In some programming languages, eval
, short for the English evaluate, is a function which evaluates a string as though it were an expression in the language, and returns a result; in others, it executes multiple lines of code as though they had been included instead of the line including the eval
. The input to eval
is not necessarily a string; it may be structured representation of code, such as an abstract syntax tree (like Lisp forms), or of special type such as code
(as in Python). The analog for a statement is exec, which executes a string (or code in other format) as if it were a statement; in some languages, such as Python, both are present, while in other languages only one of either eval
or exec
is.
Using eval
with data from an untrusted source may introduce security vulnerabilities. For instance, assuming that the get_data
function gets data from the Internet, this Python code is insecure:
An attacker could supply the program with the string "session.update(authenticated=True)"
as data, which would update the session
dictionary to set an authenticated key to be True. To remedy this, all data which will be used with eval
must be escaped, or it must be run without access to potentially harmful functions.
In interpreted languages, eval
is almost always implemented with the same interpreter as normal code. In compiled languages, the same compiler used to compile programs may be embedded in programs using the eval
function; separate interpreters are sometimes used, though this results in code duplication.
In JavaScript, eval
is something of a hybrid between an expression evaluator and a statement executor. It returns the result of the last expression evaluated.
Example as an expression evaluator:
Example as a statement executor:
One use of JavaScript's eval
is to parse JSON text, perhaps as part of an Ajax framework. However, modern browsers provide JSON.parse
as a more secure alternative for this task.
In ActionScript (Flash's programming language), eval
cannot be used to evaluate arbitrary expressions. According to the Flash 8 documentation, its usage is limited to expressions which represent "the name of a variable, property, object, or movie clip to retrieve. This parameter can be either a String or a direct reference to the object instance."[1]
ActionScript 3 does not support eval.
The ActionScript 3 Eval Library[2] and the D.eval API[3] were development projects to create equivalents to eval
in ActionScript 3. Both have ended, as Adobe Flash Player has reached its end-of-life.
Lisp was the original language to make use of an eval
function in 1958. In fact, definition of the eval
function led to the first implementation of the language interpreter.[4] Before the eval
function was defined, Lisp functions were manually compiled to assembly language statements. However, once the eval
function had been manually compiled it was then used as part of a simple read-eval-print loop which formed the basis of the first Lisp interpreter.
Later versions of the Lisp eval
function have also been implemented as compilers.
The eval
function in Lisp expects a form to be evaluated as it's argument. The resulting value of the given form will be the returned value of the call to eval
.
This is an example Lisp code:
Lisp is well known to be very flexible and so is the eval
function. For example, to evaluate the content of a string, the string would first have to be converted into a Lisp form using the read-from-string
function and then the resulting form would have to be passed to eval
:
One major point of confusion is the question, in which context the symbols in the form will be evaluated. In the above example, form1
contains the symbol +
. Evaluation of this symbol must yield the function for addition to make the example work as intended. Thus some dialects of Lisp allow an additional parameter for eval
to specify the context of evaluation (similar to the optional arguments to Python's eval
function - see below). An example in the Scheme dialect of Lisp (R5RS and later):
In Perl, the eval
function is something of a hybrid between an expression evaluator and a statement executor. It returns the result of the last expression evaluated (all statements are expressions in Perl programming), and allows the final semicolon to be left off.
Example as an expression evaluator:
Example as a statement executor:
Perl also has eval
blocks, which serves as its exception handling mechanism (see Exception handling syntax#Perl). This differs from the above use of eval
with strings in that code inside eval
blocks is interpreted at compile-time instead of run-time, so it is not the meaning of eval
used in this article.
In PHP, eval
executes code in a string almost exactly as if it had been put in the file instead of the call to eval
. The only exception is that errors are reported as coming from a call to eval
, and return statements become the result of the function.
Unlike some languages, the argument to eval
must be a string of one or more complete statements, not just expressions; however, one can get the "expression" form of eval
by putting the expression in a return statement, which causes eval
to return the result of that expression.
Unlike some languages, PHP's eval
is a "language construct" rather than a function,[5] and so cannot be used in some contexts where functions can be, like higher-order functions.
Example using echo:
Example returning a value:
In Lua 5.1, loadstring
compiles Lua code into an anonymous function.
Example as an expression evaluator:
Example to do the evaluation in two steps:
Lua 5.2 deprecates loadstring
in favor of the existing load
function, which has been augmented to accept strings. In addition, it allows providing the function's environment directly, as environments are now upvalues.
PostScript's exec
operator takes an operand — if it is a simple literal it pushes it back on the stack. If one takes a string containing a PostScript expression however, one can convert the string to an executable which then can be executed by the interpreter, for example:
PostScript's run
operator is similar in functionality but instead the interpreter interprets PostScript expressions in a file, itself.
In Python, the eval
function in its simplest form evaluates a single expression.
eval
example (interactive shell):
The eval
function takes two optional arguments, global
and locals
, which allow the programmer to set up a restricted environment for the evaluation of the expression.
