| B::Deparse - Perl compiler backend to produce perl code |
B::Deparse - Perl compiler backend to produce perl code
perl -MO=Deparse[,-d][,-fFILE][,-p][,-q][,-l] [,-sLETTERS][,-xLEVEL] prog.pl
B::Deparse is a backend module for the Perl compiler that generates perl source code, based on the internal compiled structure that perl itself creates after parsing a program. The output of B::Deparse won't be exactly the same as the original source, since perl doesn't keep track of comments or whitespace, and there isn't a one-to-one correspondence between perl's syntactical constructions and their compiled form, but it will often be close. When you use the -p option, the output also includes parentheses even when they are not required by precedence, which can make it easy to see if perl is parsing your expressions the way you intended.
While B::Deparse goes to some lengths to try to figure out what your original program was doing, some parts of the language can still trip it up; it still fails even on some parts of Perl's own test suite. If you encounter a failure other than the most common ones described in the BUGS section below, you can help contribute to B::Deparse's ongoing development by submitting a bug report with a small example.
As with all compiler backend options, these must follow directly after the '-MO=Deparse', separated by a comma but not any white space.
if ($var & 0x7f == 65) {print "Gimme an A!"}
print ($which ? $a : $b), "\n";
$name = $ENV{USER} or "Bob";
B::Deparse,-p will print
if (($var & 0)) {
print('Gimme an A!')
};
(print(($which ? $a : $b)), '???');
(($name = $ENV{'USER'}) or '???')
which probably isn't what you intended (the '???' is a sign that
perl optimized away a constant value).
perl -MO=Deparse,-P -e 'sub foo (\@) { 1 } foo @x'
will print
sub foo (\@) {
1;
}
&foo(\@x);
making clear how the parameters are actually passed to foo.
print "Hello, $world, @ladies, \u$gentlemen\E, \u\L$me!";
as
print 'Hello, ' . $world . ', ' . join($", @ladies) . ', '
. ucfirst($gentlemen) . ', ' . ucfirst(lc $me . '!');
Note that the expanded form represents the way perl handles such
constructions internally -- this option actually turns off the reverse
translation that B::Deparse usually does. On the other hand, note that
$x = "$y" is not the same as $x = $y: the former makes the value
of $y into a string before doing the assignment.
elsif, else, and continue blocks. For example, print
if (...) {
...
} else {
...
}
instead of
if (...) {
...
}
else {
...
}
The default is not to cuddle.
If LEVEL is at least 3, for loops will be translated into equivalent
while loops with continue blocks; for instance
for ($i = 0; $i < 10; ++$i) {
print $i;
}
turns into
$i = 0;
while ($i < 10) {
print $i;
} continue {
++$i
}
Note that in a few cases this translation can't be perfectly carried back into the source code -- if the loop's initializer declares a my variable, for instance, it won't have the correct scope outside of the loop.
If LEVEL is at least 5, use declarations will be translated into
BEGIN blocks containing calls to require and import; for
instance,
use strict 'refs';
turns into
sub BEGIN {
require strict;
do {
'strict'->import('refs')
};
}
If LEVEL is at least 7, if statements will be translated into
equivalent expressions using &&, ?: and do {}; for instance
print 'hi' if $nice;
if ($nice) {
print 'hi';
}
if ($nice) {
print 'hi';
} else {
print 'bye';
}
turns into
$nice and print 'hi';
$nice and do { print 'hi' };
$nice ? do { print 'hi' } : do { print 'bye' };
Long sequences of elsifs will turn into nested ternary operators, which B::Deparse doesn't know how to indent nicely.
use B::Deparse;
$deparse = B::Deparse->new("-p", "-sC");
$body = $deparse->coderef2text(\&func);
eval "sub func $body"; # the inverse operation
B::Deparse can also be used on a sub-by-sub basis from other perl programs.
$deparse = B::Deparse->new(OPTIONS)
Create an object to store the state of a deparsing operation and any options. The options are the same as those that can be given on the command line (see OPTIONS); options that are separated by commas after -MO=Deparse should be given as separate strings. Some options, like -u, don't make sense for a single subroutine, so don't pass them.
