Download the PHP package iliaal/phpser without Composer

phpser

A PHP serialization extension in C, targeting read-heavy cache workloads where decode time matters more than encode time or payload size.

Why phpser?

PHP cache workloads pay decode cost on every read and encode cost once per write. igbinary has been the default for over a decade, but it leaves performance on the table for common cache shapes: database rowsets, packed numeric arrays, deep-nested structures, and same-class DTO batches (Laravel queue payloads, cached models).

phpser optimizes for decode. It uses pointer-equality dict interning with a bounded content fallback, reuses decoded zend_strings by refcount, pre-sizes hash tables, writes straight into packed zval storage, and emits tagged scalar runs. On the current ARM benchmark, phpser beats igbinary on encode and decode in all ten cases. Integer ranges collapse to constant-size affine runs and decode 91-92% faster, shuffled integer arrays decode 78% faster, deep nesting decodes 26% faster, and DTO batches decode 62-65% faster.

Rowsets keep the pointer-equality fast path for shared strings, while columnar encoding also deduplicates low-cardinality strings and equal packed-string vectors by content. In rowset_distinct_1000, where equal repeated strings have separate allocations, phpser is 56% smaller, 63% faster to encode, and 53% faster to decode than igbinary.

📖 Design writeup: phpser: a fast, secure binary serializer for PHP cache workloads covers what the decoder does differently and why decode time is the metric to optimize. The interactive benchmark page compares phpser against igbinary, native serialize(), and msgpack across every cache shape.

Install

On a minimal PHP image (e.g. php:8.x-cli from Docker Hub), PIE needs a few build tools installed first:

Install unzip on Debian. Composer shells out to /usr/bin/unzip to extract PIE's prebuilt-binary zip. Without it, composer silently falls back to PHP's ZipArchive, which puts the .so at a path PIE doesn't check, and install fails with ExtensionBinaryNotFound even though the download succeeded.

From source

Pre-built binaries

Pre-built .dlls for Windows (PHP 8.2-8.5, TS/NTS, x86 and x64) and .sos for Linux glibc (x86_64, arm64) and macOS arm64 (PHP 8.4-8.5) are attached to each GitHub release. PIE fetches the matching binary automatically and builds from source when no asset matches.

Usage

Basic round-trip. The encoded payload is opaque bytes; treat it as a binary blob in storage (no JSON-safety, no UTF-8 guarantees):

HMAC-signed mode for untrusted storage (memcached, redis, files, cookies). The signed entry points wrap the payload in a constant-time HMAC-SHA256 frame; tampered or foreign-keyed input is rejected before any decoding work runs:

allowed_classes option on both unserialize entry points. Same shape as PHP's native unserialize($payload, ['allowed_classes' => ...]):

When decoding attacker-controlled bytes, use one of the two restricted modes or the signed entry point. See SECURITY.md for the full threat model.

Signed sessions. Set a key and every session write carries an HMAC-SHA256 tag bound to the session ID; a read whose tag does not verify is rejected before decoding, and PHP destroys the session:

Quote the value: an unquoted key containing = (base64) is a php.ini syntax error that drops every line after it. The key must be at least 32 bytes. A shorter one (including On, which INI turns into "1") fails closed: every session_start() fails, new sessions included, and PHP destroys each existing session it reads, so a typo during rollout takes sessions down for everyone. Off, no, false, null, and none become an empty value, which means unsigned; yes and true become "1", like On. Use a distinct key per application and environment: the tag binds a blob to its session ID, not to the app that wrote it. phpser.session_key is PHP_INI_SYSTEM | PHP_INI_PERDIR, so ini_set() from application code cannot change it. Set it, and the handler, for every script that shares the session store: in php.ini or an FPM pool's php_admin_value, not a .user.ini that covers only part of the tree. A script that runs without the key, or switches session.serialize_handler (which any script can, unless it is locked with php_admin_value), reads the store unverified. Lock both settings with php_admin_value where you can.

No application code changes, with two consequences to know about:

• Signing does not stop rollback. The tag binds the data to the session ID, not to a point in time, so anyone who can write the store can put back an older, validly signed copy of a session (for example, from before a role was revoked). That includes a user who saved their own session file and still holds the old ID; regenerating the ID does not stop them, because they can recreate the old record. Keep authorization decisions (roles, revocations, completed 2FA) in server-side state the store cannot override, and check it on each request.
• session_encode() output is bound to the current session ID. session_decode() of that string under a different ID fails and destroys the target session.

