Re: [Jolie-devel] Jolie 1.2 - Inline trees
A service-oriented programming language.
Brought to you by:
fmontesi
Menu
▾
▴
Re: [Jolie-devel] Jolie 1.2 - Inline trees
|
From: Saverio G. <sav...@gm...> - 2015-03-14 16:56:03
|
I imagine case 1) as a representations of a static tree-like structure. The
order it is "unfolded" when created does not change the final output.
An example of it are XML structures (but not JSON, in JS you can write "var
myJSON = { "a": i++, "b": i++, "c": i++ }", but JS is evil!).
Right now the ITC is not a declaration but a procedure of creation of the
structure. That is why IMHO ITC is a constrained version of 2)
where we only support sequential composition of assignments, with the
difference that instead of writing the composition with ";" we write ",".
Plus, since we already allow assignments to accept aliases and variables in
ITCs I do not see any easy application of static checks on this kind of
structures.
E.g., think of a variable used in the creation of the tree but also used by
one or more parallel threads (globals, parallel composition, etc.). We
cannot say anything
on the content of that tree (damn, dynamic typing :P).
Are we sure we want to avoid option 2)?
P.S. I realised only now that the name I used previously is ICT but the
acronym is Inline Tree Creation, my bad.
On 14 March 2015 at 12:53, Fabrizio Montesi <fam...@gm...> wrote:
> I see ict as different than with, in that with is just a preprocessing
> construct, whereas ict is an expression (and hence has a runtime
> semantics).
>
> I'm scared by having ict too general, it would probably make any static
> analysis more complicated because we may have lots of side effects in an
> ict. With, instead, is completely transparent because it is just a
> preprocessor instruction so only the parser sees it.
>
> The current ict is neither option 1 or 2.
>
> You can assign generic expressions to subnodes so the order matters:
>
> x << { .x = i++, .y = i++ }
>
> I think in general we should allow also for nested deep copies and, very
> important for writing efficient code, aliases: .x -> z
>
> Ideally ict will become powerful enough that we can use it for the
> implementation of the protocol configuration block in communication ports
> (which are now implemented as a with instead, IIRC).
>
> However that will cause problems because we're using the ; there as a
> separator, so bye bye backward compatibility.
>
> My reasoning for not using ; as a separator for ict is that it may look
> like somebody can write in there a generic code block, but that's not true.
>
> Not using any separator as in types may be the best solution, but it just
> looks weird in code imho... :-/
>
> Or, maybe you guys can convince me of using ; ;-)
>
> This is very soft ground, so please continue sharing your thoughts. Once
> the syntax is out for the first release of Jolie 1.2 I'd like it to be
> stable.
>
> Sent from my phone.
> On Mar 14, 2015 9:50 AM, "Saverio Giallorenzo" <
> sav...@gm...> wrote:
>
>> I see your point. I think it depends on which kind of construct we want
>> the inline tree creation (ICT) to be. I see two options:
>>
>> 1) ICT is used only to assign constant values. In this case it does not
>> matter the order in which the assignments are computed as they are
>> independent. That is the case to use commas "," or even nothing, like the
>> syntax of types.
>>
>> 2) ICT is "with" on steroids. The result of its evaluation returns a tree
>> but its body is a Jolie procedure, hence it can contain ";", "|" and input
>> choices to compose statements in which ".any.path" is interpreted and
>> evaluated in a similar way as "with". In that case we could have something
>> like:
>>
>> animals << {
>> .pet[ 0 ].name = "cat";
>> getAnimal@PetShop()( .pet[ 1 ].name );
>> {
>> .wild[ i++ ].name = "tiger" |
>> .wild[ i++ ].name = "lion"
>> } // super-dangerous non-deterministic block :)
>> }
>>
>>
>> Personally, I find the second option intriguing, but I do not have a
>> precise idea of the possibile havocs it might lead to (maybe none).
>>
>> What do you think about it?
>>
>> On 14 March 2015 at 09:17, Matthias Dieter Wallnöfer <
>> mwa...@ya...> wrote:
>>
>>> We could also try to imitate "with", which requires the semicolons, but
>>> its block content is also not limited to assignments.
>>>
>>> > with ( animals ){
>>> > .pet[ 0 ].name = "cat";
>>> > .pet[ 1 ].name = "dog";
>>> > println@Console(.pet[0].name)();
>>> > .wild[ 0 ].name = "tiger";
>>> > .wild[ 1 ].name = "lion"
>>> > }
>>>
>>> Saverio Giallorenzo schrieb:
>>> > If I may toss an idea, what about making the syntax of inline trees as
>>> > close al possible to that of type declaration?