$deparse->ambient_pragmas(strict => 'all', '$[' => $[);
The compilation of a subroutine can be affected by a few compiler directives, pragmas. These are:
Ordinarily, if you use B::Deparse on a subroutine which has been compiled in the presence of one or more of these pragmas, the output will include statements to turn on the appropriate directives. So if you then compile the code returned by coderef2text, it will behave the same way as the subroutine which you deparsed.
However, you may know that you intend to use the results in a particular context, where some pragmas are already in scope. In this case, you use the ambient_pragmas method to describe the assumptions you wish to make.
Not all of the options currently have any useful effect. See BUGS for more details.
The parameters it accepts are:
$deparse->ambient_pragmas(strict => 'subs refs');
$deparser->ambient_pragmas(re => 'eval');
$deparser->ambient_pragmas(warnings => [qw[void io]]);
If one of the values is the string ``FATAL'', then all the warnings in that list will be considered fatal, just as with the warnings pragma itself. Should you need to specify that some warnings are fatal, and others are merely enabled, you can pass the warnings parameter twice:
$deparser->ambient_pragmas(
warnings => 'all',
warnings => [FATAL => qw/void io/],
);
See the perllexwarn manpage for more information about lexical warnings.
They exist principally so that you can write code like:
{ my ($hint_bits, $warning_bits);
BEGIN {($hint_bits, $warning_bits) = ($^H, ${^WARNING_BITS})}
$deparser->ambient_pragmas (
hint_bits => $hint_bits,
warning_bits => $warning_bits,
'$[' => 0 + $[
); }
which specifies that the ambient pragmas are exactly those which are in scope at the point of calling.
$body = $deparse->coderef2text(\&func)
$body = $deparse->coderef2text(sub ($$) { ... })
Return source code for the body of a subroutine (a block, optionally preceded by a prototype in parens), given a reference to the sub. Because a subroutine can have no names, or more than one name, this method doesn't return a complete subroutine definition -- if you want to eval the result, you should prepend ``sub subname '', or ``sub '' for an anonymous function constructor. Unless the sub was defined in the main:: package, the code will include a package declaration.
use warnings,
use strict 'refs', use bytes, and use integer. ($[, which
behaves like a pragma, is also supported.)
Excepting those listed above, we're currently unable to guarantee that B::Deparse will produce a pragma at the correct point in the program. (Specifically, pragmas at the beginning of a block often appear right before the start of the block instead.) Since the effects of pragmas are often lexically scoped, this can mean that the pragma holds sway over a different portion of the program than in the input file.
In fact, the above is a specific instance of a more general problem: we can't guarantee to produce BEGIN blocks oruse declarations in
exactly the right place. So if you use a module which affects compilation
(such as by over-riding keywords, overloading constants or whatever)
then the output code might not work as intended.
This is the most serious outstanding problem, and will require some help from the Perl core to fix.
If a keyword is over-ridden, and your program explicitly calls the built-in version by using CORE::keyword, the output of B::Deparse will not reflect this. If you run the resulting code, it will call the over-ridden version rather than the built-in one. (Maybe there should be an option to always print keyword calls asCORE::name.)
Some constants don't print correctly either with or without -d.
For instance, neither B::Deparse nor Data::Dumper know how to print
dual-valued scalars correctly, as in:
use constant E2BIG => ($!=7); $y = E2BIG; print $y, 0+$y;
An input file that uses source filtering probably won't be deparsed into
runnable code, because it will still include the use declaration
for the source filtering module, even though the code that is
produced is already ordinary Perl which shouldn't be filtered again.
There are probably many more bugs on non-ASCII platforms (EBCDIC).
Stephen McCamant <smcc@CSUA.Berkeley.EDU>, based on an earlier version by Malcolm Beattie <mbeattie@sable.ox.ac.uk>, with contributions from Gisle Aas, James Duncan, Albert Dvornik, Robin Houston, Dave Mitchell, Hugo van der Sanden, Gurusamy Sarathy, Nick Ing-Simmons, and Rafael Garcia-Suarez.
| B::Deparse - Perl compiler backend to produce perl code |