Read normalized OJA data from the itaposts DuckDB store

Loads Online Job Advertisement (OJA) data from the shared DuckDB store maintained by the itaposts package and returns it in the same three-table shape produced by normalize_ojv, with the skill-side column names lower-cased so they match the names the rest of skillviz consumes.

Usage

read_oja_itaposts(
  con,
  snapshots = NULL,
  region_code = NULL,
  years = NULL,
  months = NULL,
  verbose = TRUE
)

Arguments

con: A connection to the itaposts DuckDB store, as returned by itaposts::oja_connect().
snapshots: Character vector of snapshot_id values to include (e.g. "ITC4_2026_2"), or NULL (default) for all snapshots in the store.
region_code: Character vector of NUTS-2 region codes to include (e.g. "ITC4"), or NULL (default) for all regions.
years: Integer vector of years to include, or NULL (default) for all available years. Intersected with snapshots against the store's snapshot metadata.
months: Integer vector of months to include, or NULL (default) for all available months.
verbose: Logical scalar. If TRUE (default), itaposts reports the row count of each returned table.

Value

A named list with three data.table elements, all keyed on general_id:

postings: One row per general_id, with the CP2021 and ESCO occupation hierarchies already joined in (including idesco_level_4), the geography, contract, education, sector, experience, working-hours, salary and source attributes, and companyname.
skills: Long-format skill rows (one per posting/skill pair) with lower-cased ESCO skill attribute columns.
companies: The general_id/companyname pair, with missing and empty company names dropped.

Details

This is the recommended way to load OJA data into skillviz. normalize_ojv reads the raw vendor ZIP archives directly and duplicates read/dedup logic that now lives in itaposts; it is retained for backward compatibility only.

The connection is caller-owned, following the itaposts idiom:


con <- itaposts::oja_connect()
on.exit(itaposts::oja_disconnect(con), add = TRUE)

itaposts::oja_normalised() re-emits the skill columns under their historical Lightcast SHOUTING names. This function lower-cases them, so IDESCOSKILL_LEVEL_3, ESCOSKILL_LEVEL_3, ESCO_V0101_OBSOLETE, ESCO_v0101_DESCRIPTION, ESCO_V0101_URI, ESCO_V0101_SKILLSTYPE, ESCO_V0101_REUSETYPE, ESCO_V0101_GREEN and ESCO_V0101_LANGUAGE become idescoskill_level_3, escoskill_level_3, esco_v0101_obsolete, esco_v0101_description, esco_v0101_uri, esco_v0101_skillstype, esco_v0101_reusetype, esco_v0101_green and esco_v0101_language. Only columns actually present are renamed, so the function tolerates changes to the itaposts shim's column set.

pillar_softskills and esco_v0101_ict are not returned: the vendor dropped them in the data_v2 delivery and itaposts has no substitute for them. Downstream code must treat them as optional.

Examples