geruh commented on code in PR #2951: URL: https://github.com/apache/iceberg-python/pull/2951#discussion_r2725943373
########## tests/benchmark/test_memory_benchmark.py: ########## @@ -0,0 +1,286 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +"""Memory benchmarks for manifest cache efficiency. + +These benchmarks reproduce the manifest cache memory issue described in: +https://github.com/apache/iceberg-python/issues/2325 + +The issue: When caching manifest lists as tuples, overlapping ManifestFile objects +are duplicated across cache entries, causing O(N²) memory growth instead of O(N). + +Run with: uv run pytest tests/benchmark/test_memory_benchmark.py -v -s -m benchmark +""" + +import gc +import tracemalloc +from datetime import datetime, timezone + +import pyarrow as pa +import pytest + +from pyiceberg.catalog.memory import InMemoryCatalog +from pyiceberg.manifest import _manifest_cache + + +def generate_test_dataframe() -> pa.Table: + """Generate a PyArrow table for testing, similar to the issue's example.""" + n_rows = 100 # Smaller for faster tests, increase for more realistic benchmarks + + return pa.table( + { + "event_type": ["playback"] * n_rows, + "event_origin": ["origin1"] * n_rows, + "event_send_at": [datetime.now(timezone.utc)] * n_rows, + "event_saved_at": [datetime.now(timezone.utc)] * n_rows, + "id": list(range(n_rows)), + "reference_id": [f"ref-{i}" for i in range(n_rows)], + } + ) + + [email protected] +def memory_catalog(tmp_path_factory: pytest.TempPathFactory) -> InMemoryCatalog: + """Create an in-memory catalog for memory testing.""" + warehouse_path = str(tmp_path_factory.mktemp("warehouse")) + catalog = InMemoryCatalog("memory_test", warehouse=f"file://{warehouse_path}") + catalog.create_namespace("default") + return catalog + + [email protected](autouse=True) +def clear_caches() -> None: + """Clear caches before each test.""" + _manifest_cache.clear() + gc.collect() + + [email protected] +def test_manifest_cache_memory_growth(memory_catalog: InMemoryCatalog) -> None: + """Benchmark memory growth of manifest cache during repeated appends. + + This test reproduces the issue from GitHub #2325 where each append creates + a new manifest list entry in the cache, causing memory to grow. + + With the old caching strategy (tuple per manifest list), memory grew as O(N²). + With the new strategy (individual ManifestFile objects), memory grows as O(N). + """ + df = generate_test_dataframe() + table = memory_catalog.create_table("default.memory_test", schema=df.schema) + + tracemalloc.start() + + num_iterations = 50 + memory_samples: list[tuple[int, int, int]] = [] # (iteration, current_memory, cache_size) + + print("\n--- Manifest Cache Memory Growth Benchmark ---") + print(f"Running {num_iterations} append operations...") + + for i in range(num_iterations): + table.append(df) + + # Sample memory at intervals + if (i + 1) % 10 == 0: + current, _ = tracemalloc.get_traced_memory() + cache_size = len(_manifest_cache) + + memory_samples.append((i + 1, current, cache_size)) + print(f" Iteration {i + 1}: Memory={current / 1024:.1f} KB, Cache entries={cache_size}") + + tracemalloc.stop() + + # Analyze memory growth + if len(memory_samples) >= 2: + first_memory = memory_samples[0][1] + last_memory = memory_samples[-1][1] + memory_growth = last_memory - first_memory + growth_per_iteration = memory_growth / (memory_samples[-1][0] - memory_samples[0][0]) + + print("\nResults:") + print(f" Initial memory: {first_memory / 1024:.1f} KB") + print(f" Final memory: {last_memory / 1024:.1f} KB") + print(f" Total growth: {memory_growth / 1024:.1f} KB") + print(f" Growth per iteration: {growth_per_iteration:.1f} bytes") + print(f" Final cache size: {memory_samples[-1][2]} entries") + + # With efficient caching, growth should be roughly linear (O(N)) + # rather than quadratic (O(N²)) as it was before + # Memory growth includes ManifestFile objects, metadata, and other overhead + # We expect about 5-10 KB per iteration for typical workloads + # The key improvement is that growth is O(N) not O(N²) + # Threshold of 15KB/iteration based on observed behavior - O(N²) would show ~50KB+/iteration + assert growth_per_iteration < 15000, ( Review Comment: nit: non blocking can make a constant ########## tests/benchmark/test_memory_benchmark.py: ########## @@ -0,0 +1,286 @@ +# Licensed to the Apache Software Foundation (ASF) under one +# or more contributor license agreements. See the NOTICE file +# distributed with this work for additional information +# regarding copyright ownership. The ASF licenses this file +# to you under the Apache License, Version 2.0 (the +# "License"); you may not use this file except in compliance +# with the License. You may obtain a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, +# software distributed under the License is distributed on an +# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY +# KIND, either express or implied. See the License for the +# specific language governing permissions and limitations +# under the License. +"""Memory benchmarks for manifest cache efficiency. + +These benchmarks reproduce the manifest cache memory issue described in: +https://github.com/apache/iceberg-python/issues/2325 + +The issue: When caching manifest lists as tuples, overlapping ManifestFile objects +are duplicated across cache entries, causing O(N²) memory growth instead of O(N). Review Comment: awesome! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