APCu. When APCu is loaded, phpser registers itself as an APCu serializer, so APCu stores arrays and objects as phpser frames:

Set apc.serializer globally, in the php.ini the FPM master reads, the same for every pool. APCu binds its serializer once per process: at startup when the value is php (the default), otherwise at the first request that runs with APCu enabled. With opcache.preload, a non-root FPM master runs that request itself before forking, so every pool inherits the php.ini value; a root master with opcache.preload_user preloads in a forked child, so each worker binds at its own first request. A pool's php_admin_value[apc.serializer] can therefore be ignored. The APCu row in the phpser section of phpinfo() reports what phpser observed, not what will happen: inactive when APCu bound php at startup, used by APCu in this process or the parent it forked from once APCu has called phpser in that process or in a parent before it forked, not yet used before then (or when another serializer is bound), and not registered when APCu was not loaded at startup or phpser came from dl(). phpser registers with APCu only when it is loaded at startup through an extension= line.

Without phpser.key the frames are plain and APCu shared memory is trusted, as with APCu's own php serializer. With it, every value APCu passes through the serializer (objects always, and arrays) carries an HMAC-SHA256 tag bound to the key and to the context apcu. A value whose tag does not verify (written under another key, by another serializer, or before the key was set) is never decoded and raises an E_WARNING: apcu_fetch() returns false, as for a miss, but apcu_entry() returns null without calling its generator until the entry expires or is deleted. Under an error handler that turns warnings into exceptions (as Laravel's does), the fetch throws instead, so a hostile pool that overwrites an entry can make another pool's reads of it throw. Scalars, and arrays that a process with no resolvable serializer stored natively, are read back without a check.

This is a defense against object injection between PHP-FPM pools that share one master, and therefore one APCu shared memory segment: give each pool its own php_admin_value[phpser.key] and one pool cannot make another pool's apcu_fetch() instantiate objects. It can still make the other pool's apcu_entry() return null for an entry it overwrites. It does not authenticate APCu data in general, so do not keep authorization data (roles, permissions) in APCu on a master shared with untrusted pools. A key shorter than 32 bytes fails closed: apcu_store() of an array or object returns false with a warning, and nothing is stored unsigned.

Keyed values for other stores. phpser_serialize_keyed($value, $context) and phpser_unserialize_keyed($payload, $context, $options) sign with phpser.key, so the key stays in server configuration instead of application code:

The context names the store and must be 1 to 255 bytes, contain no NUL byte, and not be session. A payload signed for redis does not verify as memcached or apcu. Under the same key and context, the output is byte-identical to what the C API below produces, so PHP code and an extension can share a store.

C API. Other extensions (cache clients, session stores) can serialize through phpser with phpser_api_serialize() and phpser_api_unserialize() from the installed header ext/phpser/php_phpser_api.h. Both follow phpser.key: plain frames when it is unset, context-bound signed frames when it is set. Failed verification returns PHPSER_API_ERR_TAG, distinct from other failures, so a client that falls back to raw bytes on a decode error can treat it as a miss instead. The header documents the ownership and error contract. A consumer declares ZEND_MOD_REQUIRED("phpser"), and extension=phpser must come before the consumer's extension= line: the dependency orders startup among loaded extensions but does not load phpser.

✨ Features

• Signed payloads: phpser_serialize_signed($value, $key) wraps the payload in an HMAC-SHA256 frame; phpser_unserialize_signed($payload, $key) verifies in constant time and rejects tampered or foreign-keyed input before any decoding runs. Use it whenever the store crosses a trust boundary (memcached, redis, files, cookies), anywhere an attacker who can write to the store could feed your decoder a crafted payload. Both sides reject an empty key, since a keyless HMAC is forgeable.
• Signed sessions: phpser.session_key turns on HMAC-SHA256 signing for the session handler. The tag is bound to the session ID, so a forged, corrupted, or copied-from-another-session blob fails verification and is never decoded. One INI line next to session.serialize_handler = phpser, no code changes. It does not stop rollback to an older blob signed for the same ID, so keep authorization state outside the session.
• APCu serializer and keyed signing: apc.serializer = phpser stores APCu arrays and objects as phpser frames. Set phpser.key and the values APCu serializes carry an HMAC-SHA256 tag bound to the key and the apcu context, verified before decoding, so one FPM pool cannot make another pool's apcu_fetch() instantiate objects. Scalars are not signed. phpser_serialize_keyed() / phpser_unserialize_keyed() apply the same key to Redis, memcached, or any other store, and php_phpser_api.h exposes it to other extensions.
• Untrusted input limits: both unserialize entry points take allowed_classes in the same shape as native unserialize(): false rejects all classes, an array allowlists specific ones, true is the default. Disallowed classes decode as __PHP_Incomplete_Class with the original name preserved and are never instantiated. Recursion depth is capped at 512 on encode (throws) and decode (returns null). Duplicate assoc keys collapse to last-write-wins instead of phantom buckets. Wire-controlled key comparisons and collision chains consume bounded work budgets. TABLE/ROWSET rows and fallback object-property copies have a separate materialization budget. Exhausting a budget rejects the frame; these limits do not bound autoloaders or application hooks. See SECURITY.md for exact limits and class-resolution costs.
• Supported versions: PHP 8.2+ (8.3, 8.4, 8.5, master). BSD 3-Clause.

