NAME
Devel::Assert - assertions support for perls >= 5.8.1
SYNOPSIS
use Devel::Assert;
sub my_sqrt{
my $num = shift;
assert($num > 0);
return sqrt($num);
}
#this will fail with built-in sqrt's message, 'cos of assert() is compiled as no-op
my_sqrt(-12);
use Devel::Assert -all, -verbose;
sub my_sqrt{
my $num = shift;
assert($num > 0);
return sqrt($num);
}
#this will fail with "Assertion '$num > 0' failed
# guessing variables... $num = '-12'
# at ..., line ..."
my_sqrt(-12);
FEATURES
assert() function is replaced with no-op at compile stage for release builds
precise diagnostics for what and where happened
Not a source filter (magic for syntax is provided by Devel::Declare)
DESCRIPTION
When writing code, you always make some assumptions (correct or not) about incoming data, results of internal or external function call, etc. Using assertions is a way to document these assumptions, thus making your program more reliable by forcing them.
# old way:
#data should be non-empty array
my @values = process_data(@data); #values contains same element count as data
# new way:
assert(scalar @data);
my @values = process_data(@data);
assert(scalar @data == scalar @values);
But what about more code to execute? Often unnecessary code is beeing cut of by postfixing it with 'if DEBUG' (and declaring DEBUG as a constant), but for Devel::Assert that isn't necessary. assert
call is only really compiled for debug builds - by changing just a single import line, you can make all them no-ops, causing no overhead (except for compile time, of course).
# old way:
sub DEBUG () { 1 }
die 'no data given' if DEBUG && !@data;
my @values = process_data(@data);
die 'wrong count' if DEBUG && (scalar @data != scalar @values);
# new way:
use Devel::Assert -all;
assert(scalar @data);
my @values = process_data(@data);
assert(scalar @data == scalar @values);
Furthermore, by using -verbose
mode, you can not only get description of where failure occured, but also what has exactly failed.
# old way:
die 'no data given' if DEBUG && !@data;
my @values = process_data(@data);
if (DEBUG && (scalar @data != scalar @values)){
require Data::Dumper;
Data::Dumper->import('Dumper');
die "Data and values count mismatch: ".Dumper(\@data).Dumper(\@values);
}
# new way:
use Devel::Assert -all, -verbose;
assert(scalar @data);
my @values = process_data(@data);
assert(scalar @data == scalar @values);
So, the more debug info you need - the less code you write.
CONFIGURATION
When you're use'ing Devel::Assert, you should specify parsing mode for assert call. That's done by passing one of the following to import
:
-all
-
Starting with this point, all assert() calls are really done and your assertions are checked for all subsequently compiled modules, regardless of their import() option. This can be used in your main program, for example, before any other modules used (global debug mode).
-none
-
Starting with this point, all assert() calls are changed to no-op for all subsequently compiled modules, regardless of their import() option. This can be used in your main program, for example, before any other modules used (global release mode).
Note: first occurence of
-all
or-none
takes precedence of all further ones. - any boolean 'true'
-
In this module, assert() calls are really done and your assertions are checked (debug mode).
- any boolean 'false' (or nothing)
-
In this module, assert() calls are changed to no-op (release mode).
-verbose
-
This is special control value, that can be added to normal options. In the
-verbose
mode, in case of assertion failure, Devel::Assert will try to dump all lexical variables that are mentioned in failed code. That's done through Data::Dumper and PadWalker - without these modules,-verbose
mode will fall back to default behaviour, with warning emitted.Here is sample output:
Assertion ' scalar @y < $x ' failed, trying to determine acting variables... $x = 3; @y = ( 1, 2, { '4' => 'HASH(0x3d913c)' } ); ...and all this happened at test.pl line 10 main::__ANON__() called at test.pl line 13 main::z() called at test.pl line 17
DISABLING ASSERTIONS
To temporary disable assertions (at compile time, turning them to no-ops), you can do the following trick:
no Devel::Assert;
<some code with assertions that would not trigger even under '-all'>
use Devel::Assert q/1/;
<code with assertions turned on>
Note: 'no Devel::Assert' is not lexically scoped. It's effect prolongates till module end or next 'use Devel::Assert'.
FAILURE HOOKS & RUN-TIME CONFIGURATION
In case of assertion failure, depending on -verbose
flag was set or not, Devel::Assert calls some internal callbacks that generates messages that you see by default. By calling (always fully qualified) sub set_options
, you can change default behavior. Also, through the same call, you can change -verbose
status at run-time.
Devel::Assert::set_options(
verbose => 1,
hook_terse => \&some_sub,
hook_verbose => \&another_sub,
);
verbose
-
Change assertions verbosity status at run-time
hook_terse
-
Called with one or two arguments. First one is always a failed assertion text ('$num > 0'). Second, if present, is a message generated from
hook_verbose
.Default behaviour for
hook_terse
is to callCarp::confess
. hook_verbose
-
Called with one argument - failed assertion text. Overriding this seems to be of little use, and provided here only for completeness.
By default, calls
hook_terse
with arguments described above.
CAVEATS
One subroutine - assert
- is exported to the caller namespace.
If you place argument to assert() at more then one line, then source code line numbering will shift.
Parser detail - you must specify parentheses for assert call:
assert 3 < 5; #wrong!
assert (3 < 5); #ok
SEE ALSO
assertions - for perls >= 5.9.0 only, different syntax (code attributes).
Carp::Assert - requires nasty 'if DEBUG' suffix.
AUTHOR
Sergey Aleynikov <sergey.aleynikov@gmail.com>
LICENSE
Copyright (c) 2009 by Sergey Aleynikov. This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
Some part for parser are taken from Devel::Declare::Context::Simple, (c) Rhesa Rozendaal (?).