Debug::Filter::PrintExpr - Convert comment lines to debug print statements
use Debug::Filter::PrintExpr;
my $s = 'a scalar';
my @a = qw(this is an array);
my %h = (key1 => 'value1', key2 => 'value2', '' => 'empty', undef => undef);
my $ref = \%h;
#${$s}
#@{@a}
#%{ %h }
#${ calc: @a * 2 }
#\{$ref}
This produces an output like:
L13: $s = 'a scalar';
L14: @a = ('this', 'is', 'an', 'array');
L15: %h = ('' => 'empty', 'key1' => 'value1', 'key2' => 'value2', 'undef' => undef);
calc: @a * 2 = 8;
L17: dump($ref);
$_[0] = {
'' => 'empty',
'key1' => 'value1',
'key2' => 'value2',
'undef' => undef
};
Providing debug output often results in a couple of print statements that display the value of some expression and some kind of description. When the program development is finished, these statements must be made conditional on some variable or turned into comments.
Often the contents of arrays or hashes need to be presented in a readable way, leading to repeated lines of similar code.
C programmers use the preprocessor to solve this problem. As Perl has it's own filter mechanism for preprocessing, this leads to a similar solution in Perl.
The Filter::Simple module by Damian Conway provides a convenient way of implementing Perl filters.
Debug::Filter::PrintExpr
makes use of Filter::Simple
to transform specially formed comment lines into print statements
for various debugging purposes.
(Besides, there is Smart::Comments from Damian, that does something
very similar but more advanced.)
Just by removing the "use" of Debug::Filter::PrintExpr
completely,
disabling it partially by
no Debug::Filter::PrintExpr;
or making the usage conditional (e.g. on environment variable DEBUG) by
use if $ENV{DEBUG}, 'Debug::Filter::PrintExpr';
all these lines (or a part of them) lose their magic and remain simple comments.
The comment lines to be transformed must follow this format:
# sigil { [label:] [expression] }
or more formally must be matched by the following regexp:
qr{
^\h*\#
(?<type>[%@\$\\"#])
\{\h*
(?<label>[[:alpha:]_]\w*:)?
\h*
(?<expr>\V+)?
\}\h*$
}x
where type
represents the sigil, label
an optional label and
expr
an optional expression.
If the label is omitted, it defaults to L_n_:
, where n is the
line number in the program.
The sigil determines the evaluation context for the given expression and the output format of the result:
-
$
The expression is evaluated in scalar context. Strings are printed inside single quotes, integer and floating point numbers are printed unquoted and dual valued variables are shown in the form
dualvar(_numval_, '_stringval_')
. Undefined values are represented by the unquoted stringundef
. Hash and array references are shown in their usual string representation as e.g.ARRAY(0x19830d0)
orHASH(0xccba88)
. Blessed references are shown by the class they are belong to asblessed(class)
. -
"
The expression is evaluated in scalar context as a string.
-
#
The expression is evaluated in scalar context as a numeric value.
-
@
The expression is evaluated in list context and the elements of the list are printed like single scalars, separated by commas and gathered in parentheses.
-
%
The expression is evaluated as a list of key-value pairs and is presented in the form 'key' => value,... inside parentheses. value is formatted like a single scalar.
-
\
The expression shall evaluate to a list of references. These will be evaluated using Data::Dumper and named like parameters in a subroutine, i.e.
$_[_n_]
.
The usage and difference between #${}
, #"{}
and ##{}
is
best described by example:
my $dt = DateTime->now;
#${$dt} # Ln: $dt = blessed(DateTime);
#"{$dt} # Ln: $dt = '2019-10-27T15:54:28';
my $num = ' 42 ';
#${$num} # Ln: $num = ' 42 ';
$num + 0;
#${$num} # Ln: $num = dualvar(42, ' 42 ');
#"{$num} # Ln: $num = ' 42 ';
##{$num} # Ln: $num = 42;
The forms #${}, #"{}, ##{} and #@{} may be used for any type of expression and inside the #%{} form, arrays are permitted too. With the varibles $s, @a and %h as defined above, it is possible to use:
#@{scalar_as_array: $s}
#${array_as_scalar :@a}
#@{hash_as_array: %h}
and produce these results:
scalar_as_array: $s = ('this is a scalar');
array_as_scalar: @a = 4;
hash_as_array: %h = ('k1', 'v1', 'k2', 'v2');
Regular expressions may be evaluated too:
#@{"a<b>c<d><e>f<g>h" =~ /\w*<(\w+)>/g}
gives:
Ln: "a<b>c<d><e>f<g>h" =~ /\w*<(\w+)>/g = ('b', 'd', 'e', 'g');
If the expression is omitted, only the label will be printed.
The sigil $
should be used in this case.
Requirements for the expression are:
- It must be a valid Perl expression.
- In case of the #%{}-form, it must evaluate to a list of pairs, e.g. a hash.
A PrintExpr will be resolved to a block and therefore may be located anywhere in the program where a block is valid. Do not put it in a place, where a block is required (e.g. after a conditional) as this would break the code when running without the filter.
As a code snippet of the form {label: expr}
is a valid perl
expression and the generated code will result in a
braced expression, a simple consistency check can be done by removing
hash and sigil from the PrintExpr line:
The resulting code must still be valid and should only emit a warning
about a useless use of something in void context.
The use
statement for Debug::Filter::PrintExpr
may contain
following arguments:
-
-debug
This option causes the resulting source code after comment transformation to be written to
STDERR
.
-
$Debug::Filter::PrintExpr::handle
The filehandle that is referenced by this variable is used for printing the generated output. The default is STDERR and may be changed by the caller.
Damian Conway's module Smart::Comments provides something similar and more advanced.
While Smart::Comments has lots of features for visualizing the program flow, this module focuses on data representation. The main requirements for this module were:
- Always print the source line number or a user provide label.
- Always print the literal expression along with its evaluation.
- Give a defined context where the expression is evaluated. Especially provide scalar and list context or perform an iteration over the key-value pairs of a hash. The usage of Data::Dumper was adopted later from Damian's implementation.
- Trailing whitespace in values should be clearly visible.
- Distinguish between the numeric and string value of a variable.
- undefined values should be clearly distinguishable from empty values.
The first three requirements are not met by Smart::Comments as there is an extra effort needed to display a line number, the display of a label and the literal expression are mutual exclusive and a specific context is not enforced by the module.
All in all, the module presented here is not much more than a programming exercise.
Other related modules: Scalar::Util, Data::Dumper
Jörg Sommrey
Copyright (c) 2018-2020, Jörg Sommrey. All rights reserved.
This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.
See http://dev.perl.org/licenses/ for more information.