perlfaq7 - General Perl Language Issues ($Revision: 1.28 $, $Date: 2005/12/31 00:54:37 $)
This section deals with general Perl language issues that don't clearly fit into any of the other sections.
There is no BNF, but you can paw your way through the yacc grammar in perly.y in the source distribution if you're particularly brave. The grammar relies on very smart tokenizing code, so be prepared to venture into toke.c as well.
In the words of Chaim Frenkel: "Perl's grammar can not be reduced to BNF. The work of parsing perl is distributed between yacc, the lexer, smoke and mirrors."
They are type specifiers, as detailed in perldata:
- $ for scalar values (number, string or reference)
- @ for arrays
- % for hashes (associative arrays)
- & for subroutines (aka functions, procedures, methods)
- * for all types of that symbol name. In version 4 you used them like
- pointers, but in modern perls you can just use references.
There are couple of other symbols that you're likely to encounter that aren't really type specifiers:
- <> are used for inputting a record from a filehandle.
- \ takes a reference to something.
Note that <FILE> is neither the type specifier for files
nor the name of the handle. It is the <>
operator applied
to the handle FILE. It reads one line (well, record--see
$/ in perlvar) from the handle FILE in scalar context, or all lines
in list context. When performing open, close, or any other operation
besides <>
on files, or even when talking about the handle, do
not use the brackets. These are correct: eof(FH), seek(FH, 0,
2)
and "copying from STDIN to FILE".
Normally, a bareword doesn't need to be quoted, but in most cases
probably should be (and must be under use strict
). But a hash key
consisting of a simple word (that isn't the name of a defined
subroutine) and the left-hand operand to the =>
operator both
count as though they were quoted:
- This is like this
- ------------ ---------------
- $foo{line} $foo{'line'}
- bar => stuff 'bar' => stuff
The final semicolon in a block is optional, as is the final comma in a list. Good style (see perlstyle) says to put them in except for one-liners:
One way is to treat the return values as a list and index into it:
- $dir = (getpwnam($user))[7];
Another way is to use undef as an element on the left-hand-side:
You can also use a list slice to select only the elements that you need:
- ($dev, $ino, $uid, $gid) = ( stat($file) )[0,1,4,5];
If you are running Perl 5.6.0 or better, the use warnings
pragma
allows fine control of what warning are produced.
See perllexwarn for more details.
- {
- no warnings; # temporarily turn off warnings
- $a = $b + $c; # I know these might be undef
- }
Additionally, you can enable and disable categories of warnings. You turn off the categories you want to ignore and you can still get other categories of warnings. See perllexwarn for the complete details, including the category names and hierarchy.
- {
- no warnings 'uninitialized';
- $a = $b + $c;
- }
If you have an older version of Perl, the $^W
variable (documented
in perlvar) controls runtime warnings for a block:
- {
- local $^W = 0; # temporarily turn off warnings
- $a = $b + $c; # I know these might be undef
- }
Note that like all the punctuation variables, you cannot currently
use my() on $^W
, only local().
An extension is a way of calling compiled C code from Perl. Reading perlxstut is a good place to learn more about extensions.
Actually, they don't. All C operators that Perl copies have the same precedence in Perl as they do in C. The problem is with operators that C doesn't have, especially functions that give a list context to everything on their right, eg. print, chmod, exec, and so on. Such functions are called "list operators" and appear as such in the precedence table in perlop.
A common mistake is to write:
This gets interpreted as:
To avoid this problem, either put in extra parentheses or use the
super low precedence or
operator:
The "English" operators (and
, or
, xor
, and not
)
deliberately have precedence lower than that of list operators for
just such situations as the one above.
Another operator with surprising precedence is exponentiation. It
binds more tightly even than unary minus, making -2**2
product a
negative not a positive four. It is also right-associating, meaning
that 2**3**2
is two raised to the ninth power, not eight squared.
