Struct Contract 

pub struct Contract {

    pub time_column: Option<String>,
    pub series_key: Vec<String>,
    pub chunk: Option<Chunk>,
    pub dedup: bool,
    pub ingest_time_column: Option<String>,
    pub retention: Option<Duration>,
    pub target_file_size: u64,
    pub row_group_rows: usize,
    pub auto_maintain: bool,
    pub snapshot_retention: Duration,
    pub min_snapshots: usize,
    pub max_snapshots: usize,
    pub compact_min_files: usize,
    pub compact_cold_after: Option<Duration>,
    pub read_dedup: bool,
    pub read_dedup_strategy: String,
    pub head_format: String,
    pub head_inline_bytes: u64,
    pub format_version: FormatVersion,
    pub keep_snapshots_until: Option<i64>,
    pub rollups: Vec<String>,
    pub rollup_routing: bool,
    pub rollup_refresh_interval: Option<Duration>,
    pub version_check_interval: Duration,
    pub data_page_rows: Option<usize>,
}

The parsed contract of an existing table.

Fields

time_column: Option<String>

The timestamp column, if this is a time-series table.

series_key: Vec<String>

Columns that identify a series.

chunk: Option<Chunk>

Time partition size.

dedup: bool

Remove rows with the same (series_key, time) during compaction; the latest write wins.

ingest_time_column: Option<String>

The column holding each row’s insertion time, if the table has one.

retention: Option<Duration>

Partitions entirely older than this are dropped by maintenance.

target_file_size: u64

Compaction writes files of about this size.

row_group_rows: usize

Rows per Parquet row group.

auto_maintain: bool

Appends run inline maintenance when thresholds are crossed.

snapshot_retention: Duration

Snapshots younger than this are kept.

min_snapshots: usize

The newest snapshots kept regardless of age.

max_snapshots: usize

Snapshots beyond the newest max_snapshots are expired even if younger than the retention.

compact_min_files: usize

Inline auto-maintenance merges a size tier of a partition once it has this many files.

compact_cold_after: Option<Duration>

Inline auto-maintenance rewrites a partition whole once the table holds data newer than the partition’s end by this much (None: never).

read_dedup: bool

The Timelake reader removes copies of a (series…, time) when reading.

read_dedup_strategy: String

How the Timelake reader removes copies: merge or keys-first.

head_format: String

Format of the head object, or none to not write one.

head_inline_bytes: u64

Bytes the head may spend on optional sections (the last-value index).

format_version: FormatVersion

The table format version (see crate::FORMAT_VERSION).

keep_snapshots_until: Option<i64>

No snapshot expires before this time (milliseconds since the epoch).

rollups: Vec<String>

The table’s rollups (table names).

rollup_routing: bool

The Timelake reader may answer queries on this table from its rollups.

rollup_refresh_interval: Option<Duration>

For a rollup: inline maintenance of its raw table refreshes it once its last refresh is this old (None: never).

version_check_interval: Duration

The Timelake reader uses a table version it loaded or checked for this long before checking for a newer one.

data_page_rows: Option<usize>

Rows per Parquet data page in the files written (None: the writer’s default).

Implementations

impl Contract

pub fn from_properties(properties: &HashMap<String, String>) -> Result<Self>

Reads the contract from Iceberg table properties, with defaults for what isn’t set.

pub fn check_readable(&self) -> Result<()>

Errors when this Timelake can’t read the table correctly: its format is newer than FORMAT_VERSION in a way that breaks older readers.

pub fn check_writable(&self) -> Result<()>

Errors when this Timelake must not write to the table: it wouldn’t keep what the table’s newer format relies on.

pub fn is_time_series(&self) -> bool

Whether the table has a time column.

pub fn sort_columns(&self) -> Vec<String>

Sort key: series key columns, then time.

pub fn write_order(&self) -> Vec<String>

The sort key, then the insertion time: in this order (ties kept in commit order) the last row of each (series…, time) is its latest write.

Trait Implementations

impl Clone for Contract

fn clone(&self) -> Self

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for Contract

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.