Support cache size ratio with reader

README.md

@@ -63,25 +63,30 @@ reader = lcs.TraceReader(
 )
 
 # Step 2: Initialize cache
+# Note: cache_size as float (0.1) means 10% of the reader's total working set size in bytes.
+# To specify an absolute size, pass an integer (e.g., 1024*1024 for 1MB).
 cache = lcs.S3FIFO(
-    cache_size=1024*1024,
+    cache_size=0.1,  # 0.1 = 10% of trace's working set size (requires reader parameter)
     # Cache specific parameters
     small_size_ratio=0.2,
     ghost_size_ratio=0.8,
     move_to_main_threshold=2,
+    reader=reader,  # Required when cache_size is a float ratio
 )
 
 # Step 3: Process entire trace efficiently (C++ backend)
 req_miss_ratio, byte_miss_ratio = cache.process_trace(reader)
 print(f"Request miss ratio: {req_miss_ratio:.4f}, Byte miss ratio: {byte_miss_ratio:.4f}")
 
 # Step 3.1: Process the first 1000 requests
+# Note: cache_size as float means a ratio of the working set size (requires reader parameter)
 cache = lcs.S3FIFO(
-    cache_size=1024 * 1024,
+    cache_size=0.1,  # 10% of trace's working set size
     # Cache specific parameters
     small_size_ratio=0.2,
     ghost_size_ratio=0.8,
     move_to_main_threshold=2,
+    reader=reader,  # Required when cache_size is a float ratio
 )
 req_miss_ratio, byte_miss_ratio = cache.process_trace(reader, start_req=0, max_req=1000)
 print(f"Request miss ratio: {req_miss_ratio:.4f}, Byte miss ratio: {byte_miss_ratio:.4f}")

libcachesim/__init__.pyi

@@ -1,5 +1,5 @@
 from __future__ import annotations
-from typing import bool, int, str, tuple
+from typing import Optional, Callable, Any
 from collections.abc import Iterator
 
 from .libcachesim_python import ReqOp, TraceType, SamplerType
@@ -38,6 +38,85 @@ class CommonCacheParams:
     hashpower: int
     consider_obj_metadata: bool
 
+class ReaderInitParam:
+    ignore_obj_size: bool
+    ignore_size_zero_req: bool
+    obj_id_is_num: bool
+    obj_id_is_num_set: bool
+    cap_at_n_req: int
+    time_field: int
+    obj_id_field: int
+    obj_size_field: int
+    op_field: int
+    ttl_field: int
+    cnt_field: int
+    tenant_field: int
+    next_access_vtime_field: int
+    n_feature_fields: int
+    block_size: int
+    has_header: bool
+    has_header_set: bool
+    delimiter: str
+    trace_start_offset: int
+    binary_fmt_str: str
+    def __init__(
+        self,
+        binary_fmt_str: str = "",
+        ignore_obj_size: bool = False,
+        ignore_size_zero_req: bool = True,
+        obj_id_is_num: bool = True,
+        obj_id_is_num_set: bool = False,
+        cap_at_n_req: int = -1,
+        block_size: int = -1,
+        has_header: bool = False,
+        has_header_set: bool = False,
+        delimiter: str = ",",
+        trace_start_offset: int = 0,
+        sampler: Optional[Any] = None,
+    ): ...
+
+class AnalysisParam:
+    access_pattern_sample_ratio_inv: int
+    track_n_popular: int
+    track_n_hit: int
+    time_window: int
+    warmup_time: int
+    def __init__(
+        self,
+        access_pattern_sample_ratio_inv: int = 10,
+        track_n_popular: int = 10,
+        track_n_hit: int = 5,
+        time_window: int = 60,
+        warmup_time: int = 0,
+    ): ...
+
+class AnalysisOption:
+    req_rate: bool
+    access_pattern: bool
+    size: bool
+    reuse: bool
+    popularity: bool
+    ttl: bool
+    popularity_decay: bool
+    lifetime: bool
+    create_future_reuse_ccdf: bool
+    prob_at_age: bool
+    size_change: bool
+    def __init__(
+        self,
+        req_rate: bool = True,
+        access_pattern: bool = True,
+        size: bool = True,
+        reuse: bool = True,
+        popularity: bool = True,
+        ttl: bool = False,
+        popularity_decay: bool = False,
+        lifetime: bool = False,
+        create_future_reuse_ccdf: bool = False,
+        prob_at_age: bool = False,
+        size_change: bool = False,
+    ): ...
+
 class Cache:
     cache_size: int
     default_ttl: int
@@ -80,113 +159,164 @@ class CacheBase:
     def cache_name(self) -> str: ...
 