Although it has the same precedence as in C, Perl's ?: operator
produces an lvalue. This assigns $x to either $a or $b, depending
on the trueness of $maybe:
- ($maybe ? $a : $b) = $x;
In general, you don't "declare" a structure. Just use a (probably anonymous) hash reference. See perlref and perldsc for details. Here's an example:
- $person = {}; # new anonymous hash
- $person->{AGE} = 24; # set field AGE to 24
- $person->{NAME} = "Nat"; # set field NAME to "Nat"
If you're looking for something a bit more rigorous, try perltoot.
(contributed by brian d foy)
perlmod, perlmodlib, perlmodstyle explain modules in all the gory details. perlnewmod gives a brief overview of the process along with a couple of suggestions about style.
If you need to include C code or C library interfaces in your module, you'll need h2xs. h2xs will create the module distribution structure and the initial interface files you'll need. perlxs and perlxstut explain the details.
If you don't need to use C code, other tools such as ExtUtils::ModuleMaker and Module::Starter, can help you create a skeleton module distribution.
You may also want to see Sam Tregar's "Writing Perl Modules for CPAN" ( http://apress.com/book/bookDisplay.html?bID=14 ) which is the best hands-on guide to creating module distributions.
See perltoot for an introduction to classes and objects, as well as perlobj and perlbot.
You can use the tainted() function of the Scalar::Util module, available from CPAN (or included with Perl since release 5.8.0). See also Laundering and Detecting Tainted Data in perlsec.
Closures are documented in perlref.
Closure is a computer science term with a precise but hard-to-explain meaning. Closures are implemented in Perl as anonymous subroutines with lasting references to lexical variables outside their own scopes. These lexicals magically refer to the variables that were around when the subroutine was defined (deep binding).
Closures make sense in any programming language where you can have the return value of a function be itself a function, as you can in Perl. Note that some languages provide anonymous functions but are not capable of providing proper closures: the Python language, for example. For more information on closures, check out any textbook on functional programming. Scheme is a language that not only supports but encourages closures.
Here's a classic function-generating function:
The closure works as a function template with some customization slots left out to be filled later. The anonymous subroutine returned by add_function_generator() isn't technically a closure because it refers to no lexicals outside its own scope.
Contrast this with the following make_adder() function, in which the returned anonymous function contains a reference to a lexical variable outside the scope of that function itself. Such a reference requires that Perl return a proper closure, thus locking in for all time the value that the lexical had when the function was created.
Now &$f1($n)
is always 20 plus whatever $n you pass in, whereas
&$f2($n)
is always 555 plus whatever $n you pass in. The $addpiece
in the closure sticks around.
Closures are often used for less esoteric purposes. For example, when you want to pass in a bit of code into a function:
If the code to execute had been passed in as a string,
'$line = <STDIN>'
, there would have been no way for the
hypothetical timeout() function to access the lexical variable
$line back in its caller's scope.
This problem was fixed in perl 5.004_05, so preventing it means upgrading your version of perl. ;)
Variable suicide is when you (temporarily or permanently) lose the value of a variable. It is caused by scoping through my() and local() interacting with either closures or aliased foreach() iterator variables and subroutine arguments. It used to be easy to inadvertently lose a variable's value this way, but now it's much harder. Take this code:
If you are experiencing variable suicide, that my $f
in the subroutine
doesn't pick up a fresh copy of the $f
whose value is <foo>. The output
shows that inside the subroutine the value of $f
leaks through when it
shouldn't, as in this output:
- foobar
- foobarbar
- foobarbarbar
- Finally foo
The $f that has "bar" added to it three times should be a new $f
my $f
should create a new lexical variable each time through the loop.
The expected output is:
- foobar
- foobar
- foobar
- Finally foo
With the exception of regexes, you need to pass references to these objects. See Pass by Reference in perlsub for this particular question, and perlref for information on references.
See "Passing Regexes", below, for information on passing regular expressions.
Regular variables and functions are quite easy to pass: just pass in a reference to an existing or anonymous variable or function:
- func( \$some_scalar );
- func( \@some_array );
- func( [ 1 .. 10 ] );
- func( \%some_hash );
- func( { this => 10, that => 20 } );
- func( \&some_func );
- func( sub { $_[0] ** $_[1] } );
As of Perl 5.6, you can represent filehandles with scalar variables which you treat as any other scalar.
