Module autoload_cache 

rkyv-backed bytecode cache for autoload functions.

Single-file shard at ~/.cache/zshrs/autoloads.rkyv. Keyed by function name (not file path) — autoload bytecode is identified by the resolved function name, regardless of which fpath dir or .zwc archive it came from.

Storage layout (rkyv archived): AutoloadShard { header: { magic, format_version, zshrs_version, pointer_width, built_at_secs }, entries: HashMap<function_name, AutoloadEntry>, } AutoloadEntry { binary_mtime_at_cache, cached_at_secs, chunk_blob: Vec<u8> }

Inner chunk_blob is bincode-encoded fusevm::Chunk (same constraint as script_cache module — fusevm::Chunk is upstream and only derives serde).

Invalidation:

• zshrs binary mtime newer than binary_mtime_at_cache ⇒ entry stale (any zshrs rebuild silently invalidates the whole shard).
• There is no per-source-file mtime check here. Autoload bodies live in fpath dirs / .zwc archives and the existing compsys::cache::autoloads SQLite row tracks the source file/offset/size. Rebuild logic relies on compinit clearing the whole rkyv shard at recompute time (see AutoloadShardWriter — used by the compinit bulk-prewarm path).

Bulk-write: compinit prewarms 16k+ autoload bytecodes in one go. Per-batch shard rewrites (the SQLite-era pattern) would re-serialize 16k entries 160 times. Instead the new API exposes AutoloadShardWriter: accumulate all (name, blob) pairs in memory, then commit() writes the shard once. The single-add try_save_one path remains for the cold-start case where one autoload at a time is compiled by the interactive shell.

The on-disk shape mirrors ScriptShard — same header, same magic-version-pointer_width discipline, same atomic-rename writes.

Structs

ArchivedAutoloadEntry

An archived AutoloadEntry

ArchivedAutoloadShard

An archived AutoloadShard

ArchivedShardHeader

An archived ShardHeader

AutoloadCache

AutoloadEntry

AutoloadEntryResolver

The resolver for an archived AutoloadEntry

AutoloadShard

AutoloadShardResolver

The resolver for an archived AutoloadShard

MmappedShard

ShardHeader

ShardHeaderResolver

The resolver for an archived ShardHeader

Constants

SHARD_FORMAT_VERSION

SHARD_MAGIC

“ZRAL” little-endian.

Statics

CACHE

Functions

cache_enabled

cached_names

clear

default_cache_path

entry_count

stats

try_load

try_merge_in

Merge new entries into the existing shard. Use this from the compinit BACKFILL path (existing shard has most entries, just adding the missing ones).

try_replace_all

Replace the entire autoload shard with the given entries. Use this from compinit’s bulk pre-warm path — accumulates all (name, chunk_blob) in the entries HashMap and writes the shard exactly once.

try_save_one