Supersociedades: balance

Cobertura publicada

Registros

No disponible

Última actualización

No disponible

Frecuencia

yearly

Frescura

Ver estado

Identificador

nit

Búsqueda por nombre

No publicada

Conteo y frescura consultados desde api.latinfo.dev/sources durante el build del 2026-10-05.

Fuente primaria

Cómo se procesa

The custom streaming importer downloads the complete official CSV on the Linux high-disk runner through EasyPipe. Gate 1 uses a bounded 100-record test. Full validation checks the upstream count and unchanged rowsUpdatedAt, parses all eleven columns, canonicalizes nine-digit corporate NIT bodies to ten digits with a leading zero, and retains every report, concept, period and value. Cut dates are normalized to ISO dates. Structural tabs and newlines become spaces; other text and numeric strings are preserved. SHA-256 of the original ordered source values distinguishes rows that differ before structural normalization. GNU sort removes only identical source rows. Malformed or unexpected NITs are skipped with warnings and a reconciliation count; no identifiers are reconstructed.

The binary build and an exhaustive corporate-scope check must agree on row count. Uploads use a quarantined candidate, trusted seal, and atomic content-addressed manifest promotion. No raw CSV archive or lake output is added. The endpoint returns all matching line items as a multi-record array. The auto-generated MCP lookup returns a records array inside a provenance object with attribution, license, unit, coverage and freshness. No name search is available because these datasets contain no legal names. Financial history loads separately in the public Colombia lookup to preserve the core lookup latency. It is not declared universal-ready or mapped to normalized financial metrics.

Scale contract: bounded_local, expected input 6 GiB, largest partition 6 GiB, working-set limit 30 GiB, runner capacity requirement 40 GiB, available-memory admission 7 GiB, V8 heap 5120 MiB. Input estimates use the upstream row count and server-side sample byte-length aggregates, with headroom; working set includes CSV, canonical TSV, sort spill and binaries, not just final R2 size. Actions enforces the disk and RAM limits.

Campos publicados

All values are strings. Examples describe formats and are not asserted company facts.

The publisher FAQ states that financial figures are in thousands of Colombian pesos. Multiplying values by 1,000 is a consumer choice, not a transformation applied by this source.

Limitaciones conocidas

The source contains existing replacement characters in some concepts. These defects are retained rather than guessed. Cut dates span 2015-03-31 through 2025-12-31 in upstream aggregates; some period labels refer to earlier comparative years. Multiple filings or reporting bases can describe the same company and date and must not be treated as duplicates. Only exact original-row duplicates are removed. No company names or shareholder information are present. Wide UTF-8 fields use 16-bit lengths, declared in the index header; concepts are preserved in full. A field or complete record exceeding 65,535 bytes fails visibly rather than being cropped. The scanned CSV count must match the official upstream count, and the deduplicated TSV, binary output and exhaustive compliance check must agree exactly. Health thresholds apply to deduplicated rows, not the raw export. Dataset-specific source endpoints and the separate public history endpoint are provided; universal company financial modules are not claimed.