Yes, for example (a very naive one); from a pike string "hello world" to <string>hello world</string>

Greetings Coen

On 12-1-2011 14:03, Arne Goedeke wrote:

What do you mean by that? You want to encode arbitrary pike objects in XML?

arne

On Wed, 12 Jan 2011, Coen Schalkwijk wrote:

...

I'm looking for code to translate objects to XML and back with absolutely NO pike specific encoding of any kind. Is this already available or do I have to make this myself?

-- Met vriendelijke groet / With kind regards / mit besten Gr??en,

 Coen Schalkwijk

Software Engineer

coen.schalkwijk@rtl.nl <mailto:coen.schalkwijk@rtl.nl>

coen@rtlinteractief.nl <mailto:coen@rtlinteractief.nl>

+31 (0)35 671 8915

Its for storing objects in a database and usage from within multiple different environments. I'll look into JSON. Does pike offer Object-JSON-Object support? If so, could you help me out with a quick example?

On 12-1-2011 15:12, Arne Goedeke wrote:

I dont know if something like this exists. But have you considered using JSON? It works quite well for all standard types, has far less overhead than xml and is easier to parse.

arne

On Wed, 12 Jan 2011, Coen Schalkwijk wrote:

...

Yes, for example (a very naive one); from a pike string "hello world" to <string>hello world</string>

Greetings Coen

On 12-1-2011 14:03, Arne Goedeke wrote:

...

What do you mean by that? You want to encode arbitrary pike objects in XML?

arne

On Wed, 12 Jan 2011, Coen Schalkwijk wrote:

...

...

I'm looking for code to translate objects to XML and back with

absolutely NO pike specific encoding of any kind. Is this already available or do I have to make this myself?

...

...

-- > Met vriendelijke groet / With kind regards / mit besten

Gr??en,

...

...

...

> > Coen Schalkwijk

Software Engineer

...

 coen.schalkwijk@rtl.nl <mailto:coen.schalkwijk@rtl.nl>
 coen@rtlinteractief.nl <mailto:coen@rtlinteractief.nl>
 +31 (0)35 671 8915

...

>

Thanks for your suggestion. Is it possible in pike to do/use/consume something like 'they' do in dotNET:

[XMLNode name="A"] classA{ [XMLNode name="aVar"] string a; [XMLNode name="cookieCount"] int b; }

On 12-1-2011 16:53, Arne Goedeke wrote:

In any case you will need to know which kinds of objects you want to serialize and come up with some kind of serialization. The Pike JSON module supports encoding arbitrary objects using a callback. But you still need to specify by hand which object variables you want to serialize, etc. Also, when you need to decode the objects again from their json representation you will need to build some code that does this. there is no reasonable way to do this automatically.

You could for instance do something like this

class A { string a, b, c; mapping d, e;

string encode_json() {
    return JSON.encode(([
        "class" : "A",
        "a" : a,
        "b" : b,
        ....
        ]));
}

}

This would then be encoded as a mapping, which you would need to transform back into the object depending on the "class".

arne

On Wed, 12 Jan 2011, Coen Schalkwijk wrote:

...

Its for storing objects in a database and usage from within multiple different environments. I'll look into JSON. Does pike offer Object-JSON-Object support? If so, could you help me out with a quick example?

On 12-1-2011 15:12, Arne Goedeke wrote:

...

I dont know if something like this exists. But have you considered using JSON? It works quite well for all standard types, has far less overhead than xml and is easier to parse.

arne

On Wed, 12 Jan 2011, Coen Schalkwijk wrote:

...

Yes, for example (a very naive one); from a pike string "hello

world" to > <string>hello world</string>

...

...

Greetings Coen On 12-1-2011 14:03, Arne Goedeke wrote: What do you mean by that? You want to encode arbitrary pike

objects > > in

...

...

XML?

...

> arne > On Wed, 12 Jan 2011, Coen Schalkwijk wrote: > > > I'm looking for code to translate objects to XML and

back with > > > absolutely NO pike specific encoding of any kind. Is this already > > > available or do I have to make this myself?

...

...

...

> > -- > Met vriendelijke groet / With kind regards / mit

besten > > Gr??en,

...

...

...

> > > > Coen Schalkwijk Software Engineer > coen.schalkwijk@rtl.nl mailto:coen.schalkwijk@rtl.nl > coen@rtlinteractief.nl mailto:coen@rtlinteractief.nl > +31 (0)35 671 8915 > > > >

On 12/01/11 17:01, Coen Schalkwijk wrote:

Thanks for your suggestion. Is it possible in pike to do/use/consume something like 'they' do in dotNET:

[XMLNode name="A"] classA{ [XMLNode name="aVar"] string a; [XMLNode name="cookieCount"] int b; }

There is no real need for this when it is so easy to look into objects, iterate, etc.

