set_string with custom_mapping does rename field

[albertmink]

Intention is to override JSON data. We read the JSON data, then calculate a new value which then shall be stored in the JSON data.

When setting a string value, its fieldname is changed. Here from fieldData to fielddata. Is this a bug?

DATA ajson   TYPE REF TO zcl_ajson.
DATA path TYPE string.

ajson = zcl_ajson=>parse( iv_json = `{"fieldData":"field_value"}`
                          ii_custom_mapping = zcl_ajson_mapping=>create_camel_case( iv_first_json_upper = abap_false ) ).

" /fieldata = E
ajson->set_string( iv_path = `/fieldData`  iv_val = 'E' ).

Without custom_mapping the field name stays the same.

[sbcgua]

hmm, interesting case, and a bit controversial.

• at the parsing step the json is parsed and the mapper is assigned after. Maybe it's wrong. Maybe I'd expect the camel case converted into snake (abap-like) - field_data
• the set_string part is kind of logical ... somewhat ... to make proper casing and also because of the fact that path may be generated by structure (and not just a single value) and structure comopnent names are uppercased. So, because of this the name is forced to lowercase. Thus it is kind of expected behavior. Otherwise e.g. POSTING_DATE snake case will be converted to POSTINGDATE camel case which is wrong.

What behavior would you expect ? (in particular in the step 1) ? @jrodriguez-rc What do you think ?

P.S. There is a bug though. If you add ajson->set_string( iv_path = '/fielddata' iv_val = '...' ). (lowercased field name) - it will dump. I think due to the step 1 issue.

[albertmink]

I'm not sure, whether I understand all technical aspects properly. But I noticed, that I didn't provided the entire use case. The to_abap part was missing, see below.

Challenge: The JSON data encodes language in ISO 639 and the ABAP logic is based on the char[1] type. I would be a natural choice, if the AJSON could do the conversion. Otherwise, we need to find a workaround in ABAP coding.

So, I expected (overall) that the set_string sets the value and does NOT modify the field name.

DATA:
  ajson TYPE REF TO zcl_ajson,

  BEGIN OF ls_result,
    original_language TYPE c LENGTH 1,
  END OF ls_result.

ajson = zcl_ajson=>parse( iv_json = `{"originalLanguage":"en"}`  " language code encoded  ISO 639
                          ii_custom_mapping = zcl_ajson_mapping=>create_camel_case( ) ).

*ajson->set_string( iv_path = `/originalLanguage`  iv_val = 'E' ).

ajson->to_abap( IMPORTING ev_container = ls_result ).

[jrodriguez-rc]

It seems that the "SET" method always calls to mapper

The SET is calling to the "to_json" transformer, so the attribute is converted to lowercase and there is no underscore.

With the first code, if I try to extract the content of the json to an ABAP structure with the field "field_data" there is a "Path not found" exception because it is trying to find the field "fielddata".

it should be:

• The nodes stored in the ajson instance will be only in JSON format
• The "to_abap" mapping will be called only when calls to method "zcl_json=>zif_ajson~to_abap"
• The "to_json" will be called when you add an ABAP type

This way, if you convert JSON to JSON or ABAP to ABAP the format doesn't change.

But now, when calls to the "SET" method, even if you add a value to a node, it is converted "to_json" because the local class "lcl_abap_to_json" is called.

This needs to be carefully analyzed, to avoid calling the mapping when it doesn't need to call

[sbcgua]

Yeah, the issue appears quite tricky.

@albertmink Just for the sake of search of quick workaround:

• do you consider keeping language in ISO format and converting it separately from json ? I mean, if it were me, I'd keep the external data in ISO. SAP code is less readable and less universal. Just IMHO. But of course I don't see the full picture.
• do you consider working without converter ? I mean all automatic convertors of naming are potentially not 100% reliable, just by definition. Well. Of course it depends on the data ... so maybe it is not a strong advice :)

@jrodriguez-rc do you mean that the conversion should happen only on in/out ? Like parsing/serialization ? And the set should keep the name as is ? This makes sense to me. And also makes the whole thing more maintainable and predictable. Also aligns well with my previous (1) point about "parsing step"

[albertmink]

@sbcgua our boundary conditions are: JSON data is ISO en and ABAP logic is based on E . So we definitely need to transform the language field (separately from JSON) and we are fine with a converter.

The good thing, we could use an already existing class, that does the transformation. However, we would like to set the transformed value, where we encountered the reported issue.

[jrodriguez-rc]

About the language conversion, maybe could do it about the ABAP field type, like timestamp