Temporal provenance and truncation sensitivity in operational government API acquisitions: data, code, and reproducibility materials
收藏资源简介:
Frozen, field-minimized copy of one 400,000-record walk of the public Prozorro/OpenProcurement tender synchronization feed (Ukraine; API version 2.5), retrieved on 13 July 2026. The walk began at the modification-time offset 1 December 2023 and stopped at a fixed ceiling of 400,000 tender identifiers, so the retrieved records carry final dateModified values from 1 to 26 December 2023. It is a capped, modification-ordered acquisition, not a census of Ukrainian public procurement. The deposit preserves the exact order in which the feed returned the tenders (pseudonymized identifiers), a field-minimized JSON Lines snapshot of the 400,000 records, an empty retrieval-failure log, and deterministic derived tables: a 400,000-row tender feed-order table, a 33,148-row eligible award-lot table and a 32,800-row November-December 2023 award-lot table. It also includes offline derivation and verification code, regression tests, pinned environments, SHA-256 manifests, descriptive and diagnostic tables and figures, a data dictionary, a provenance record that states known gaps, and the original acquisition script. Names, contact details, addresses, monetary values, narrative text, documents and procedure labels are excluded; tender, buyer, lot and bid identifiers are pseudonymized with all within-archive links preserved. Version 1.2.0 changes documentation and licensing only; the frozen records, derived data and numerical outputs are identical to version 1.1.0. Data, tables, figures and documentation are licensed under CC BY 4.0 and code under MIT. See README.md, PROVENANCE.md and DATA_DICTIONARY.md in the archive.



