How to cut the separate call chain SCOOPING my application

[Gachoud Philippe]

Hi all,

going through my SCOOP basics I'm actually trying to SCOOP my application which seems to be a big work. 

Into the steps of SCOOPING my app, I have `file_processors` which are going to be `separate file_processor`

1) creating them with `create l_file_processor.make (a_sep_arg)` as far as I understood `a_sep_arg` HAS to be separate, so I have to either change the `make (an_arg: SOME)` to `make (an_arg: separate SOME)` or `make_from_separate (an_arg: separate SOME)` keeping both options. 

Am I right till here?

2) If I refactor my classes to make them able to do that, I'm reaching a point where calling things from libraries like in my case a_json_string.item ("some_json_key") I'm encountering a "Formal argument or tuple field of a separate call should be of a separate type if the actual argument or the source expression is of a reference type." and where the `item_from_separate` function is not defined :-(((

3) Where am I wrong??

4) What are the workarounds?

Thx for your answers

[Gachoud Philippe]

My dirty but working workaround is following to avoid inheriting or rewriting libs. I'm still with my questions about that

feature -- Conversion

non_sep_json_o (a_json_o: separate JSON_OBJECT): JSON_OBJECT

do

check

attached {JSON_OBJECT} non_sep_json_value (a_json_o) as l_res

then

Result := l_res

end

ensure

instance_free: Class

end

non_sep_json_value (a_json_v: separate JSON_VALUE): JSON_VALUE

local

l_json_s: JSON_STRING

l_s: STRING

do

-- - JSON_ARRAY: no assertion

-- - JSON_BOOLEAN: no assertion

-- - JSON_NULL: no assertion

-- - JSON_NUMBER: no assertion

-- - JSON_OBJECT: no assertion

-- - JSON_STRING: no assertion

if attached {separate JSON_ARRAY} a_json_v as l_json then

create {JSON_ARRAY} Result.make (l_json.count)

check

attached {JSON_ARRAY} Result as l_res

then

across

l_json as l_item

loop

check

attached {JSON_VALUE} non_sep_json_value (l_item.item) as l_non_sep_json_value

then

l_res.extend (l_non_sep_json_value)

end

end

end

elseif attached {separate JSON_BOOLEAN} a_json_v as l_json then

create {JSON_BOOLEAN} Result.make (l_json.item)

elseif attached {separate JSON_STRING} a_json_v as l_json then

create l_s.make_from_separate (l_json.item)

create {JSON_STRING} Result.make_from_string (l_s)

elseif attached {separate JSON_NULL} a_json_v as l_json then

create {JSON_NULL} Result

elseif attached {separate JSON_NUMBER} a_json_v as l_json then

if l_json.is_integer then

create {JSON_NUMBER} Result.make_integer (l_json.integer_64_item)

elseif l_json.is_real then

create {JSON_NUMBER} Result.make_real (l_json.real_64_item)

elseif l_json.is_natural then

create {JSON_NUMBER} Result.make_natural (l_json.natural_64_item)

else

create {JSON_NUMBER} Result.make_natural (0)

check

invalid_json_number: False

end

end

elseif attached {separate JSON_OBJECT} a_json_v as l_json then

create {JSON_OBJECT} Result.make_with_capacity (l_json.count)

check

attached {JSON_OBJECT} Result as l_res

then

across

l_json as l_item

loop

check

attached {JSON_VALUE} non_sep_json_value (l_item.item) as l_non_sep_json_value

then

create l_s.make_from_separate (l_item.key.item)

create l_json_s.make_from_string (l_s)

l_res.put (l_non_sep_json_value, l_json_s)

end

end

end

else

create {JSON_ARRAY} Result.make_empty

check

not_implemented: False

end

end

ensure

instance_free: Class

end

[Larry Rix]

So far—I have found ways to only pass {separate STRING} objects as arguments. The ability to do:

create  l_my_string.make_from_separate (a_sep_string)

seems to work pretty well after receiving 'a_sep_string: separate STRING' as an argument.

What I am hoping to build is a class (based on a suggestion from Jocelyn) that will serialize and then deserialize passed objects of ANY type, such that I can have a single class on either side of a Region boundary that handles that work.

[Gachoud Philippe]

Did some more digressions about the topic on https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!topic/eiffel-users/ELJkIQ8hMQI hopefully some mecanism in ANY could see the sun...

[Larry Rix]

Turns out that there is a scoop patterns library that has a great deal of solutions and patterns already built, tested, and ready to go with documentation in the Master Thesis paper plus Betrand's SCOOP paper in addition. I have shared links to all of these in other topics.

--
You received this message because you are subscribed to the Google Groups "Eiffel Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to eiffel-users...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/eiffel-users/033c748d-2b1f-42b6-a263-ac5f56debffc%40googlegroups.com.

[Gachoud Philippe]

I would have needed all this at the beginning... not pretending I got everything!

How and where can we bring all that together for posterity?

--
**********************************************
Philippe Gachoud
Puerto Williams 6657
Las Condes
Santiago de Chile
+56  934022210
ph.ga...@gmail.com
**********************************************
   

To view this discussion on the web visit https://groups.google.com/d/msgid/eiffel-users/CAJ7KgmBrQO-c1CgZXgx3KTiCWzzM0b6evGrBRu86VXahWT0%3DJA%40mail.gmail.com.