Eval::Context(3) Evalute perl code in context wraper


use Eval::Context ;

my $context = new Eval::Context(PRE_CODE => "use strict;\nuse warnings;\n") ;

# code will be evaluated with strict and warnings loaded in the context.

$context->eval(CODE => 'print "evaluated in an Eval::Context!" ;') ;
$context->eval(CODE_FROM_FILE => 'file.pl') ;


This module define a subroutine that let you evaluate Perl code in a specific context. The code can be passed directly as a string or as a file name to read from. It also provides some subroutines to let you define and optionally share variables and subroutines between your code and the code you wish to evaluate. Finally there is some support for running your code in a safe compartment.

Don't play with fire!

Don't start using this module, or any other module, thinking it will let you take code from anywhere and be safe. Read perlsec, Safe, Opcode, Taint and other security related documents. Control your input.


Subroutines that are not part of the public interface are marked with [p].


Create an Eval::Context object. The object is used as a repository of ``default'' values for your code evaluations. The context can be used many times. The values can be temporarily overridden during the "eval" call.

  my $context = new Eval::Context() ; # default context
  my $context = new Eval::Context
                NAME              => 'libraries evaluation context',
                PACKAGE           => 'libraries',
                SAFE              => {...} ;
                PRE_CODE          => "use strict ;\n"
                POST_CODE         => sub{},
                PERL_EVAL_CONTEXT => undef,
                INSTALL_SUBS      => {...},
                INSTALL_VARIABLES => [...],
                EVAL_SIDE_PERSISTENT_VARIABLES => {...},
                INTERACTION => {...},
                DISPLAY_SOURCE_IN_CONTEXT => 1, #useful when debuging
                ) ;


@named_arguments - setup data for the object

All the arguments optional. The argument passed to "new" can also be passed to "eval". All arguments are named.

  • NAME - use when displaying information about the object.

    Set automatically to 'Anonymous' if not set. The name will also be reported by perl if an error occurs during your code evaluation.

  • PACKAGE - the package the code passed to "eval" will evaluated be in.

    If not set, a unique package name is generated and used for every "eval" call.

  • REMOVE_PACKAGE_AFTER_EVAL - When set the content of the package after evaluation will be erase

    The default behavior is to remove all data from after the call to "eval".

  • PRE_CODE - code prepended to the code passed to eval
  • POST_CODE - code appended to the code passed to eval
  • PERL_EVAL_CONTEXT - the context to eval code in (void, scalar, list).

    This option Works as ``wantarray'' in perlfunc. It will override the context in which "eval" is called.

  • INSTALL_SUBS - subs that will be available in the eval.

    A hash where the keys are a function names and the values a code references.

  • SAFE

    This argument must be a hash reference. if the hash is empty, a default safe compartment will be used. Read Safe documentation for more information.

            SAFE => {} # default safe environment

    You can have a finer control over the safe compartment Eval::Context that will be used.

            my $compartment = new Safe('ABC') ;
            my $context = new Eval::Context
                    SAFE => # controlling the safe environment
                            PACKAGE     => 'ABC', 
                            PRE_CODE    => "use my module ;\n" # code we consider safe
                            USE_STRICT  => 0,                # set to 1 by default
                            COMPARTMENT => $compartment ,   # use default if not passed
                            } ,
            $context->eval(CODE => .....) ;
  • COMPARTMENT - a Safe object, you create, that will be used by Eval::Context
  • USE_STRICT - Controls if strict is used in the Safe compartment

    The default is to use strict. Note that ``Safe'' in perldoc default is to NOT use strict (undocumented).

  • PRE_CODE - safe code you want to evaluate in the same context as the unsafe code

    This let you, for example, use certain modules which provide subroutines to be used in the evaluated code. The default compartment is quite restrictive and you can't even use strict in it without tuning the safe compartment.

  • A few remarks:

    - See <http://rt.cpan.org/Ticket/Display.html?id=31090> on RT

    - Pass the same package name to your safe compartment and to Eval::Context.