The exec
statement (or the exec
function in Python 3.x) executes statements:
exec
example (interactive shell):
The most general form for evaluating statements/expressions is using code objects. Those can be created by invoking the compile
function and by telling it what kind of input it has to compile: an "exec
" statement, an "eval
" statement or a "single
" statement:
compile
example (interactive shell):
D is a statically compiled language and therefore does not include an "eval
" statement in the traditional sense, but does include the related "mixin
" statement. The difference is that, where "eval
" interprets a string as code at runtime, with a "mixin
" the string is statically compiled like ordinary code and must be known at compile time. For example:
void main
The above example will compile to exactly the same assembly language instructions as if "num++;
" had been written directly instead of mixed in. The argument to mixin doesn't need to be a string literal, but arbitrary expressions resulting in a string value, including function calls, that can be evaluated at compile time.
ColdFusion's evaluate
function lets you evaluate a string expression at runtime.
It is particularly useful when you need to programmatically choose the variable you want to read from.
The Ruby programming language interpreter offers an eval
function similar to Python or Perl, and also allows a scope, or binding, to be specified.
Aside from specifying a function's binding, eval
may also be used to evaluate an expression within a specific class definition binding or object instance binding, allowing classes to be extended with new methods specified in strings.
def get_binding(a) bindingendeval('a+1',get_binding(3)) # (evaluates to 4, because 'a' in the context of get_binding is 3)
Most standard implementations of Forth have two variants of eval
: EVALUATE
and INTERPRET
.
Win32FORTH code example:
In REALbasic, there is a class called RBScript which can execute REALbasic code at runtime. RBScript is very sandboxed—only the most core language features are there, and you have to allow it access to things you want it to have. You can optionally assign an object to the context property. This allows for the code in RBScript to call functions and use properties of the context object. However, it is still limited to only understanding the most basic types, so if you have a function that returns a Dictionary or MySpiffyObject, RBScript will be unable to use it. You can also communicate with your RBScript through the Print and Input events.
Microsoft's VBScript, which is an interpreted language, has two constructs. Eval
is a function evaluator that can include calls to user-defined functions. (These functions may have side-effects such as changing the values of global variables.) Execute
executes one or more colon-separated statements, which can change global state.
Both VBScript and JScript eval
are available to developers of compiled Windows applications (written in languages which do not support Eval) through an ActiveX control called the Microsoft Script Control, whose Eval method can be called by application code. To support calling of user-defined functions, one must first initialize the control with the AddCode method, which loads a string (or a string resource) containing a library of user-defined functions defined in the language of one's choice, prior to calling Eval.
Visual Basic for Applications (VBA), the programming language of Microsoft Office, is a virtual machine language where the runtime environment compiles and runs p-code. Its flavor of Eval supports only expression evaluation, where the expression may include user-defined functions and objects (but not user-defined variable names). Of note, the evaluator is different from VBS, and invocation of certain user-defined functions may work differently in VBA than the identical code in VBScript.
As Smalltalk's compiler classes are part of the standard class library and usually present at run time, these can be used to evaluate a code string.
The Tcl programming language has a command called eval
, which executes the source code provided as an argument. Tcl represents all source code as strings, with curly braces acting as quotation marks, so that the argument to eval
can have the same formatting as any other source code.
bs has an eval
function that takes one string argument. The function is both an expression evaluator and a statement executor. In the latter role, it can also be used for error handling. The following examples and text are from the bs
man page as appears in the UNIX System V Release 3.2 Programmer's Manual.[6]
The eval command is present in all Unix shells, including the original "sh" (Bourne shell). It concatenates all the arguments with spaces, then re-parses and executes the result as a command.
In Windows PowerShell, the Invoke-Expression
Cmdlet serves the same purpose as the eval function in programming languages like JavaScript, PHP and Python.The Cmdlet runs any Windows PowerShell expression that is provided as a command parameter in the form of a string and outputs the result of the specified expression.Usually, the output of the Cmdlet is of the same type as the result of executing the expression. However, if the result is an empty array, it outputs $null
. In case the result is a single-element array, it outputs that single element. Similar to JavaScript, Windows PowerShell allows the final semicolon to be left off.
Example as an expression evaluator:
In 1966 IBM Conversational Programming System (CPS) introduced a microprogrammed function EVAL
to perform "interpretive evaluation of expressions which are written in a modified Polish-string notation" on an IBM System/360 Model 50.[7] Microcoding this function was "substantially more" than five times faster compared to a program that interpreted an assignment statement.[8]
In theoretical computer science, a careful distinction is commonly made between eval and apply. Eval is understood to be the step of converting a quoted string into a callable function and its arguments, whereas apply is the actual call of the function with a given set of arguments. The distinction is particularly noticeable in functional languages, and languages based on lambda calculus, such as LISP and Scheme. Thus, for example, in Scheme, the distinction is between
where the form (f x) is to be evaluated, and
where the function f is to be called with argument x.
Eval and apply are the two interdependent components of the eval-apply cycle, which is the essence of evaluating Lisp, described in SICP.[9]
In category theory, the eval morphism is used to define the closed monoidal category. Thus, for example, the category of sets, with functions taken as morphisms, and the cartesian product taken as the product, forms a Cartesian closed category. Here, eval (or, properly speaking, apply) together with its right adjoint, currying, form the simply typed lambda calculus, which can be interpreted to be the morphisms of Cartesian closed categories.