Before Perl 5.6, you had to use the *FH
or \*FH
notations.
These are "typeglobs"--see Typeglobs and Filehandles in perldata
and especially Pass by Reference in perlsub for more information.
To pass regexes around, you'll need to be using a release of Perl
sufficiently recent as to support the qr// construct, pass around
strings and use an exception-trapping eval, or else be very, very clever.
Here's an example of how to pass in a string to be regex compared
using qr//:
Notice how qr// allows flags at the end. That pattern was compiled
at compile time, although it was executed later. The nifty qr//
notation wasn't introduced until the 5.005 release. Before that, you
had to approach this problem much less intuitively. For example, here
it is again if you don't have qr//:
Make sure you never say something like this:
or someone can sneak shell escapes into the regex due to the double interpolation of the eval and the double-quoted string. For example:
- $pattern_of_evil = 'danger ${ system("rm -rf * &") } danger';
- eval "\$string =~ /$pattern_of_evil/";
Those preferring to be very, very clever might see the O'Reilly book, Mastering Regular Expressions, by Jeffrey Friedl. Page 273's Build_MatchMany_Function() is particularly interesting. A complete citation of this book is given in perlfaq2.
To pass an object method into a subroutine, you can do this:
- call_a_lot(10, $some_obj, "methname")
- sub call_a_lot {
- my ($count, $widget, $trick) = @_;
- for (my $i = 0; $i < $count; $i++) {
- $widget->$trick();
- }
- }
Or, you can use a closure to bundle up the object, its method call, and arguments:
You could also investigate the can() method in the UNIVERSAL class (part of the standard perl distribution).
(contributed by brian d foy)
Perl doesn't have "static" variables, which can only be accessed from the function in which they are declared. You can get the same effect with lexical variables, though.
You can fake a static variable by using a lexical variable which goes
out of scope. In this example, you define the subroutine counter
, and
it uses the lexical variable $count
. Since you wrap this in a BEGIN
block, $count
is defined at compile-time, but also goes out of
scope at the end of the BEGIN block. The BEGIN block also ensures that
the subroutine and the value it uses is defined at compile-time so the
subroutine is ready to use just like any other subroutine, and you can
put this code in the same place as other subroutines in the program
text (i.e. at the end of the code, typically). The subroutine
counter
still has a reference to the data, and is the only way you
can access the value (and each time you do, you increment the value).
The data in chunk of memory defined by $count
is private to
counter
.
- BEGIN {
- my $count = 1;
- sub counter { $count++ }
- }
- my $start = count();
- .... # code that calls count();
- my $end = count();
In the previous example, you created a function-private variable
because only one function remembered its reference. You could define
multiple functions while the variable is in scope, and each function
can share the "private" variable. It's not really "static" because you
can access it outside the function while the lexical variable is in
scope, and even create references to it. In this example,
increment_count
and return_count
share the variable. One
function adds to the value and the other simply returns the value.
They can both access $count
, and since it has gone out of scope,
there is no other way to access it.
- BEGIN {
- my $count = 1;
- sub increment_count { $count++ }
- sub return_count { $count }
- }
To declare a file-private variable, you still use a lexical variable. A file is also a scope, so a lexical variable defined in the file cannot be seen from any other file.
See Persistent Private Variables in perlsub for more information. The discussion of closures in perlref may help you even though we did not use anonymous subroutines in this answer. See Persistent Private Variables in perlsub for details.
local($x) saves away the old value of the global variable $x
and assigns a new value for the duration of the subroutine which is
visible in other functions called from that subroutine. This is done
at run-time, so is called dynamic scoping. local() always affects global
variables, also called package variables or dynamic variables.
my($x) creates a new variable that is only visible in the current
subroutine. This is done at compile-time, so it is called lexical or
static scoping. my() always affects private variables, also called
lexical variables or (improperly) static(ly scoped) variables.
