NAME¶
Lexical::Failure - User-selectable lexically-scoped failure signaling
VERSION¶
This document describes Lexical::Failure version 0.000006
SYNOPSIS¶
package Your::Module;
# Set up this module for lexical failure handling...
use Lexical::Failure;
# Each time module is imported, set up failure handler...
sub import {
my ($package, %named_arg) = @_;
ON_FAILURE( $named_arg{'fail'} );
}
# Then, in the module's subs/methods, call fail() to fail...
sub inverse_square {
my ($n) = @_;
if ($n == 0) {
fail "Can't invert zero";
}
return 1/$n**2;
}
sub load_file {
my ($filename) = @_;
fail 'No such file: ', $filename
if ! -r $filename;
local (@ARGV, $/) = $filename;
return readline;
}
DESCRIPTION¶
This module sets up two new keywords: "fail" and
"ON_FAILURE", with which you can quickly create modules whose
failure signaling is lexicially scoped, under the control of client code.
Normally, modules specify some fixed mechanism for error handling and require
client code to adapt to that policy. One module may signal errors by returning
"undef", or perhaps some special "error object". Another
may "die" or "croak" on failure. A third may set a flag
variable. A fourth may require the client code to set up a callback, which is
executed on failure.
If you are using all four modules, your own code now has to check for failure in
four different ways, depending on where the failing component originated. If
you would rather that
all components throw exceptions, or all return
"undef", you will probably have to write wrappers around 3/4 of
them, to convert from their "native" failure mechanism to your
preferred one.
Lexical::Failure offers an alternative: a simple mechanism with which module
authors can generically specify "fail here with this message" (using
the "fail" keyword), but then allow each block of client code to
decide how that failure is reported to it within its own lexical scope (using
the "ON_FAILURE" keyword).
Module authors can still provide a default failure signaling mechanism, for when
client code does not specify how errors are to be reported. This is handy for
ensuring backwards compatibility in existing modules that are converted to
this new failure signaling approach.
INTERFACE¶
Accessing the API¶
To install the new "fail" and "ON_FAILURE" keywords, simple
load the module:
use Lexical::Failure;
Changing the names of the API keywords
To avoid name conflicts, you can change the name of either (or both) of the
keywords that the module sets up, by passing a named argument when loading the
module. The name of the argument should be the standard name of the keyword
you want to rename, and the value of the argument should be a string
containing the new name. For example:
use Lexical::Failure (
fail => 'return_error',
ON_FAILURE => 'set_error_handler',
);
sub import {
my ($package, %named_arg) = @_;
set_error_handler( $named_arg{'fail'} );
}
sub inverse_square {
my ($n) = @_;
return_error "Can't invert zero" if $n == 0;
return 1/$n**2;
}
Signaling failure with "fail"¶
Once the module is loaded, you simply use the "fail" keyword in place
of "return", "return undef", "die",
"croak", "confess", or any other mechanism by which you
would normally indicate failure.
You can call "fail" with any number of arguments, including none, and
these will be passed to whichever failure handler the client code eventually
selects (see below).
Note that "fail" is a keyword, not a subroutine (that is, it's like
"return" itself, and not something you can call as part of a larger
expression).
Specifying a lexically scoped failure handler with "ON_FAILURE"¶
You set up a failure-signaling interface for client code by placing the
"ON_FAILURE" keyword in your module's "import()"
subroutine (or in a subroutine called from your "import()").
The keyword expects one argument, which specifies how failures in the module are
to be handled in the lexical scope where your module was loaded. The single
argument can be:
- •
- a string containing the name of a named failure handler
- •
- a reference to a variable, into which failure signals will be stored
- •
- a reference to a subroutine, which will be used as a callback and invoked
whenever a failure is to be signalled
- •
- "undef" (or no argument at all), in which case
"ON_FAILURE" does nothing. This means you don't need to bother
checking whether a failure specifier was passed in to your
"import()". Just pass in the resulting "undef"
value...and it's ignored.