To map special names to variables, just use a mapping:

mapping XMLNames = ([ classA : ([ "a" : "aVar", "b" : "cookieCount" ]), classB : ([ "a" : "aVar", "b" : "somethingElse" ]) ]);

mapping default_XMLNames = ([ "a" : "aVar" ]);

object o = get_object(); mapping o_xml_names = XMLNames[object_program(o)] || default_XMLNames;

foreach (o; string var_name; mixed val) { string name = o_xml_names[var_name]; if (!name) continue; /* add to xml */ }

This way seems preferable to editing and cluttering old code with annotations.

There is also Program.implements() which is probably good to use.

Met vriendelijke groet / With kind regards / mit besten Grüßen,

Coen Schalkwijk Software Engineer

coen.schalkwijk@rtl.nl mailto:coen.schalkwijk@rtl.nl

coen@rtlinteractief.nl mailto:coen@rtlinteractief.nl

+31 (0)35 671 8915

Hi All,

Creating the XML is no problem, I build a nice method that easily and neatly builds a typed XML. Now I'm working on the translation back to pike and that's when I ran into some 'issues', having mainly to do with translating XML to an object. Currently I have only basic types in the objects I want to store (arrays/mappings with int/string keys and values) which can easily translate back to the pike equivalents (provided I use some less pretty code for handling mapping and array creation). but I'd like to have the subject of object XML serialization covered before the expected function request arises :)

If, a generic type of, annotation was available/possible, it would be far more easier to accomplish this (/serialization to any type of output) without having to know or store everything about the objects involved.

Greetings,

Coen

On 12-1-2011 19:44, Johan Björklund wrote:

On 12/01/11 17:01, Coen Schalkwijk wrote:

...

Thanks for your suggestion. Is it possible in pike to do/use/consume something like 'they' do in dotNET:

[XMLNode name="A"] classA{ [XMLNode name="aVar"] string a; [XMLNode name="cookieCount"] int b; }

There is no real need for this when it is so easy to look into objects, iterate, etc.

To map special names to variables, just use a mapping:

mapping XMLNames = ([ classA : ([ "a" : "aVar", "b" : "cookieCount" ]), classB : ([ "a" : "aVar", "b" : "somethingElse" ]) ]);

mapping default_XMLNames = ([ "a" : "aVar" ]);

object o = get_object(); mapping o_xml_names = XMLNames[object_program(o)] || default_XMLNames;

foreach (o; string var_name; mixed val) { string name = o_xml_names[var_name]; if (!name) continue; /* add to xml */ }

This way seems preferable to editing and cluttering old code with annotations.

There is also Program.implements() which is probably good to use.

Another problem is that 'string s;' or 'string s=0;' evaluate to 'int' when using %t and 'zero_type()' doesn't help either.

Not a problem for pike but for any other language using the generated XML.

As a workaround, have you considered using typeof()/_typeof() ?

----8<----8<----8<----

string s; string t=0; string u=""; string v="foo"; zero_type(s);

(1) Result: 0

zero_type(t);

(2) Result: 0

zero_type(u);

(3) Result: 0

zero_type(v);

(4) Result: 0

typeof(s);

(5) Result: string

typeof(t);

(6) Result: string

typeof(u);

(7) Result: string

typeof(v);

(8) Result: string

_typeof(s);

(9) Result: zero

_typeof(t);

(10) Result: zero

_typeof(u);

(11) Result: string

_typeof(v);

(12) Result: string ---->8---->8---->8----

Pity that a zero integer value also results in a zero type:

int t; _typeof(t);

zero

t=0; _typeof(t);

zero

t=2; _typeof(t);

int(2..2)

But I'm glad it works fine for 'true' objects :)

On 13-1-2011 15:21, Bertrand LUPART wrote:

...

Another problem is that 'string s;' or 'string s=0;' evaluate to 'int' when using %t and 'zero_type()' doesn't help either.

Not a problem for pike but for any other language using the generated XML.

As a workaround, have you considered using typeof()/_typeof() ?

----8<----8<----8<----

...

string s; string t=0; string u=""; string v="foo"; zero_type(s);

(1) Result: 0

...

zero_type(t);

(2) Result: 0

...

zero_type(u);

(3) Result: 0

...

zero_type(v);

(4) Result: 0

...

typeof(s);

(5) Result: string

...

typeof(t);

(6) Result: string

...

typeof(u);

(7) Result: string

...

typeof(v);

(8) Result: string

...

_typeof(s);

(9) Result: zero

...

_typeof(t);

(10) Result: zero

...