For instance:
- sub visible {
- print "var has value $var\n";
- }
- sub dynamic {
- local $var = 'local'; # new temporary value for the still-global
- visible(); # variable called $var
- }
- sub lexical {
- my $var = 'private'; # new private variable, $var
- visible(); # (invisible outside of sub scope)
- }
- $var = 'global';
- visible(); # prints global
- dynamic(); # prints local
- lexical(); # prints global
Notice how at no point does the value "private" get printed. That's because $var only has that value within the block of the lexical() function, and it is hidden from called subroutine.
In summary, local() doesn't make what you think of as private, local variables. It gives a global variable a temporary value. my() is what you're looking for if you want private variables.
See Private Variables via my() in perlsub and Temporary Values via local() in perlsub for excruciating details.
If you know your package, you can just mention it explicitly, as in $Some_Pack::var. Note that the notation $::var is not the dynamic $var in the current package, but rather the one in the "main" package, as though you had written $main::var.
Alternatively you can use the compiler directive our() to bring a dynamic variable into the current lexical scope.
In deep binding, lexical variables mentioned in anonymous subroutines are the same ones that were in scope when the subroutine was created. In shallow binding, they are whichever variables with the same names happen to be in scope when the subroutine is called. Perl always uses deep binding of lexical variables (i.e., those created with my()). However, dynamic variables (aka global, local, or package variables) are effectively shallowly bound. Consider this just one more reason not to use them. See the answer to What's a closure?.
my() and local() give list context to the right hand side
of =
. The <FH> read operation, like so many of Perl's
functions and operators, can tell which context it was called in and
behaves appropriately. In general, the scalar() function can help.
This function does nothing to the data itself (contrary to popular myth)
but rather tells its argument to behave in whatever its scalar fashion is.
If that function doesn't have a defined scalar behavior, this of course
doesn't help you (such as with sort()).
To enforce scalar context in this particular case, however, you need merely omit the parentheses:
You should probably be using lexical variables anyway, although the issue is the same here:
Why do you want to do that? :-)
If you want to override a predefined function, such as open(), then you'll have to import the new definition from a different module. See Overriding Built-in Functions in perlsub. There's also an example in Class::Template in perltoot.
If you want to overload a Perl operator, such as +
or **
,
then you'll want to use the use overload
pragma, documented
in overload.
If you're talking about obscuring method calls in parent classes, see Overridden Methods in perltoot.
When you call a function as &foo
, you allow that function access to
your current @_ values, and you bypass prototypes.
The function doesn't get an empty @_--it gets yours! While not
strictly speaking a bug (it's documented that way in perlsub), it
would be hard to consider this a feature in most cases.
When you call your function as &foo()
, then you do get a new @_,
but prototyping is still circumvented.
Normally, you want to call a function using foo()
. You may only
omit the parentheses if the function is already known to the compiler
because it already saw the definition (use but not require),
or via a forward reference or use subs
declaration. Even in this
case, you get a clean @_ without any of the old values leaking through
where they don't belong.
This is explained in more depth in the perlsyn. Briefly, there's no official case statement, because of the variety of tests possible in Perl (numeric comparison, string comparison, glob comparison, regex matching, overloaded comparisons, ...). Larry couldn't decide how best to do this, so he left it out, even though it's been on the wish list since perl1.
Starting from Perl 5.8 to get switch and case one can use the Switch extension and say:
- use Switch;
after which one has switch and case. It is not as fast as it could be because it's not really part of the language (it's done using source filters) but it is available, and it's very flexible.
But if one wants to use pure Perl, the general answer is to write a construct like this:
- for ($variable_to_test) {
- if (/pat1/) { } # do something
- elsif (/pat2/) { } # do something else
- elsif (/pat3/) { } # do something else
- else { } # default
- }
Here's a simple example of a switch based on pattern matching, this time lined up in a way to make it look more like a switch statement. We'll do a multiway conditional based on the type of reference stored in $whatchamacallit:
- SWITCH: for (ref $whatchamacallit) {
- /^$/ && die "not a reference";
- /SCALAR/ && do {
- print_scalar($$ref);
- last SWITCH;
- };
- /ARRAY/ && do {
- print_array(@$ref);
- last SWITCH;
- };
- /HASH/ && do {
- print_hash(%$ref);
- last SWITCH;
- };
- /CODE/ && do {
- warn "can't print function ref";
- last SWITCH;
- };
- # DEFAULT
- warn "User defined type skipped";
- }
See perlsyn/"Basic BLOCKs and Switch Statements" for many other
examples in this style.