>>> >
>>> > E.g., we have type
>>> >
>>> > type myType: string {
>>> > .x[ 1, * ]: int
>>> > .y[ 1, 3 ]: void {
>>> > .value*: double
>>> > .comment: string
>>> > }
>>> > }
>>> >
>>> > and we write
>>> >
>>> > myTree << "Root" {
>>> > .x[0]=1
>>> > .x[1]=2
>>> > .y << {
>>> > .value[0]=1.1
>>> > .value[1]=2.1
>>> > .value[2]=3.1
>>> > .comment = "This is a comment"
>>> > }
>>> > }
>>> >
>>> > Granted a comma does not make such a big difference, as an user I would
>>> > wander why one syntax needs the comma and the other does not.
>>> >
>>> > BTW, should the deep-copy operator work inside inline trees or shall we
>>> > allow only assignments?
>>> >
>>> > Best,
>>> > sg
>>> >
>>> >
>>> > On 12 March 2015 at 10:10, Matthias Dieter Wallnöfer
>>> > <mwa...@ya... <mailto:mwa...@ya...>> wrote:
>>> >
>>> > ";" is more about sequential execution for us, "," is our delimiter
>>> > (consider also "for") so I would stuck with ",".
>>> >
>>> > Matthias
>>> >
>>> >
>>> > Fabrizio Montesi <fam...@gm... <mailto:fam...@gm...
>>> >>
>>> > schrieb am 9:30 Donnerstag, 12.März 2015:
>>> >
>>> >
>>> >
>>> > Hi all,
>>> >
>>> > Any ideas on my last comment in the e-mail (comma vs semi-colon)?
>>> > I'd like some feedback on that particular syntax before making it
>>> > official on the docs site.
>>> >
>>> > Even an "I like it" would make me feel better. Break the silence.
>>> ;-)
>>> >
>>> > Cheers,
>>> > Fabrizio
>>> >
>>> >
>>> > On Sat, Mar 7, 2015 at 3:44 PM, Fabrizio Montesi
>>> > <fam...@gm... <mailto:fam...@gm...>> wrote:
>>> >
>>> > Hi all,
>>> > >
>>> > >I committed a series of patches I had lingering around for a while
>>> > on inline trees.
>>> > >
>>> > >Now we can write stuff like this:
>>> > >
>>> > >a.left << "Left" { .x = 1, .y = 2, .y.left = "y_l", .y.right =
>>> "y_r" };
>>> > >
>>> > >
>>> > >valueToPrettyString@StringUtils( { .x = 1, .y = 2 } )( s );
>>> > >
>>> > >
>>> > >
>>> > >
>>> > >This is obtained by introducing a new kind of expression
>>> > (InlineTreeExpression) that takes a value for the root node and a
>>> > series of assignments in between curly brackets { }.
>>> > >
>>> > >
>>> > >You can thus write an inline tree wherever you can write an
>>> expression.
>>> > >
>>> > >
>>> > >Let me know what you think.
>>> > >
>>> > >
>>> > >In particular, I tried to follow what they do in other language by
>>> > using the comma to separate assignments, but I'm not very happy
>>> that
>>> > it is inconsistent with what we do in protocol configurations where
>>> > we use ; instead.
>>> > >
>>> > >
>>> > >Feedback is welcome.
>>> > >
>>> > >
>>> > >Cheers,
>>> > >Fabrizio
>>> >
>>> >
>>> >
>>> ------------------------------------------------------------------------------
>>> > Dive into the World of Parallel Programming The Go Parallel
>>> Website,
>>> > sponsored
>>> > by Intel and developed in partnership with Slashdot Media, is your
>>> > hub for all
>>> > things parallel software development, from weekly thought
>>> leadership
>>> > blogs to
>>> > news, videos, case studies, tutorials and more. Take a look and
>>> join the
>>> > conversation now. http://goparallel.sourceforge.net/
>>> >
>>> > _______________________________________________
>>> > Jolie-devel mailing list
>>> > Jol...@li...
>>> > <mailto:Jol...@li...>
>>> > https://lists.sourceforge.net/lists/listinfo/jolie-devel
>>> >
>>> >
>>> ------------------------------------------------------------------------------
>>> > Dive into the World of Parallel Programming The Go Parallel
>>> Website,
>>> > sponsored
>>> > by Intel and developed in partnership with Slashdot Media, is your
>>> > hub for all
>>> > things parallel software development, from weekly thought
>>> leadership
>>> > blogs to
>>> > news, videos, case studies, tutorials and more. Take a look and
>>> join the
>>> > conversation now. http://goparallel.sourceforge.net/
>>> > _______________________________________________
>>> > Jolie-devel mailing list
>>> > Jol...@li...
>>> > <mailto:Jol...@li...>
>>> > https://lists.sourceforge.net/lists/listinfo/jolie-devel
>>> >
>>> >
>>>
>>>
>>
|