Output both labels and IDs

[haraldgroven]

--fid output as labels and --cid outputs IDs. But what if I need convert JSON-stat and retain both IDs and labels? E.g. sometimes the a Municipality ID is useful and in other cases I need the Municipality name. I may of course write a parser for such cases or run the conversion twice and merge results line by line. But it would be much easier if a flag which would be able to output --fid --cid combined.

[badosa]

--fid output as labels and --cid outputs IDs.

Well, not exactly: both have to do with using IDs instead of labels. The difference is that --cid applies to categories while --fid applies to dimensions. They can be combined.

If you need a richer non-JSON-stat JSON response, I'd recommend to use --meta (jsonstat2arrobj).

[haraldgroven]

ok.

Consider following fragments:

curl https://json-stat.org/samples/oecd.json | jsonstat2arrobj --meta --stream | jq '.'
//  
  {
    "value": 3.003021166,
    "year": "2014",
    "area": "Norway",
    "concept": "unemployment rate"
  },

IDs

url https://json-stat.org/samples/oecd.json | jsonstat2arrobj --cid --meta --stream | jq '.'

// 
    {
      "value": 3.003021166,
      "year": "2014",
      "area": "NO",
      "concept": "UNR"
    },

Desired output (with some name convention of distinguishing labels from IDs):

  {
    "value": 3.003021166,
    "year": "2014",
    "area": "Norway",
    "area_id": "NO",    
    "concept": "unemployment rate"
    "concept_id": "UNR"
  },

Is it possible to add this as a feature for jsonstat2arrobj, or are there any elegant one liners which may achieve the same result from the JSON header added by --meta ?

[badosa]

Using

jsonstat2arrobj oecd.json arrobj.json --meta --cid

it shouldn't be difficult to translate category IDs into labels. For instance, assuming the JSON document is stored as j, the category ID for dimension "area" and observation 101 can be retrieved as

j.meta.dimensions["area"].categories.label[j.meta.dimensions["area"].categories.id.indexOf(j.data[100]["area"])]

Not a "one liner" but easy to implement in any programming language. Shouldn't this suffice?

Your approach is heavier (all observations must have both, id and label). And it's problematic (take into account that dimension IDs can be a string of any kind, not necessarily a single word string).

If I was to develop a heavy approach (to ease processing), wouldn't it make more sense something like

{
  "value": 3.003021166,
  "year": {"id": "2014", "label": "2014"},
  "area": {"id": "NO", "label": "Norway"},
  "concept": {"id": "UNR", "label": "unemployment rate"}
}

?