Fix model_json and nitweb #2327
Conversation
| return array | ||
| end | ||
|
|
||
| redef class JsonSerializer |
There was a problem hiding this comment.
To answer your question about the 2 serialization versions, the serialization API is designed for the clients to subclass the engines to customize their behavior. Here you should be able to create a subclass of JsonSerializer instead of a redef (I'd name the subclass FullModelSerializer). You can then use the distinct class (which is also the visitor) to change the serialization behavior without adding methods. You would not need the custom serialize_full, core_serialize_full and accept_full_json_serializer, instead you can use type checks in core_serialize_to:
class FullModelSerializer
super JsonSerializer
end
redef class MPropDef
redef fun core_serialize_to(v) do
super
v.serialize_attribute("is_intro", is_intro)
if v isa FullModelSerializer then
v.serialize_attribute("mclassdef", to_mentity_ref(mclassdef))
v.serialize_attribute("mproperty", to_mentity_ref(mproperty))
v.serialize_attribute("intro", to_mentity_ref(mproperty.intro))
v.serialize_attribute("intro_mclassdef", to_mentity_ref(mproperty.intro.mclassdef))
v.serialize_attribute("mmodule", to_mentity_ref(mclassdef.mmodule))
v.serialize_attribute("mgroup", to_mentity_ref(mclassdef.mmodule.mgroup))
v.serialize_attribute("mpackage", to_mentity_ref(mclassdef.mmodule.mpackage))
end
end
end
This would reduce code duplication and you could remove the intrude importation. The class could also have attributes to further customize the serialization of the model for different usage. And this would even be compatible with a binary engine by simply sharing a common interface!
| v.serialize_attribute("mdoc", mdoc_or_fallback) | ||
| v.serialize_attribute("visibility", visibility.to_s) | ||
| v.serialize_attribute("modifiers", collect_modifiers) | ||
| v.serialize_attribute("location", location) |
|
Reroll with @xymus suggestions on the two serialization versions |
| @@ -355,16 +338,11 @@ end | |||
| redef class POSetElement[E] | |||
There was a problem hiding this comment.
Just a note for the future, POSetElement defines serialization services in its introduction module. However, it writes the fat private attributes (tos, froms) including element, count & poset, instead of flattening the public getters (greaters, smallers) as done here, so the resulting JSON objects differs. This is not a problem, but the difference may cause surprises in the future.
Conflicts would arise if this module is compiled/joined with the compiler or interpreter, or if the deserialization services of POSetElements are changed. (see #2311) Until then, everything is fine.
| @@ -26,7 +26,7 @@ module model_json | |||
|
|
|||
| import model::model_collect | |||
| import json::static | |||
| import json | |||
| intrude import json::serialization_write | |||
There was a problem hiding this comment.
Do you still need the intrude?
Morriar
force-pushed
the
fix-nitweb
branch
2 times, most recently
from
f8b7846 to
4052fdb
Compare
Morriar
force-pushed
the
fix-nitweb
branch
2 times, most recently
from
cdc2fbe to
75234c2
Compare
Was broken since #2312 * Migrate `model_json` to new serialization API * Add tests to `model_json` to ensure this doesn't happen anymore * Migrate `web` package and `nitweb` to new API @xymus a lot of serialization for you. Note: for each entity, I needed two deserialization version `full` and no full. Wasn't sure on how to handle this concern properly with your API Pull-Request: #2327 Reviewed-by: Alexis Laferrière <[email protected]>
|
test this please |
When working with test suites, one can now use `NIT_TESTING_PATH` to retrieve the test suite path.
It can be used to access files based on the current test suite location:
~~~nit
class MyTest
super TestSuite
fun test_suite_path do
assert "NIT_TESTING_PATH".environ.basename == "my_test_suite.nit"
end
end
~~~
Useful for test suites based on model loading like in #2327.
Pull-Request: #2331
Was broken since #2312 * Migrate `model_json` to new serialization API * Add tests to `model_json` to ensure this doesn't happen anymore * Migrate `web` package and `nitweb` to new API @xymus a lot of serialization for you. Note: for each entity, I needed two deserialization version `full` and no full. Wasn't sure on how to handle this concern properly with your API Pull-Request: #2327 Reviewed-by: Alexis Laferrière <[email protected]>
Signed-off-by: Alexandre Terrasa <[email protected]>
Signed-off-by: Alexandre Terrasa <[email protected]>
Signed-off-by: Alexandre Terrasa <[email protected]>
Was broken since #2312 * Migrate `model_json` to new serialization API * Add tests to `model_json` to ensure this doesn't happen anymore * Migrate `web` package and `nitweb` to new API @xymus a lot of serialization for you. Note: for each entity, I needed two deserialization version `full` and no full. Wasn't sure on how to handle this concern properly with your API Pull-Request: #2327 Reviewed-by: Alexis Laferrière <[email protected]>
Was broken since #2312
model_jsonto new serialization APImodel_jsonto ensure this doesn't happen anymorewebpackage andnitwebto new API@xymus a lot of serialization for you. Note: for each entity, I needed two deserialization version
fulland no full. Wasn't sure on how to handle this concern properly with your API