Module table_view 

TableV Module - Windowed View over a Table

TableV is a row-aligned, zero-copy view over a Table, holding the window [offset .. offset+len) across all columns.

Purpose

• Stream or batch rows without copying full columns.
• Do per-partition / per-window work with stable schema metadata (Fields).
• Glue for chunked processing alongside SuperTableV and ArrayV, but focused on windowed zero-copy access within a single Table batch.

Behaviour

• Columns are exposed as ArrayV windows with the same (offset, len).
• Schema (Fields) is retained and shared; column data is not copied.
• Slicing a TableV (from_self) is O(1) - metadata-only.
• to_table() materialises an owned Table copy of the window.

When to use

• Sliding windows, micro-batching, streaming sinks/sources.
• Thread-local views where each worker consumes a row range.
• Building higher-level chunked tables (SuperTableV) from base tables.

Example

use minarrow::{Table, TableV, fa_i32};
use minarrow::ColumnSelection;  // Re-exported at crate root
use std::sync::Arc;

// Build a simple 2-column table
let a = fa_i32!("a", 1, 2, 3, 4, 5);
let b = fa_i32!("b", 10, 20, 30, 40, 50);
let mut tbl = Table::new("T".to_string(), vec![a,b].into());

// View rows 1..4 (3 rows total)
let tv = TableV::from_table(tbl, 1, 3);
assert_eq!(tv.n_rows(), 3);
assert_eq!(tv.n_cols(), 2);
// Access a column window via ColumnSelection trait
let col0 = tv.col_ix(0).unwrap();
assert_eq!(col0.get::<minarrow::IntegerArray<i32>>(0), Some(2));

// Materialise the window as an owned Table copy
let owned = tv.to_table();
assert_eq!(owned.n_rows, 3);

Structs

TableV

TableView