eval

I have a script that creates new objects based on the value of a form
field. Basically, the code looks like this.

eval ('new ' + objType.value + '(val1, val2, val3');

objType is a select with the different types of objects you can create
as values.

I really don't like using eval, and it's causing problems, like if I
do something like the following:

eval ('localVar = new ' + objType.value + '(val1, val2, val3');

localVar should refer to a variable local to the calling object, but
with the eval approach it ends up being the local variable for the
scope from which the eval was first executed.

I could use a seitch statement, but this approach means I now have an
extra place where I need to change my code so I'd like to afford this
if possible.

Is there another approach to creating a new instance of an object
dependant on the value of some variable?

Gordon schreef:

I have a script that creates new objects based on the value of a form
field. Basically, the code looks like this.

eval ('new ' + objType.value + '(val1, val2, val3');

objType is a select with the different types of objects you can create
as values.

I really don't like using eval, and it's causing problems, like if I
do something like the following:

eval ('localVar = new ' + objType.value + '(val1, val2, val3');

localVar should refer to a variable local to the calling object, but
with the eval approach it ends up being the local variable for the
scope from which the eval was first executed.

Hi Gordon,

So the scope is the problem, not eval? Right?

I guess your eval-code is in some function.

Without knowing anything about your code, and who calls what, what about
returning the object?
So:
eval ('var localVar = new ' + objType.value + '(val1, val2, val3');
return localVar;

And when calling the routine:
theObj = myEvalRoutine(vars here);

Could that help?

Regards,
Erwin Moller

I could use a seitch statement, but this approach means I now have an
extra place where I need to change my code so I'd like to afford this
if possible.

Is there another approach to creating a new instance of an object
dependant on the value of some variable?

On Jun 12, 4:32 pm, Erwin Moller
<Since_humans_read_this_I_am_spammed_too_m... (AT) spamyourself (DOT) com> wrote:

Gordon schreef:



I have a script that creates new objects based on the value of a form
field. Basically, the code looks like this.

eval ('new ' + objType.value + '(val1, val2, val3');

objType is a select with the different types of objects you can create
as values.

I really don't like using eval, and it's causing problems, like if I
do something like the following:

eval ('localVar = new ' + objType.value + '(val1, val2, val3');

localVar should refer to a variable local to the calling object, but
with the eval approach it ends up being the local variable for the
scope from which the eval was first executed.

Hi Gordon,

So the scope is the problem, not eval? Right?

I guess your eval-code is in some function.

Without knowing anything about your code, and who calls what, what about
returning the object?
So:
eval ('var localVar = new ' + objType.value + '(val1, val2, val3');
return localVar;

And when calling the routine:
theObj = myEvalRoutine(vars here);

Could that help?

Regards,
Erwin Moller



I could use a seitch statement, but this approach means I now have an
extra place where I need to change my code so I'd like to afford this
if possible.

Is there another approach to creating a new instance of an object
dependant on the value of some variable?

The eval was a quick and dirty alternative to a switch statement for
each kind of object that i want to be able to instantize. The problem
is that using the eval is screwing up the scoping rules that normally
apply in javascript that involve coming up with inelegant workarounds,
such as putting stuff that I'd normally keep in the scope of the cose
I'm running in the global scope instead.

Of the two choices the switch statement is probably the better one,
but it adds an extra point in the code that has to be changed should,
for example, I choose to add a new object type. It also adds n lines
of code per object type, and I'd like to find a better solution. I
guess what I really want to be able to do is something like:

var localVar = new objType.value ()

But when I try that I just get an objType.value is not a function
message.

On Jun 12, 5:03 pm, Gordon wrote:
<snip>

The eval was a quick and dirty alternative to a switch
statement for each kind of object that i want to be able
to instantize. The problem is that using the eval is
screwing up the scoping rules that normally apply in
javascript that involve coming up with inelegant workarounds,

The scope used in an eval execution context is the same scope as the
execution context from which eval is called, so what you are saying
does not make sense.

such as putting stuff that I'd normally keep in the scope
of the cose I'm running in the global scope instead.

And that makes even less sense.

Of the two choices the switch statement is probably the better
one, but it adds an extra point in the code that has to be
changed should, for example, I choose to add a new object type.
It also adds n lines of code per object type, and I'd like to
find a better solution. I guess what I really want to be able
to do is something like:

var localVar = new objType.value ()

But when I try that I just get an objType.value is not a
function message.