Typically, then, you have your "import()" subroutine accept an
argument through which client code indicates its desired failure mode:
package Your::Module;
sub import {
my ($package, %named_arg) = @_;
ON_FAILURE $named_arg{'fail'};
}
Then the client code can specify different reporting strategies in different
lexical scopes:
# Hereafter, report failures by returning undef...
use Your::Module fail => 'undef';
{
# But in this block, make errors fatal...
use Your::Module fail => 'croak';
{
# And in here, set a flag...
my $nested_error_flag;
use Your::Module fail => \$nested_error_flag;
{
# And in here, any error is quietly loggged...
use Your::Module fail => sub { $logger->error(@_) };
}
}
# Back to croaking errors here
}
# Back to returning undef here
Each "use Your::Module" invokes "Your::Module::import()",
whereupon the call to "ON_FAILURE" installs the specified failure
handler into the lexical scope in which "use Your::Module" occurred.
The installed handler is specific to Your::Module, so if two or more modules
are each using Lexical::Failure, client code can set failure-signaling
policies for each module independently in the same scope.
Named failure handlers
If "ON_FAILURE" is passed a string, that string is treated as the name
of a predefined failure handler. Lexical::Failure provides six standard named
handlers:
- "ON_FAILURE 'null'"
- Specifies that each "fail @args" should act like:
return;
That is: return "undef" in scalar context or return an empty list
in list context.
Note that, this context-sensitive behaviour can occasionally lead to subtle
errors. For example, if these three subroutines are using "ON_FAILURE
'null'" failure signaling:
my %personal_data = (
name => get_name(),
age => get_age(),
status => get_status(),
);
then if any of them fails, it will return an empty list, messing up the
initialization of the hash. In such cases, "ON_FAILURE 'undef'"
is a better alternative.
- "ON_FAILURE 'undef'"
- Specifies that each "fail @args" should act like:
return undef;
Note that to get this behaviour, the argument needs to be 'undef' (a five
letter string), not "undef" (the special undefined value).
Note too that, when this handler is selected, "fail" returns an
"undef" even in list context. This can be problematical, as an
"undef" is (to many people's surprise) true in list
context. For example, if "get_results()" returns
"undef" on failure, the conditional test of this "if"
will still be true:
if (my @results = get_results($data)) {
....
}
because @results will then contain one element (the "undef"), and
a non-empty array always evaluates true in boolean context.
For this reason it's usually better to use "ON_FAILURE 'null'"
instead.
- "ON_FAILURE 'die'"
- Specifies that each "fail @args" should act like:
die @args;
- "ON_FAILURE 'croak'"
- Specifies that each "fail @args" should act like:
Carp::croak(@args);
- "ON_FAILURE 'confess'"
- Specifies that each "fail @args" should act like:
Carp::confess(@args);
- "ON_FAILURE 'failobj'"
- Specifies that each "fail @args" should act like:
return Lexical::Failure::Objects->new(
msg => ( @args == 1 ? $args[0] : "@args" ),
context => [caller 1]
);
In other words, "ON_FAILURE 'failobj'" causes "fail" to
return a special object encapsulating the arguments passed to
"fail" and the call context in which the "fail"
occurred.
See the documentation of Lexical::Failure::Objects for more details on this
alternative.
You can also set up other named failure handlers of your own devising (see
"Specifying additional named failure handlers").
Variables as failure handlers
If "ON_FAILURE" is passed a reference to a scalar, array, or hash,
that variable becomes the "receiver" of subsequent failure reports,
as follows:
- "ON_FAILURE \$scalar"
- Specifies that "fail @args" should act like:
$scalar = [@args];
return undef;
- "ON_FAILURE \@array"
- Specifies that "fail @args" should act like:
push @array, [@args];
return undef;
- "ON_FAILURE \%hash"
- Specifies that "fail @args" should act like:
$hash{ $CURRENT_SUBNAME } = [@args];
return undef;
Subroutines as failure handlers
"ON_FAILURE" can also be passed a reference to a subroutine, which
then acts like a callback when failures are signalled.
In other words:
ON_FAILURE $subroutine_ref;
causes "fail @args" to act like:
return $subroutine_ref->(@args);
The availability of this alternative means that client code can create entirely
new failure-signaling behaviours whenever needed. For example:
# Signal failure by logging an error and returning negatively...
use Your::Module fail => sub { $logger->error(@_); return -1; };
# Signal failure by returning undef/empty list,
# except in one critical case...
use Your::Module fail => sub {
my $msg = "@_";
croak $msg if $msg =~ /dangerous/;
return;
};
# The very first failure is instantly (and unluckily) fatal...
use Your::Module fail => sub { carp(@_); exit(13) };
Restricting how client code can signal failure
Because the call to "ON_FAILURE" must occur in your module's
"import()" subroutine, you always have ultimate control over what
types of failure signaling the client code may request from your module.