Sometimes you should change the positions of the constant and the variable.
For example, let's say you wanted to test which of many answers you were
given, but in a case-insensitive way that also allows abbreviations.
You can use the following technique if the strings all start with
different characters or if you want to arrange the matches so that
one takes precedence over another, as "SEND"
has precedence over
"STOP"
here:
- chomp($answer = <>);
- if ("SEND" =~ /^\Q$answer/i) { print "Action is send\n" }
- elsif ("STOP" =~ /^\Q$answer/i) { print "Action is stop\n" }
- elsif ("ABORT" =~ /^\Q$answer/i) { print "Action is abort\n" }
- elsif ("LIST" =~ /^\Q$answer/i) { print "Action is list\n" }
- elsif ("EDIT" =~ /^\Q$answer/i) { print "Action is edit\n" }
A totally different approach is to create a hash of function references.
The AUTOLOAD method, discussed in Autoloading in perlsub and AUTOLOAD: Proxy Methods in perltoot, lets you capture calls to undefined functions and methods.
When it comes to undefined variables that would trigger a warning
under use warnings
, you can promote the warning to an error.
- use warnings FATAL => qw(uninitialized);
Some possible reasons: your inheritance is getting confused, you've
misspelled the method name, or the object is of the wrong type. Check
out perltoot for details about any of the above cases. You may
also use print ref($object)
to find out the class $object
was
blessed into.
Another possible reason for problems is because you've used the
indirect object syntax (eg, find Guru "Samy"
) on a class name
before Perl has seen that such a package exists. It's wisest to make
sure your packages are all defined before you start using them, which
will be taken care of if you use the use statement instead of
require. If not, make sure to use arrow notation (eg.,
Guru->find("Samy")
) instead. Object notation is explained in
perlobj.
Make sure to read about creating modules in perlmod and the perils of indirect objects in Method Invocation in perlobj.
If you're just a random program, you can do this to find out what the currently compiled package is:
- my $packname = __PACKAGE__;
But, if you're a method and you want to print an error message that includes the kind of object you were called on (which is not necessarily the same as the one in which you were compiled):
You can use embedded POD to discard it. Enclose the blocks you want
to comment out in POD markers. The <=begin> directive marks a section
for a specific formatter. Use the comment
format, which no formatter
should claim to understand (by policy). Mark the end of the block
with <=end>.
- # program is here
- =begin comment
- all of this stuff
- here will be ignored
- by everyone
- =end comment
- =cut
- # program continues
The pod directives cannot go just anywhere. You must put a pod directive where the parser is expecting a new statement, not just in the middle of an expression or some other arbitrary grammar production.
See perlpod for more details.
Use this code, provided by Mark-Jason Dominus:
- sub scrub_package {
- no strict 'refs';
- my $pack = shift;
- die "Shouldn't delete main package"
- if $pack eq "" || $pack eq "main";
- my $stash = *{$pack . '::'}{HASH};
- my $name;
- foreach $name (keys %$stash) {
- my $fullname = $pack . '::' . $name;
- # Get rid of everything with that name.
- undef $$fullname;
- undef @$fullname;
- undef %$fullname;
- undef &$fullname;
- undef *$fullname;
- }
- }
Or, if you're using a recent release of Perl, you can just use the Symbol::delete_package() function instead.
Beginners often think they want to have a variable contain the name of a variable.
- $fred = 23;
- $varname = "fred";
- ++$$varname; # $fred now 24
This works sometimes, but it is a very bad idea for two reasons.
The first reason is that this technique only works on global variables. That means that if $fred is a lexical variable created with my() in the above example, the code wouldn't work at all: you'd accidentally access the global and skip right over the private lexical altogether. Global variables are bad because they can easily collide accidentally and in general make for non-scalable and confusing code.
