in the generated error message.

The message-suffix string is appended to the end of the error message. If not "", it should normally start with a newline and two spaces to add extra fields to the message (see Error Message Conventions).

Changed in version 6.90.0.18 of package base: Added the message-suffix optional argument.

The unquoted-printing-string? procedure returns #t if v is a unquoted-printing string, #f otherwise. The unquoted-printing-string creates a unquoted-printing string value that encapsulates the string s, and unquoted-printing-string-value returns the string within a unquoted-printing string.

Added in version 6.10.0.2 of package base.

10.2.3 Handling Exceptions

Any procedure that takes one argument can be an exception handler. Normally, an exception handler escapes from the context of the raise call via abort-current-continuation or some other escape mechanism. To propagate an exception to the “previous” exception handler—that is, the exception handler associated with the rest of the continuation after the point where the called exception handler was associated with the continuation—an exception handler can simply return a result instead of escaping, in which case the raise call propagates the value to the previous exception handler (still in the dynamic extent of the call to raise, and under the same barrier, if any). If an exception handler returns a result and no previous handler is available, the uncaught-exception handler is used.

A call to an exception handler is parameterize-breaked to disable breaks, and it is wrapped with call-with-exception-handler to install an exception handler that reports both the original and newly raised exceptions via the error display handler and then escapes via the error escape handler.

The default uncaught-exception handler prints an error message using the current error display handler (see error-display-handler), unless the argument to the handler is an instance of exn:break:hang-up. If the argument to the handler is an instance of exn:break:hang-up or exn:break:terminate, the default uncaught-exception handler then calls the exit handler with 1, which normally exits or escapes. For any argument, the default uncaught-exception handler then escapes by calling the current error escape handler (see error-escape-handler). The call to each handler is parameterized to set error-display-handler to the default error display handler, and it is parameterize-breaked to disable breaks. The call to the error escape handler is further parameterized to set error-escape-handler to the default error escape handler; if the error escape handler returns, then the default error escape handler is called.

When the current error display handler is the default handler, then the error-display call is parameterized to install an emergency error display handler that logs an error (see log-error) and never fails.

The new exception handler processes an exception only if one of the pred-expr procedures returns a true value when applied to the exception, otherwise the exception handler is invoked from the continuation of the with-handlers expression (by raising the exception again). If an exception is handled by one of the handler-expr procedures, the result of the entire with-handlers expression is the return value of the handler.

When an exception is raised during the evaluation of bodys, each predicate procedure pred-expr is applied to the exception value; if a predicate returns a true value, the corresponding handler-expr procedure is invoked with the exception as an argument. The predicates are tried in the order that they are specified.

Before any predicate or handler procedure is invoked, the continuation of the entire with-handlers expression is restored, but also parameterize-breaked to disable breaks. Thus, breaks are disabled by default during the predicate and handler procedures (see Breaks), and the exception handler is the one from the continuation of the with-handlers expression.

The exn:fail? procedure is useful as a handler predicate to catch all error exceptions. Avoid using (lambda (x) #t) as a predicate, because the exn:break exception typically should not be caught (unless it will be re-raised to cooperatively break). Beware, also, of catching and discarding exceptions, because discarding an error message can make debugging unnecessarily difficult; instead of discarding an error message, consider logging it via log-error or a logging form created by define-logger.

10.2.4 Configuring Default Handling

The error escape handler is normally called directly by an exception handler, in a parameterization that sets the error display handler and error escape handler to the default handlers, and it is normally parameterize-breaked to disable breaks. To escape from a run-time error in a different context, use raise or error.

Due to a continuation barrier around exception-handling calls, an error escape handler cannot invoke a full continuation that was created prior to the exception, but it can abort to a prompt (see call-with-continuation-prompt) or invoke an escape continuation (see call-with-escape-continuation).

The default error display handler displays its first argument to the current error port (determined by the current-error-port parameter) and extracts a stack trace (see continuation-mark-set->context) to display from the second argument if it is an exn value but not an exn:fail:user value.

The default error display handler in DrRacket also uses the second argument to highlight source locations.

To report a run-time error, use raise or procedures like error, instead of calling the error display handler directly.

The integer argument to the handler specifies the maximum number of characters that should be used to represent the value in the resulting string. The default error value conversion handler prints the value into a string (using the current global port print handler; see global-port-print-handler). If the printed form is too long, the printed form is truncated and the last three characters of the return string are set to “...”.

If the string returned by an error value conversion handler is longer than requested, the string is destructively “truncated” by setting the first extra position in the string to the null character. If a non-string is returned, then the string "..." is used. If a primitive error string needs to be generated before the handler has returned, the default error value conversion handler is used.

Calls to an error value conversion handler are parameterized to re-install the default error value conversion handler, and to enable printing of unreadable values (see print-unreadable).

10.2.5 Built-in Exception Types

Exceptions raised by Racket form a hierarchy under exn:

exn

exn:fail

exn:fail:contract

exn:fail:contract:arity

exn:fail:contract:divide-by-zero

exn:fail:contract:non-fixnum-result

exn:fail:contract:continuation

exn:fail:contract:variable

exn:fail:syntax

exn:fail:syntax:unbound

exn:fail:syntax:missing-module

exn:fail:read

exn:fail:read:eof

exn:fail:read:non-char

exn:fail:filesystem

exn:fail:filesystem:exists

exn:fail:filesystem:version

exn:fail:filesystem:errno

exn:fail:filesystem:missing-module

exn:fail:network

exn:fail:network:errno

exn:fail:out-of-memory

exn:fail:unsupported

exn:fail:user

exn:break

exn:break:hang-up

exn:break:terminate

This structure type implements the prop:exn:srclocs property.

The default module name resolver raises this exception only when it is given a syntax object as its second argument, and the default load handler raises this exception only when the value of current-module-path-for-load is a syntax object (in which case both the exprs field and the path field are determined by the syntax object.

This structure type implements the prop:exn:missing-module property.

The default module name resolver raises this exception only when it is not given a syntax object as its second argument, and the default load handler raises this exception only when the value of current-module-path-for-load is not a syntax object.

This structure type implements the prop:exn:missing-module property.

The property value must be a procedure that accepts a single value—the structure type instance from which to extract source locations—and returns a list of srclocs. Some error display handlers use only the first returned location.

• source — An arbitrary value identifying the source, often a path (see Paths).

• line — The line number (counts from 1) or #f (unknown).

• column — The column number (counts from 0) or #f (unknown).

• position — The starting position (counts from 1) or #f (unknown).

• span — The number of covered positions (counts from 0) or #f (unknown).

See Printing Compiled Code for information about the treatment of srcloc values that are embedded in compiled code.

The property value must be a procedure that accepts a single value—the structure type instance from which to extract source locations—and returns a module path.

10.2.6 Additional Exception Functions

Added in version 6.3 of package base.