For example, to prevent client code from requesting "return undef"
behaviours:
sub import {
my ($package, %named_arg) = @_;
croak "Can't specify 'undef' as a failure handler"
if $named_arg{'fail'} eq 'undef';
ON_FAILURE $named_arg{'fail'};
}
or to quietly convert 'die' behaviours into (much more useful) 'croak'
behaviours:
sub import {
my ($package, %named_arg) = @_;
$named_arg{'fail'} =~ s/^die$/croak/;
ON_FAILURE $named_arg{'fail'};
}
Specifying a module's default failure handler
In any scope where no explicit failure signaling behaviour has been specified,
Lexical::Failure defaults to its standard 'croak' behaviour (see "Named
failure handlers").
However, you can also specify a different default for your module, by adding a
named argument when you load Lexical::Failure:
# Default to full confession on failure...
use Lexical::Failure default => 'confess';
# Default to 'return undef or empty list' on failure...
use Lexical::Failure default => 'null';
# Default to instant fatality on failure...
use Lexical::Failure default => sub { carp(@_); exit() };
The values allowed for the 'default' option are somewhat more restrictive than
those which can be passed directly to "ON_FAILURE"; you can specify
only standard named handlers (see "Named failure handlers") or a
subroutine reference.
If you need your default to be a non-standard named handler (see
"Specifying additional named failure handlers") or a reference to a
variable, you must arrange that in your "import()" instead. For
example:
sub import {
my ($package, %named_arg) = @_;
# Install failure signaling, if specified...
if (defined $named_arg{'fail'}) {
ON_FAILURE $named_arg{'fail'};
}
# Otherwise, default to pushing errors onto a package variable
# (yeah, this is a HORRIBLE idea, but it's what our boss decided!)
else {
ON_FAILURE \@Your::Module::errors;
}
}
Specifying additional named failure handlers
The six standard named failure handlers provide convenient declarative shortcuts
for client code. That is, instead of constantly having to create messy
subroutines like:
use Your::Module
fail => sub {
return Lexical::Failure::Objects->new(
msg => (@_ == 1 ? $_[0] : "@_"),
context => [caller 1],
);
};
client code can just request:
use Your::Module fail => 'failobj';
However, you may wish to offer a similar declarative interface for other
failure-signaling behaviours that your client code is likely to need. For
example:
use Your::Module fail => 'logged';
use Your::Module fail => 'exit';
use Your::Module fail => 'loud undef';
Lexical::Failure provides a simple way to set up extra named handlers like
these. You just specify the name and associated callback for each when loading
the module:
package Your::Module;
use Lexical::Failure handlers => {
'logged' => sub { $logger->error(@_); },
'exit' => sub { say @_; exit; },
'loud undef' => sub { carp(@_); return undef; },
};
The 'handlers' option expects a reference to a hash, in which each key is the
name of a new named failure handler, and each corresponding value is a
reference to a subroutine implementing the behaviour of that named handler.
Once specified, any of the new handler names may be passed to
"ON_FAILURE" to specify that "fail" should use the
corresponding callback to signal failures.
Note that any extra named handlers defined in this way are only available from
the module in which they are defined.
DIAGNOSTICS¶
- "Unknown failure handler: %s"
- You called "ON_FAILURE" with a string as the handler
specification. However, that string was not one of the standard named
handlers ('confess', 'croak', 'die', 'failobj', 'undef', or 'null'), nor
any of the extra handlers you may have specified with a 'handlers' option
when loading Lexical::Failure.
Did you perhaps misspell the handler name?
- "Unknown default failure handler: %s"
- When loading Lexical::Failure, you specified a default handler for all
scopes like so:
use Lexical::Failure default => 'SOME_STRING';
However, the string you specified did not match the name of any of the
standard handlers ('confess', 'croak', 'die', 'failobj', 'undef', or
'null') nor the name of any handler you had specified yourself using the
'handlers' option.