_typeof(u);

(11) Result: string

...

_typeof(v);

(12) Result: string ---->8---->8---->8----

On Fri, Jan 14, 2011 at 10:42:52AM +0100, Coen Schalkwijk wrote:

...bumped right into a new problem:

mapping(string:mixed) map = ([]); string s; map["test_null"] = s; _typeof(map["test_null"]);

...

zero

which is fine by me, but:

typeof(map["test_null"]);

...

mixed

which is not so nice :( and if the variable s has a value:

why?

map["test_null"] = "no_so_null"; typeof(map["test_null"]);

...

mixed

but then: _ typeof(map["test_null"]);

...

string

this is the expected behaviour. typeof() gives the type of the variable which you above defined as mixed. _typeof() gives the type of the value, which, as we can see, is a string.

Am I asking something extremely stupid if I'd like to know if it's possible to determine the (original) type of the string in the mapping?

what do you mean by original type?

greetings, martin.

On Fri, Jan 14, 2011 at 11:39:21AM +0100, Coen Schalkwijk wrote:

I have a mapping containing different kind of objects. In order to succesfully deserialize any object. I need to know the 'original' type of the object to put in the XML so that the receiving/deserializing language knows whether 1 is a string or an int.

as far as i can tell you are getting that information.

...

...

mapping(string:mixed) map = ([]);

you are declaring that the type of the value is mixed, and typeof() is telling you exactly that:

...

...

typeof(map["test_null"]);

...

mixed

what other information did you expect here?

greetings, martin.

Not only do you not need 'string' to recreate a NULL, but there is nothing 'string' about map["test_null"].

* Variables have types. These can be examined with typeof(). * Values have types. These can be examed with _typeof(). * Variables contains values.

That's all there is. A value does not keep a log of variables it has been stored in, or their types. Consider the following code:

string a; mixed b=3; b=a;

You have two variables, a and b. There types are "string" and "mixed", because that's what you declared them to be. The variable types are properties of the variables themselves, and unrelated to the value stored in them (except that it limits the types of values that you are allowed to store there). The typeof() function can in fact be evaluated on a variable that isn't even instantiated (such as during compilation), and thus does not contain a value at all.

Now, let's consider the values during execution. The variable "a" is uninitialized, and thus contains the "zero" value. If you evaluate a, and call _typeof() on the result, it will return "zero". Because that's the type of this value. It is not, has never been, and will never be a string. The variable "a" can be used to store values of type "string", but it doesn't do so now. For "a" to evaluate to a value of type "string", you need to store such a value in it. Evaluating a variable simple fetches the value stored there, it doesn't alter its type in any way.

The variable b is initialized to "3". So if you evaluate b at this point, and call _typeof() on the result, you will get "int(3..3)". This is the type of the value "3", and has nothing to do with the variable b itself. You just get the 3, and look at its type.

Finally, b is assigned a new value. The expression producing the value is a variable access to "a", so the zero value is fetched from variable "a". It's just a value, and has no connection to "a". The zero value thus computed is then stored in "b". If "b" is evaluated, the resulting value will now be the zero, and if you call _typeof() on it, the result will be "zero". Neither "b", nor the value store in it, has anything to do with strings (well, except that "mixed" is a supertype of "string"), so there is no reason why any type function should return "string".

Now, if you're dealing with introspection, you might want to somehow internalize the fact that "variable a is of type string and contains zero". And that's perfectly fine. But then you need to keep in mind that the variable type and value are two separate features of the variable, and so you need to internalize them separately. You can not magically conjure one up from the other. So what you'd need to do is not make a mapping ([ "x":obj->x ]), but rather ([ "x":({typeof(obj->x),obj->x}) ]). Note also that since typeof() operates on declared types, obj needs to be declared with the actual type of object (as opposed to "object") for this to work.

But that's just if the declared type of variables are actually relevant to the algorithm somehow, of course. And in your case it didn't seem to be.

On Wed, Jan 12, 2011 at 02:15:17PM +0100, Coen Schalkwijk wrote:

Yes, for example (a very naive one); from a pike string "hello world" to <string>hello world</string>

xmlrpc has something like that for the basic types, even functions but not arbitrary objects, though maybe it is enough for you to extend...

greetings, martin.

On Wed, Jan 12, 2011 at 02:30:22PM +0000, Martin Nilsson (Opera Mini - AFK!) @ Pike (-) developers forum wrote:

string encode_xml(mixed data) { string ret = "<pikedata>"; foreach( encode_value(data);; int c ) ret += sprintf("<%d/>", c); return ret + "</pikedata>"; }

but encode_value is pike specific so not useful for sharing data with non-pike sysyems.

greetings, martin.