 # Core cache algorithms
+class LHD(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
 class LRU(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class FIFO(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class LFU(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class ARC(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class Clock(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, init_freq: int = 0, n_bit_counter: int = 1, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class Random(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 # Advanced algorithms
 class S3FIFO(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, small_size_ratio: float = 0.1, ghost_size_ratio: float = 0.9, move_to_main_threshold: int = 2, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class Sieve(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class LIRS(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class TwoQ(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, a_in_size_ratio: float = 0.25, a_out_size_ratio: float = 0.5, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class SLRU(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class WTinyLFU(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, main_cache: str = "SLRU", window_size: float = 0.01, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class LeCaR(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, update_weight: bool = True, lru_weight: float = 0.5, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class LFUDA(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class ClockPro(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, init_ref: int = 0, init_ratio_cold: float = 0.5, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class Cacheus(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 # Optimal algorithms
 class Belady(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 class BeladySize(CacheBase):
     def __init__(
-        self, cache_size: int, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, n_samples: int = 128, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+# Probabilistic algorithms
+class LRUProb(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, prob: float = 0.5, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+class FlashProb(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, ram_size_ratio: float = 0.05, disk_admit_prob: float = 0.2, ram_cache: str = "LRU", disk_cache: str = "FIFO", admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+# Size-based algorithms
+class Size(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+class GDSF(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+# Hyperbolic algorithms
+class Hyperbolic(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+# Extra deps
+class ThreeLCache(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, objective: str = "byte-miss-ratio", admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+class GLCache(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, segment_size: int = 100, n_merge: int = 2, type: str = "learned", rank_intvl: float = 0.02, merge_consecutive_segs: bool = True, train_source_y: str = "online", retrain_intvl: int = 86400, admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
+    ): ...
+
+class LRB(CacheBase):
+    def __init__(
+        self, cache_size: int | float, default_ttl: int = 25920000, hashpower: int = 24, consider_obj_metadata: bool = False, objective: str = "byte-miss-ratio", admissioner: Optional["AdmissionerBase"] = None, reader: Optional[ReaderProtocol] = None
     ): ...
 
 # Plugin cache
 class PluginCache(CacheBase):
     def __init__(
         self,
-        cache_size: int,
+        cache_size: int | float,
+        cache_init_hook: Callable,
+        cache_hit_hook: Callable,
+        cache_miss_hook: Callable,
+        cache_eviction_hook: Callable,
+        cache_remove_hook: Callable,
+        cache_free_hook: Optional[Callable] = None,
         cache_name: str = "PythonHookCache",
         default_ttl: int = 25920000,
         hashpower: int = 24,
         consider_obj_metadata: bool = False,
-        cache_init_hook=None,
-        cache_hit_hook=None,
-        cache_miss_hook=None,
-        cache_eviction_hook=None,
-        cache_remove_hook=None,
-        cache_free_hook=None,
+        admissioner: Optional["AdmissionerBase"] = None,
+        reader: Optional[ReaderProtocol] = None,
     ): ...
     def set_hooks(self, init_hook, hit_hook, miss_hook, eviction_hook, remove_hook, free_hook=None): ...
 
@@ -246,3 +376,37 @@ class Util:
     def process_trace(
         cache: CacheBase, reader: ReaderProtocol, start_req: int = 0, max_req: int = -1
     ) -> tuple[float, float]: ...
+
+# Admissioners
+class AdmissionerBase:
+    def __init__(self, _admissioner): ...
+    def clone(self): ...
+    def update(self, req: Request, cache_size: int): ...
+    def admit(self, req: Request) -> bool: ...
+    def free(self): ...
+
+class BloomFilterAdmissioner(AdmissionerBase):
+    def __init__(self): ...
+
+class ProbAdmissioner(AdmissionerBase):
+    def __init__(self, prob: Optional[float] = None): ...
+
+class SizeAdmissioner(AdmissionerBase):
+    def __init__(self, size_threshold: Optional[int] = None): ...
+
+class SizeProbabilisticAdmissioner(AdmissionerBase):
+    def __init__(self, exponent: Optional[float] = None): ...
+
+class AdaptSizeAdmissioner(AdmissionerBase):
+    def __init__(self, max_iteration: Optional[int] = None, reconf_interval: Optional[int] = None): ...
+
+class PluginAdmissioner(AdmissionerBase):
+    def __init__(
+        self,
+        admissioner_name: str,
+        admissioner_init_hook: Callable,
+        admissioner_admit_hook: Callable,
+        admissioner_clone_hook: Callable,
+        admissioner_update_hook: Callable,
+        admissioner_free_hook: Callable,
+    ): ...