nirs4all.visualization.prediction_cache module

PredictionCache - Caching layer for expensive prediction computations.

This module provides caching for aggregated predictions to avoid redundant computations when multiple charts use the same data.

class nirs4all.visualization.prediction_cache.CacheKey(aggregate: str | None, rank_metric: str, rank_partition: str, display_partition: str, group_by: Tuple[str, ...] | None, filters: Tuple[Tuple[str, Any], ...])

Bases: object

Immutable cache key for prediction queries.

class nirs4all.visualization.prediction_cache.PredictionCache(max_entries: int = 50)

Bases: object

LRU-style cache for prediction query results.

Caches expensive computations like aggregations to speed up repeated chart rendering with the same parameters.

The cache stores: - Aggregated prediction results (keyed by aggregate + filters) - Top-k results (keyed by full query parameters)

Cache invalidation happens when: - Max size is exceeded (LRU eviction) - clear() is called explicitly - Predictions data is modified (user must call clear())

Example

>>> cache = PredictionCache(max_entries=100)
>>> # First call computes and caches
>>> result = cache.get_or_compute(key, compute_fn)
>>> # Second call returns cached result
>>> result = cache.get_or_compute(key, compute_fn)  # Fast!

__len__() → int

Return number of cached entries.

clear() → None

Clear all cached entries.

get(key: CacheKey) → Any | None

Get cached result if available.

Parameters:

key – Cache key to look up.

Returns:

Cached result or None if not found.

get_or_compute(key: CacheKey, compute_fn: callable) → Any

Get cached result or compute and cache.

This is the primary method for cache usage. It handles the cache lookup, computation, and storage in one call.

Parameters:

• key – Cache key.

• compute_fn – Function to call if cache miss (no args).

Returns:

Result from cache or computation.

get_stats() → Dict[str, Any]

Get cache statistics.

Returns:

Dictionary with hits, misses, evictions, hit_rate, etc.

static make_key(aggregate: str | None, rank_metric: str, rank_partition: str = 'val', display_partition: str = 'test', group_by: str | List[str] | None = None, **filters) → CacheKey

Create a cache key from query parameters.

Parameters:

• aggregate – Aggregation column name or None.

• rank_metric – Metric for ranking.

• rank_partition – Partition for ranking.

• display_partition – Partition for display.

• group_by – Grouping column(s).

• **filters – Additional filter criteria.

Returns:

CacheKey for the query.

put(key: CacheKey, value: Any) → None

Store result in cache.

Parameters:

• key – Cache key.

• value – Result to cache.