procedures to combine a few indicators into one

[semio]

use case

For example, we have these indicators:

• population_male by geo, year
• population_female by geo, year
• population_total by geo, year

and we want to combine these into one multidimensional indicator: population by gender, geo, year.

proposal

We need a procedure for combining indicators, as well as a procedure to extract entities from the combined result.

1. the combine_indicators procedure

procedure: combine_indicators
ingredients:
  - input_ingredient
result: output_ingredient
options:
    dictionary:  # a list of indicators to combine, can be inline or a json file
        new_indicator_1:
            - indicator: base_indicator_1_1
              entities:
                  gender: male
                  an_other_entity: foo
            - indicator: base_indicator_1_2
              entities:
                 gender: female
                 an_other_entity: bar
        new_indicator_2:
            - indicator: base_indicator_2_1
              entities:
                 gender: male
                 an_other_entity: foo
            - ...

Considering similar indicators have similar name pattern, we can add a pattern option

procedure: combine_indicators
ingredients:
 - input_ingredient
result: output_ingredient
options:
   pattern_dictionary:
       new_indicator_1:
           pattern: indicator_([0-9a-z]*)_([0-9a-z]*)  # will combine all indicators with this pattern.
           entities: ["gender", "an_other_entities"]  # assign the matched groups to the entities.

for above example, if we have 2 indicators:

• indicator_male_foo
• indicator_female_baz

then indicator_male_foo will be assigned gender = male and an_other_entities = foo and indicator_female_baz will be assigned gender = female and an_other_entities = baz

2. extract_entities

see #56