SCHEMAXIS · Declared work. Results you can check.Documentation ↗ · Contact
schemaxis/Agent’s Hands

Documentation / generated/apachearrow-operations.md
Core 11629ae · View or edit Markdown on GitHub ↗

Apachearrow operation reference

Generated from the current declarations. Do not edit this file by hand.

Types, limits and defaults below describe resolved values. Runtime expressions must produce these types. Unknown options are rejected; defaults apply only when omitted. Constructor options must be constant after document composition.

Index#

prepareTable

prepareTable#

Filter, select, rename and prepare an explicitly typed nullable table as independently verified Parquet and a readable report.

Common mistakes and assumptions:

  • Keep source row order, exact wide integers, Unicode text, null masks and UTC microseconds.
  • Only declared int32-to-int64 widening and explicit null replacement; no arbitrary expressions or metadata.

Native binding: pyarrow.csv.read_csv / compute / Table / parquet.write_table. Output kind: ["file","text"].

Inputs: none.

Option Meaning Type / constraints Default
source Required absolute local regular UTF-8 .csv file, at most 8 MiB, with an exact header. File symlinks, BOM, URLs and malformed quoting are rejected. Bytes are captured for standalone verification; changed current bytes invalidate workflow reuse. {"type":"string","minLength":1,"maxLength":4096} required
columns Declare every source field in CSV header order, 1–32 columns and at most 200000 cells/10000 rows. Names are unique. Missing/extra source cells fail even in rows later filtered out. No field metadata or units are accepted. {"type":"array","items":{"type":"object","properties":{"name":{"type":"string","minLength":1,"maxLength":40,"required":true},"type":{"type":"enum","values":["bool","int32","int64","float64","utf8","timestamp_us_utc"],"required":true},"nullable":{"type":"boolean","default":false}}},"minItems":1,"maxItems":32} required
where Optional list of at most eight predicates, default empty (retain all rows). All predicates must pass. Filters inspect source values before selection, cast or null replacement and retain source row order. {"type":"array","items":{"type":"object","properties":{"column":{"type":"string","minLength":1,"maxLength":40,"required":true},"test":{"type":"enum","values":["eq","ne","lt","le","gt","ge","isNull","isValid"],"required":true},"value":{"type":"string","maxLength":4096}}},"maxItems":8} []
select Required ordered list of 1–32 distinct source columns. Each maps column to a unique output name. Omitted source columns are deliberately excluded; output rows retain source order. {"type":"array","items":{"type":"object","properties":{"column":{"type":"string","minLength":1,"maxLength":40,"required":true},"name":{"type":"string","minLength":1,"maxLength":40,"required":true},"cast":{"type":"enum","values":["int64"]},"fillNull":{"type":"string","maxLength":4096}}},"minItems":1,"maxItems":32} required