mini\Table\Index\BTreeLeafPage final class

Documentation

Parsed leaf page - stores raw page data and parses lazily.

Uses object pooling to avoid GC overhead during scans.

Page format:

  • Header: type(1) + count(2) + offsets((n+1) * 2) + rowIdCounts(n * 2)
  • Entry: rowIds(8 each) + key - no per-entry metadata

All header metadata unpacked in single call for efficiency. Uses 0-based indexing: offsets at 0..n, rowIdCounts at n+1..2n. Arrays reused via pooling - count tracks valid entries, stale data ignored.

Properties (9)

private static array $pool
private static int $poolCount
public string $data

Raw page data

public int $count

Entry count

public array $meta
public array $entries
public bool $entriesBuilt

Whether entries array is valid for current page data

public array $writeBuffer

Parent internal node (null for root). Not serialized.

Methods (12)

public static fromRaw()

Documentation missing

public buildEntries()

Build and cache all entries for efficient scan iteration.

public getKeyAt()

Get just the key at 0-based index (for binary search comparisons).

public getEntry()

Get single entry by 0-based index (for yielding rowIds after match).

public release()

Documentation missing

public asString()

Serialize to binary page format using precomputed meta.

public calculateSize()

Calculate the size this page will have when serialized.

public static fromPool()

Get a leaf page from pool (or create new).

public rebuildMeta()

Rebuild meta from current entries array.

public importWriteBuffer()

Import writeBuffer into entries, maintaining sorted order.

public splitOversized()

Split this leaf into pages that fit within maxSize.

private static fromBatch()

Create a new leaf page from a batch of entries.

Source

src/Table/Index/BTreeIndex.php:20-445