Storage
timenet.storage ¶source
Shared physical storage helpers for TimeF readers.
parquet ¶source
Bounded caches for packed payload tables and Signal values row groups.
PayloadTableCache ¶source
Cache complete packed payload files up to a byte budget, without retaining file handles.
The least recently used files leave first. The newest file stays even when it alone exceeds the budget, so a large file is still read once.
read_table ¶source
read_table(
version: DatasetVersion, path: str, schema: Schema
) -> Table
Read a payload file once while it remains in the cache.
The file's schema is checked once, when the file enters the cache.
Parameters:
| Name | Type | Description | Default |
|---|---|---|---|
version
|
DatasetVersion
|
The release that holds the payload file. |
required |
path
|
str
|
The payload file, relative to the release root. |
required |
schema
|
Schema
|
The exact Arrow schema the payload file must have, including nullability. |
required |
Returns:
| Type | Description |
|---|---|
Table
|
The cached payload table. |
Raises:
| Type | Description |
|---|---|
TimeFFormatError
|
If the payload file cannot be read or has a different schema. |
SignalRowGroupCache ¶source
Cache values arrays and a bounded number of open Signal shard files.
read_values ¶source
read_values(
version: DatasetVersion, path: str, row_group: int
) -> Array
Read and resolve the values column once per cached row group.
Returns:
| Type | Description |
|---|---|
Array
|
The cached values array. |
Raises:
| Type | Description |
|---|---|
TimeFFormatError
|
If the file, row group, or values column cannot be read. |