Symbolic references are forbidden under the use strict
pragma.
They are not true references and consequently are not reference counted
or garbage collected.
The other reason why using a variable to hold the name of another
variable is a bad idea is that the question often stems from a lack of
understanding of Perl data structures, particularly hashes. By using
symbolic references, you are just using the package's symbol-table hash
(like %main::
) instead of a user-defined hash. The solution is to
use your own hash or a real reference instead.
- $USER_VARS{"fred"} = 23;
- $varname = "fred";
- $USER_VARS{$varname}++; # not $$varname++
There we're using the %USER_VARS hash instead of symbolic references. Sometimes this comes up in reading strings from the user with variable references and wanting to expand them to the values of your perl program's variables. This is also a bad idea because it conflates the program-addressable namespace and the user-addressable one. Instead of reading a string and expanding it to the actual contents of your program's own variables:
- $str = 'this has a $fred and $barney in it';
- $str =~ s/(\$\w+)/$1/eeg; # need double eval
it would be better to keep a hash around like %USER_VARS and have variable references actually refer to entries in that hash:
- $str =~ s/\$(\w+)/$USER_VARS{$1}/g; # no /e here at all
That's faster, cleaner, and safer than the previous approach. Of course, you don't need to use a dollar sign. You could use your own scheme to make it less confusing, like bracketed percent symbols, etc.
- $str = 'this has a %fred% and %barney% in it';
- $str =~ s/%(\w+)%/$USER_VARS{$1}/g; # no /e here at all
Another reason that folks sometimes think they want a variable to contain the name of a variable is because they don't know how to build proper data structures using hashes. For example, let's say they wanted two hashes in their program: %fred and %barney, and that they wanted to use another scalar variable to refer to those by name.
- $name = "fred";
- $$name{WIFE} = "wilma"; # set %fred
- $name = "barney";
- $$name{WIFE} = "betty"; # set %barney
This is still a symbolic reference, and is still saddled with the problems enumerated above. It would be far better to write:
- $folks{"fred"}{WIFE} = "wilma";
- $folks{"barney"}{WIFE} = "betty";
And just use a multilevel hash to start with.
The only times that you absolutely must use symbolic references are when you really must refer to the symbol table. This may be because it's something that can't take a real reference to, such as a format name. Doing so may also be important for method calls, since these always go through the symbol table for resolution.
In those cases, you would turn off strict 'refs'
temporarily so you
can play around with the symbol table. For example:
All those functions (red(), blue(), green(), etc.) appear to be separate, but the real code in the closure actually was compiled only once.
So, sometimes you might want to use symbolic references to directly manipulate the symbol table. This doesn't matter for formats, handles, and subroutines, because they are always global--you can't use my() on them. For scalars, arrays, and hashes, though--and usually for subroutines-- you probably only want to use hard references.
(contributed by brian d foy)
The "bad interpreter" message comes from the shell, not perl. The actual message may vary depending on your platform, shell, and locale settings.
If you see "bad interpreter - no such file or directory", the first line in your perl script (the "shebang" line) does not contain the right path to perl (or any other program capable of running scripts). Sometimes this happens when you move the script from one machine to another and each machine has a different path to perl---/usr/bin/perl versus /usr/local/bin/perl for instance. It may also indicate that the source machine has CRLF line terminators and the destination machine has LF only: the shell tries to find /usr/bin/perl<CR>, but can't.
If you see "bad interpreter: Permission denied", you need to make your script executable.
In either case, you should still be able to run the scripts with perl explicitly:
- % perl script.pl
If you get a message like "perl: command not found", perl is not in your PATH, which might also mean that the location of perl is not where you expect it so you need to adjust your shebang line.
Copyright (c) 1997-2006 Tom Christiansen, Nathan Torkington, and other authors as noted. All rights reserved.
This documentation is free; you can redistribute it and/or modify it under the same terms as Perl itself.
Irrespective of its distribution, all code examples in this file are hereby placed into the public domain. You are permitted and encouraged to use this code in your own programs for fun or for profit as you see fit. A simple comment in the code giving credit would be courteous but is not required.