    - If you really want to be on the safe side, control your input. When you use a module, are you sure the module hasn't been fiddle with?

    - Leave strict on. Even for trivial code.

  • INSTALL_VARIABLES - ``Give me sugar baby'' Ash.

    Eval::Context has mechanisms you can use to set and share variables with the code you will evaluate. There are two sides in an Eval::Context. The caller-side, the side where the calls to "eval" are made and the eval-side, the side where the code to be evaluated is run.

  • How should you get values back from the eval-side

    Although you can use the mechanisms below to get values from the eval-side, the cleanest way is to get the results directly from the "eval" call.

            my $context = new Eval::Context() ;
            my ($scalr_new_value, $a_string) =
                            INSTALL_VARIABLES =>[[ '$scalar'  => 42]] ,
                            CODE => "\$scalar++ ;\n (\$scalar, 'a string') ;",
                            ) ;
  • initializing variables on the eval side

    You can pass INSTALL_VARIABLES to "new" or "eval". You can initialize different variables for each run of "eval".

            my $context = new Eval::Context
                    INSTALL_VARIABLES =>
                            # variables on eval-side    #initialization source
                            [ '$data'                => 42],
                            [ '$scalar'              => $scalar_caller_side ],
                            [ '%hash'                => \%hash_caller_side ]
                            [ '$hash'                => \%hash_caller_side ],
                            [ '$object'              => $object ],
                            ] ,
                    ) ;

    The variables will be my variables on the eval-side.

    You can declare variables of any of the base types supported by perl. The initialization data , on the caller-side, is serialized and deserialized to make the values available on the eval-side. Modifying the variables on the eval-side does not modify the variables on the caller-side. The initialization data can be scalars or references and even my variables.

  • Persistent variables

    When evaluating code many times in the same context, you may wish to have variables persist between evaluations. Eval::Context allows you to declare, define and control such state variables.

    This mechanism lets you control which variables are persistent. Access to the persistent variables is controlled per "eval" run. Persistent variables are my variables on the eval-side. Modifying the variables on the eval-side does not modify the variables on the caller-side.

    Define persistent variables:

            # note: creating persistent variables in 'new' makes little sense as
            # it will force those values in the persistent variables for every run.
            # This may or may not be what you want.
            my $context = new Eval::Context() ;
                    INSTALL_VARIABLES =>
                            [ '$scalar'  => 42                 => $Eval::Context::PERSISTENT ] ,
                            # make %hash and $hash available on the eval-side. both are
                            # initialized from the same caller-side hash
                            [ '%hash'    => \%hash_caller_side => $Eval::Context::PERSISTENT ] ,
                            [ '$hash'    => \%hash_caller_side => $Eval::Context::PERSISTENT ] ,
                    CODE => '$scalar++',
                    ) ;

    Later, use the persistent value:

                    INSTALL_VARIABLES =>
                            [ '$scalar'  => $Eval::Context::USE => $Eval::Context::PERSISTENT ] ,
                            # here you decided %hash and $hash shouldn't be available on the eval-side
                    CODE => '$scalar',
                    ) ;

    $Eval::Context::USE means ``make the persistent variable and it's value available on the eval-side''. Any other value will reinitialize the persistent variable. See also REMOVE_PERSISTENT in "eval".

  • Manually synchronizing caller-side data with persistent eval-side data

    Although the first intent of persistent variables is to be used as state variables on the eval-side, you can get persistent variables values on the caller-side. To change the value of an eval-side persistent variable, simply reinitialize it with INSTALL_VARIABLES next time you call "eval".

            my $context = new Eval::Context
                            INSTALL_VARIABLES =>
                                    ['%hash' => \%hash_caller_side => $Eval::Context::PERSISTENT] 
                                    ] ,
                            ) ;
            $context->Eval(CODE => '$hash{A}++ ;') ;
            # throws exception if you request a non existing variable
            my %hash_after_eval = $context->GetPersistantVariables('%hash') ;
  • Getting the list of all the PERSISTENT variables

            my @persistent_variable_names = $context->GetPersistantVariablesNames() ;
  • Creating persistent variables on the eval-side

    The mechanism above gave you fine control over persistent variables on the eval-side. The negative side is that only the variables you made persistent exist on the eval-side. Eval::Context has another mechanism that allows the eval-side to store variables between evaluations without the caller-side declaration of the variables.