Did you perhaps misspell the handler name?
- "Can't call ON_FAILURE after compilation"
- Lexical failure handlers must be specified at compile-time (usually in
your module's "import()" subroutine). However, you called
"ON_FAILURE" at runtime.
Move the call into your module's "import()", or into some other
subroutine that "import()" calls.
- "Can't call ON_FAILURE outside a subroutine"
- You probably attempted to set up a lexical handler at the top level of
your module's source code. For example:
package Your::Module;
use Lexical::Failure;
ON_FAILURE('die');
The lexical hinting mechanism that Lexical::Failure uses only works when
"ON_FAILURE" is called from within your module's
"import()" subroutine (or from a subroutine that
"import()" itself calls).
To achieve the "set a default handler for my module" effect
intended in the previous example, rewrite it either as:
package Your::Module;
use Lexical::Failure;
sub import { ON_FAILURE('die'); }
or simply:
package Your::Module;
use Lexical::Failure default => 'die';
- "Missing rename for %s"
- You tried to rename either "fail" or "ON_FAILURE" as
part of your "use Lexical::Failure" call, but forgot to include
the new name for the subroutine (i.e. you left out the argument expected
after 'fail' or 'ON_FAILURE').
- "Missing specification for %s"
- You tried to specify either the 'default' or 'handlers' option as part of
your "use Lexical::Failure" call, but forgot to include the
corresponding default value or handlers hash (i.e. you left out the
argument expected after 'default' or 'handlers').
- "Value for %s option must be a %s"
- You passed a keyword "=>" value pair to
"use Lexical::Failure", but the value was of the wrong type for
that particular keyword. See "Changing the names of the API
keywords" or "Specifying a module's default failure
handler" or "Specifying additional named failure handlers"
for the correct usage.
- "Handlers in 'handlers' hash must all be code references"
- The 'handlers' option to "use Lexical::Failure" expects a
reference to a hash in which each value is a code reference. At least one
of the values in the hash you passed was something else.
- "Unexpected argument (%s)"
- "use Lexical::Failure" accepts only four arguments:
use Lexical::Failure (
fail => $NEW_NAME,
ON_FAILURE => $NEW_NAME,
default => $HANDLER_NAME,
handlers => \%HANDLER_HASH,
);
You attempted to pass it something else. Or perhaps you misspelled one of
the above keywords?
- "Invalid handler type (%s) in call to ON_FAILURE"
- The argument passed to "ON_FAILURE" must be either a string
(i.e. the name of a named handler) or a reference to a subroutine (i.e.
the handler itself) or a reference to a variable (i.e. the lvalue into
which error messages are to be assigned).
You passed it something else (probably a regex or a reference to a
reference).
CONFIGURATION AND ENVIRONMENT¶
Lexical::Failure requires no configuration files or environment variables.
DEPENDENCIES¶
Requires the modules: Scope::Upper, Keyword::Simple, PadWalker, and
Test::Effects.
Also requires the Lexical::Failure::Objects helper module included in its
distribution.
INCOMPATIBILITIES¶
None reported.
BUGS AND LIMITATIONS¶
No bugs have been reported.
Please report any bugs or feature requests to
"bug-lexical-failure@rt.cpan.org", or through the web interface at
<
http://rt.cpan.org>.
AUTHOR¶
Damian Conway "<DCONWAY@CPAN.org>"
LICENCE AND COPYRIGHT¶
Copyright (c) 2013, Damian Conway "<DCONWAY@CPAN.org>". All
rights reserved.
This module is free software; you can redistribute it and/or modify it under the
same terms as Perl itself. See perlartistic.
DISCLAIMER OF WARRANTY¶
BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR THE
SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTHERWISE
STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PROVIDE THE
SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED OR
IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO
THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH YOU. SHOULD THE SOFTWARE
PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR, OR
CORRECTION.
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING WILL ANY
COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR REDISTRIBUTE THE
SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING
OUT OF THE USE OR INABILITY TO USE THE SOFTWARE (INCLUDING BUT NOT LIMITED TO
LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR
THIRD PARTIES OR A FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER
SOFTWARE), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.