As is should, but if you provided some context for you code you might
get a decent answer to your question. Which will be how a bracket
notation property access can be employed here. For example, if all
your constructors were globally declared functions and environment was
a web browser were no alternative means of accessing the global object
had been created those constructors could be references as members of
the object referred to by - window -. I.E.:-

var localVar = new window[objType.value]();

Gordon <gordon.mcvey (AT) ntlworld (DOT) com> writes:

I have a script that creates new objects based on the value of a form
field. Basically, the code looks like this.

eval ('new ' + objType.value + '(val1, val2, val3');

Reduce the scope of eval as much as possible (preferably to nothing .
This can be reduced to:
new (eval(objType.value))(val1, val2, val3);
(or, more readable:
var Constructor = eval(objType.value);
... new Constructor(val1, val2, val3) ...
)

objType is a select with the different types of objects you can create
as values.

Who created the objType objects? Why do they contain the names of
functions instead of the functions themselves?
There can be good reasons for it, but if it is possible to rewrite
the data so that the objType contains the actual constructor function
instead of, or in addition to, its name, then many things become
simpler.

localVar should refer to a variable local to the calling object,

What does that mean? Objects do not have variables. Scopes have
variables, and objects have properties.

but with the eval approach it ends up being the local variable for
the scope from which the eval was first executed.

That's what the code would mean without the eval too:

localVar = new (eval(objType.value))(val1,val2,val3);

assigns to the "localVar" variable in the current scope chain.

I could use a seitch statement, but this approach means I now have an
extra place where I need to change my code so I'd like to afford this
if possible.

Your problem is that you need to lookup a function by its name.
The current solution is to look it up in the scope, and store
the function in the global scope.

That causes the same token to do double duty: as a function name and
property of the global object, and as a data value in your data. I
prefer to keep those two namespaces (language/data) separate if
possible.

You could create a lookup-table:

var functionTable = {
objType1: Type1Constructor,
objType2: Type2Constructor,
...
objTypeN: typeNConstructor
};

Then instead of eval, you could do an explicit lookup:

localVar = new (functionTable[objType.value])(val1, val2, val3);

Is there another approach to creating a new instance of an object
dependant on the value of some variable?

To create new objects from a constructor, first find the constructor
function.
That is the part you are using eval for. I recommend against it, but
it might be so much easier than alternatives that you choose to stick
with it. Just keep the scope of eval down, in this case to a single
variable (which can then be replaced by "window[objType.value]"

Then use the "new" operator. If you know the number of arguments (or a
reasonable upper bound on them), then there is no need for eval in that.

Remember that, unlike, e.g., Java, you can use any variable to refer to
the constructor function in the "new" expression, so

function MyType(){}
var t1 = new MyType();
var x = MyType;
var t2 = new x();

is perfectly valid and there is no difference in properties and
prototype chain between the values of t1 and t2.

That is why you can separate the lookup of the constructor from
the creation of a new instance. Constructors are not types, they
are just function value like every other function.

/L
--
Lasse Reichstein Nielsen - lrn (AT) hotpop (DOT) com
DHTML Death Colors: <URL:http://www.infimum.dk/HTML/rasterTriangleDOM.html>
'Faith without judgement merely degrades the spirit divine.'

On Jun 12, 6:54 pm, Lasse Reichstein Nielsen <l... (AT) hotpop (DOT) com> wrote:

Gordon <gordon.mc... (AT) ntlworld (DOT) com> writes:
I have a script that creates new objects based on the value of a form
field. Basically, the code looks like this.

eval ('new ' + objType.value + '(val1, val2, val3');

Reduce the scope of eval as much as possible (preferably to nothing .
This can be reduced to:
new (eval(objType.value))(val1, val2, val3);
(or, more readable:
var Constructor = eval(objType.value);
... new Constructor(val1, val2, val3) ...
)

objType is a select with the different types of objects you can create
as values.

Who created the objType objects? Why do they contain the names of
functions instead of the functions themselves?
There can be good reasons for it, but if it is possible to rewrite
the data so that the objType contains the actual constructor function
instead of, or in addition to, its name, then many things become
simpler.

localVar should refer to a variable local to the calling object,

What does that mean? Objects do not have variables. Scopes have
variables, and objects have properties.

but with the eval approach it ends up being the local variable for
the scope from which the eval was first executed.

That's what the code would mean without the eval too:

localVar = new (eval(objType.value))(val1,val2,val3);

assigns to the "localVar" variable in the current scope chain.

I could use a seitch statement, but this approach means I now have an
extra place where I need to change my code so I'd like to afford this
if possible.

Your problem is that you need to lookup a function by its name.
The current solution is to look it up in the scope, and store
the function in the global scope.