    To allow the eval-side to store any variable, add this to you "new" call.

            my $context = new Eval::Context
                    PACKAGE => 'my_package',
                            SAVE => { NAME => 'SavePersistent', VALIDATOR => sub{} },
                            GET  => { NAME => 'GetPersistent',  VALIDATOR => sub{} },
                    ) ;

    The eval-side can now store variables between calls to "eval"

            SavePersistent('name', $value) ;

    later in another call to "eval":

            my $variable = GetPersistent('name') ;

    By fine tuning EVAL_SIDE_PERSISTENT_VARIABLES you can control what variables are stored by the eval-side. This should seldom be used and only to help those storing data from the eval-side.

    You may have notices in the code above that a package name was passed as argument to "new". This is very important as the package names that are automatically generated differ for each "eval" call. If you want to run all you eval-side code in different packages (Eval::Context default behavior), you must tell Eval::Context where to store the eval-side values. This is done by setting CATEGORY

    The validator sub can verify if the value to be stored are valid, E.G.: variable name, variable value is within range, ...

    Here is an example of code run in different packages but can share variables. Only variables which names start with A are valid.

            new Eval::Context
                            CATEGORY => 'TEST',
                            SAVE => 
                                    NAME => 'SavePersistent',
                                    VALIDATOR => sub 
                                            my ($self, $name, $value, $package) = @_ ;
                                                    "SavePersistent: name '$name' doesn't start with A!"
                                                    )  unless $name =~ /^A/ ;
                            GET => {NAME => 'GetPersistent',VALIDATOR => sub {}},
                    ) ;
            $context->eval(CODE => 'SavePersistent('A_variable', 123) ;') ;


            $context->eval(CODE => 'GetPersistent('A_variable') ;') ;
  • Shared variables

    You can also share references between the caller-side and the eval-side.

            my $context = 
                    new Eval::Context
                            INSTALL_VARIABLES =>
                                    # reference to reference only
                                    [ '$scalar' => \$scalar           => $Eval::Context::SHARED ],
                                    [ '$hash'   => \%hash_caller_side => $Eval::Context::SHARED ],
                                    [ '$object' => $object            => $Eval::Context::SHARED ],
                                    ] ,
                            ) ;

    Modification of the variables on the eval-side will modify the variable on the caller-side. There are but a few reasons to share references. Note that you can share references to my variables.


    Lets you define subs used to interact with the user.

            INTERACTION      =>
                    INFO     => \&sub,
                    WARN     => \&sub,
                    DIE      => \&sub,
                    EVAL_DIE => \&sub,
  • This sub will be used when displaying information.
  • This sub will be used when a warning is displayed.
  • Used when an error occurs.
  • Used when an error occurs during code evaluation.
  • FILE - the file where the object has been created.

    This is practical if you want to wrap the object.

    FILE and LINE will be set automatically if not set.

  • LINE - the line where the object has been created. Set automatically if not set.
  • DISPLAY_SOURCE_IN_CONTEXT - if set, the code to evaluated will be displayed before evaluation


  • an Eval::Context object.

[p] Setup

Helper sub called by new.

[p] CheckOptionNames

Verifies the named options passed as arguments with a list of valid options. Calls {INTERACTION}{DIE} in case of error.

[p] SetInteractionDefault

Sets {INTERACTION} fields that are not set by the user.

[p] CanonizeName

Transform a string into a a string with can be used as a package name or file name usable within perl code.


