csv

In normal mode, decode returns a stream of ok/error tuples, extracting all valid rows even with malformed lines, as emphasized in the README for handling fickle CSV formats.

Uses fast binary matching to parse ~500k rows per second on modest hardware, ensuring it rarely bottlenecks data pipelines, per performance benchmarks.

Designed to work with Elixir streams and enumerables, fitting composably into data processing pipelines without extra overhead, following UNIX philosophy.

Supports custom separators, escape characters, field transformations, and formula unescaping, allowing adaptation to various CSV dialects as shown in decode options.

Allows custom data type encoding via the CSV.Encode protocol, enabling polymorphic output for complex structures, detailed in the encoding section.