Skip to content

How to get the rows behind a number

oda.records() returns the rows a summed indicator adds up, with the same filters and unit as oda.get(). Pass columns= to bring extra table columns along, such as a CRS project title.

List the activities behind a figure

                                  sector_name                                                  project_title  value
3                 Basic health infrastructure                   FOURNITURE EQUIPEMENT MEDICAL SANTE PUBLIQUE   6.83
1                             Basic nutrition                                             P209 - AAP - Kenya   1.08
20  Promotion of mental health and well-being  FID - Lever les obstacles aux soins de santé mentale au Kenya   0.65

import oda_data as oda

rows = oda.records(
    "oecd.crs.bilateral_oda",
    years=2023,
    providers="FRA",
    recipients="KEN",
    sectors=oda.sectors.health,
    columns=["project_title", "crs_id"],
)
top = rows.sort_values("value", ascending=False)
print(top[["sector_name", "project_title", "value"]].head(3).round(2).to_string())

Each row carries every dimension of the indicator, each code with its name beside it, then the extra columns, then the indicator, basis, unit, and value columns. records() has no by= and no group totals.

Check that the rows add up

The rows of a records() call sum to the get() call with the same filters.

9.500114 9.500114

import oda_data as oda

filters = dict(years=2023, providers="FRA", recipients="KEN", sectors=oda.sectors.health)

rows = oda.records("oecd.crs.bilateral_oda", **filters)
total = oda.get("oecd.crs.bilateral_oda", by=["provider"], **filters)
print(round(rows.value.sum(), 6), round(total.value.iloc[0], 6))

Choose extra columns

columns= takes any column of the table the indicator reads. Ask for a name that does not exist and the error lists every column the rows hold. On the CRS these include project_title, short_description, long_description, crs_id, project_number, agency_name, channel_reported_name, and keywords.

A column that records() already produces, such as provider_name, raises UnsupportedQueryError, since two columns would share one name.

Convert the rows

unit= converts each row with its provider's factor for its year, as get() does before it sums.

                    sector_name currency    prices  base_year     value
0   Basic drinking water supply      EUR  constant       2023  0.030000
1  Health personnel development      EUR  constant       2023  0.277365
2              Higher education      EUR  constant       2023  0.055259

import oda_data as oda

rows = oda.records(
    "oecd.crs.bilateral_oda",
    years=2023,
    providers="FRA",
    recipients="KEN",
    unit=oda.constant("EUR", 2023),
)
print(rows[["sector_name", "currency", "prices", "base_year", "value"]].head(3))

Which indicators have records

records() works on sum and method indicators, which covers every CRS, Multisystem, DAC1, and DAC2A count and one.crs.imputed_multilateral_oda. DAC1 and DAC2A publish one row per provider, year, and line, so their records hold no more detail than get(). A ratio such as oecd.dac1.oda_gni_ratio and a derived indicator such as one.dac1.core_oda raise UnsupportedQueryError. Indicators gives each indicator's kind.

See also