Evaluates Perl code, passed as a string or read from a file, in the context.

        my $context = new Eval::Context(PRE_CODE => "use strict;\nuse warnings;\n") ;
        $context->eval(CODE => 'print "evaluated in an Eval::Context!";') ;
        $context->eval(CODE_FROM_FILE => 'file.pl') ;

Call context

Evaluation context of the code (void, scalar, list) is the same as the context this subroutine was called in or in the context defined by PERL_EVAL_CONTEXT if that option is present.


NOTE: You can override any argument passed to "new". The override is temporary during the duration of this call.

@named_arguments - Any of "new" options plus the following.
  • CODE - a string containing perl code (valid code or an exception is raised)
  • CODE_FROM_FILE - a file containing perl code

    A list of regex used to match the persistent variable names to be removed, persistent variable removal is done before any variable installation is done

  • FILE and LINE - will be used in the evaluated code 'file_name' set to the caller's file and line by default

NOTE: CODE or CODE_FROM_FILE is mandatory.


  • What the code to be evaluated returns

[p] VerifyAndCompleteOptions

Helper sub for "eval".

[p] EvalCleanup

Handles the package cleanup or persistent variables cleanup after a call to "eval".

[p] GetPackageName

Returns a canonized package name. the name is either passed as argument from the caller or a temporary package name.

[p] EvalSetup

Handles the setup of the context before eval-side code is evaluated. Sets the variables and the shared subroutines.

[p] VerifyCodeInput

Verify that CODE or CODE_FROM_FILE are properly set.

[p] RemovePersistent

Handles the removal of persistent variables.

[p] GetCallContextWrapper

Generates perl code to wrap the code to be evaluated in the right calling context.

[p] SetupSafeCompartment

If running in safe mode, setup a safe compartment from the argument, otherwise defines the evaluation package.

[p] GetInstalledVariablesCode

Generates variables on the eval-side from the INSTALL_VARIABLES definitions. Dispatches the generation to specialize subroutines.

[p] GetPersistentVariablesSetFromCaller

Generates code to make persistent variables, defined on the caller-side available on the eval-side.

[p] GetSharedVariablesSetFromCaller

Handles the mechanism used to share variables (references) between the caller-side and the eval-side.

Shared variables must be defined and references. If the shared variable is undef, the variable that was previously shared, under the passed name, is used if it exists or an exception is raised.

Also check that variables are not PERSISTENT and SHARED.

[p] GetVariablesSetFromCaller

Generates code that creates local variables on the eval-side


Arguments - none

Returns - the list of existing persistent variables names

        my @persistent_variable_names = $context->GetPersistantVariablesNames() ;


  • @variable_names - list of variable names to retrieve

Returns - list of values corresponding to the input names

This subroutine will return whatever the caller-site set or the eval-side modified. Thus if you created a %hash persistent variable, a hash (not a hash reference) will be returned.

If you request multiple values, list flattening will be in effect. Be careful.

        my $context = new Eval::Context
                        INSTALL_VARIABLES =>
                                ['%hash' => \%hash_caller_side => $Eval::Context::PERSISTENT] 
                                ] ,
                        ) ;
        $context->Eval(CODE => '$hash{A}++ ;') ;
        # may throw exception
        my %hash_after_eval = $context->GetPersistantVariables('%hash') ;

[p] SetEvalSidePersistenceHandlers

Set the code needed to handle eval-side persistent variables.

[p] RemoveEvalSidePersistenceHandlers

Removes eval-side persistent variable handlers. Used after calling "eval" so the next "eval" can not access eval-side persistent variables without being allowed to do so.


I have reported a very strange error when Safe and Carp are used together. <http://rt.cpan.org/Ticket/Display.html?id=31090>. The error can be reproduced without using Eval::Context.


        Khemir Nadim ibn Hamouda
        CPAN ID: NKH
        mailto:[email protected]


This program is free software; you can redistribute it and/or modify it under the same terms as Perl itself.


You can find documentation for this module with the perldoc command.

    perldoc Eval::Context

You can also look for information at: