Struct CsvReader 

pub struct CsvReader { /* private fields */ }

Zero-copy CSV parser for byte slices.

Design

CsvReader operates directly on a &[u8] byte slice — no file I/O, no intermediate String allocation per field during the scan phase. Fields are referenced as sub-slices of the original input and parsed in a single pass.

Type inference

Each column’s type is inferred from the first data row:

• All digits (optionally signed, one optional .) → Float
• All digits (optionally signed, no .) → Int (but stored as Float for numeric safety — explicit Int columns can be forced via CsvConfig)
• "true" / "false" / "1" / "0" → Bool
• Anything else → Str

Example

let csv = b"name,age,score\nAlice,30,9.5\nBob,25,8.1";
let df = CsvReader::new(CsvConfig::default()).parse(csv)?;
assert_eq!(df.nrows(), 2);

Implementations

impl CsvReader

pub fn new(config: CsvConfig) -> Self

Create a new CsvReader with the given configuration.

pub fn parse(&self, input: &[u8]) -> Result<DataFrame, DataError>

Parse a CSV byte slice into a DataFrame.

Errors

Returns DataError::InvalidOperation if:

• The input is empty.
• A data row has fewer fields than the header.