That causes the same token to do double duty: as a function name and
property of the global object, and as a data value in your data. I
prefer to keep those two namespaces (language/data) separate if
possible.

You could create a lookup-table:

var functionTable = {
objType1: Type1Constructor,
objType2: Type2Constructor,
...
objTypeN: typeNConstructor
};

Then instead of eval, you could do an explicit lookup:

localVar = new (functionTable[objType.value])(val1, val2, val3);

Is there another approach to creating a new instance of an object
dependant on the value of some variable?

To create new objects from a constructor, first find the constructor
function.
That is the part you are using eval for. I recommend against it, but
it might be so much easier than alternatives that you choose to stick
with it. Just keep the scope of eval down, in this case to a single
variable (which can then be replaced by "window[objType.value]"

Then use the "new" operator. If you know the number of arguments (or a
reasonable upper bound on them), then there is no need for eval in that.

Remember that, unlike, e.g., Java, you can use any variable to refer to
the constructor function in the "new" expression, so

function MyType(){}
var t1 = new MyType();
var x = MyType;
var t2 = new x();

is perfectly valid and there is no difference in properties and
prototype chain between the values of t1 and t2.

That is why you can separate the lookup of the constructor from
the creation of a new instance. Constructors are not types, they
are just function value like every other function.

/L
--
Lasse Reichstein Nielsen - l... (AT) hotpop (DOT) com
DHTML Death Colors: <URL:http://www.infimum.dk/HTML/rasterTriangleDOM.html
'Faith without judgement merely degrades the spirit divine.'

That was a big help, thanks. I'm not very familiar with eval so didn't
know you could use it like that. It seems that the scope rules are
the same after all, just not what I was expecting in the part of the
code I was working on. It does look a little nicer using the item =
new (eval (objType)) (params) syntax I still would rather eliminate
the eval altogether if at all possible though.

Gordon wrote:

On Jun 12, 5:38 pm, Henry <rcornf... (AT) raindrop (DOT) co.uk> wrote:
On Jun 12, 5:03 pm, Gordon wrote:
[...]
var localVar = new objType.value ()
But when I try that I just get an objType.value is not a
function message.
As is should, but if you provided some context for you code you might
get a decent answer to your question. Which will be how a bracket
notation property access can be employed here. For example, if all
your constructors were globally declared functions and environment was
a web browser were no alternative means of accessing the global object
had been created those constructors could be references as members of
the object referred to by - window -. I.E.:-

var localVar = new window[objType.value]();

Here's the complete code. Bear in mind it's in a very early stage of
development, is very rough and ready, and has lots of code in the
middle of being rewritten

[ca. 400 lines of uncommented clueless source code]

Richard's implicitly asking for "some context" was certainly not to be
misunderstood as a request for your dumping core here and letting others
do your homework, in their leisure-time, for free.

http://www.catb.org/~esr/faqs/smart-questions.html


PointedEars
--
var bugRiddenCrashPronePieceOfJunk = (
navigator.userAgent.indexOf('MSIE 5') != -1
&& navigator.userAgent.indexOf('Mac') != -1
) // Plone, register_function.js:16

In comp.lang.javascript message <4852CC47.3010500 (AT) PointedEars (DOT) de>, Fri,
13 Jun 2008 21:36:39, Thomas 'PointedEars' Lahn <PointedEars (AT) web (DOT) de>
posted:

Richard's implicitly asking for "some context" was certainly not to be
misunderstood as a request for your dumping core here and letting others
do your homework, in their leisure-time, for free.

There is no reason why others should not be allowed to make that choice.
Even for those collecting news via dial-up modem, 400 lines of code foes
not take significant time or money to download. It is observed that,
instead of mellowing with age, you are getting increasingly vicious.

Also, when you have lost your temper to an even greater degree than
usual, your spelling degenerates. The first word of that quote is not
how one should spell "Henry".

KETFOB.

However, Gordon, if you had taken sufficient heed of newsgroup FAQ 2.3,
you would have replaced those tabs by double-spaces, and kept most of
your code lines more-or-less within the recommended 72 columns.

Opera 9.5 is available.

--
(c) John Stockton, nr London, UK. ?@merlyn.demon.co.uk Turnpike v6.05 MIME.
Web <URL:http://www.merlyn.demon.co.uk/> - FAQish topics, acronyms, & links.
Proper <= 4-line sig. separator as above, a line exactly "-- " (SonOfRFC1036)
Do not Mail News to me. Before a reply, quote with ">" or "> " (SonOfRFC1036)