Bench (PHP 8.4.25 aarch64, idle box, 1000 iters, median of 35)

phpser encodes 19-85% faster and decodes 26-92% faster than igbinary across all ten cases. Integer ranges collapse to a 7-byte affine run and decode 91-92% faster; shuffled integers (packed_rand_10k) are 62% smaller, 32% faster to encode, and 78% faster to decode. Deep nesting is 29% faster to encode and 26% faster to decode with a six-byte size difference.

The table's rowset_100 and rowset_1000 reuse PHP literal strings, so the pointer-equality intern path remains the cheapest case, and every row shares one literal tags array, which phpser encodes once and shares by refcount on decode. Columnar TAG_TABLE also performs a bounded content-cardinality scan for separately allocated strings and equal packed-string vectors, so rowset_distinct_1000 (distinct string and tags allocations on every row) is 56% smaller, 63% faster to encode, and 53% faster to decode than igbinary.

DTO workloads (Laravel-queue-style payloads, single-class arrays) are 27-34% smaller, 62-65% faster to decode, 19-20% faster to encode than igbinary. Wire-v2 TAG_OBJECT_SLOTS drops the per-property key indices and installs declared values straight into property slots; the dict dedups prop names once, and the class-entry lookup cache amortizes zend_lookup_class_ex across same-typed batches.

bench.php --extended adds single values, wrapped envelopes, signed payloads, collector-on retained decode, and Laravel-app shapes (Eloquent models whose attributes and original share one array, a document wallet, a queue job, a catalog, JWKS). Run it with -d opcache.enable_cli=1 for the opcache configuration.

Sizes are byte-identical on x86 (the wire format is architecture-neutral); the ns/op columns are from an idle aarch64 box, median of 35. For the full four-way picture, phpser vs igbinary vs native serialize() vs msgpack, with size, encode, and decode side by side on every shape, see the interactive benchmark page. Regenerate it with php ... bench.php --html > docs/index.html.

Design highlights

• Pointer-equality dict intern: encoding checks *zend_string == *zend_string first and hashes the bytes only on a miss. Intern cost is near zero for rowset-shaped data where PHP literals share interned zend_strings. Columnar rowsets additionally scan for at most eight distinct string values and prebind equal packed-string vectors, preserving dictionary reuse when database drivers return separate allocations for equal content.
• Front-loaded string dictionary: same shape as igbinary's compact_strings, except phpser emits the table once at the head and reference by varint index from values. Trade-off: not streamable.
• Integer-run compression: constant-stride runs (ranges, constant fills) collapse to a base+step affine tag a few bytes long, bounded by a one-million-element budget the encoder and decoder enforce identically. Other integer runs and rowset id columns store zigzag deltas whenever those encode smaller; wrapping mod-2^64 arithmetic keeps every int64 sequence exact with no overflow checks.
• Refcount reuse of zend_strings on decode: a per-decode cache parallels the dict. The first reference allocates; later ones addref.
• Eligible shared arrays stay shared: a nested array held in several places (a literal list on every row, an Eloquent model's attributes and original) can reuse one encoded and decoded table. Sharing requires a traversal without PHP references; the wire-format section describes eligibility and retries after objects acquire identities.
• GC-quiet decode: a clean decode adds one garbage-collector root for the result instead of one per decoded object, so large object graphs don't trigger collector runs while the application holds them.
• Hardware HMAC: signed payloads use ARMv8 Crypto Extensions or SHA-NI when available and cache the per-key HMAC state. php --ri phpser shows the backend in use (armv8, sha-ni, or ext/hash with the reason).
• HT_IS_PACKED flag check: layout comes from the flag, without scanning buckets.
• arPacked stride awareness: PHP 8+ packed arrays store zvals directly, not Buckets, so the stride is 16, not 32.
• Sparse-packed fallback: arrays with holes (post-unset) preserve original int keys via Assoc rather than silently re-indexing.

Where phpser diverges from igbinary

igbinary is the closest reference point. phpser takes measurable performance in these areas, all shipped:

1. Pre-sized HT + direct arPacked writes on decode. When the wire format declares PACKED_LEN N, allocate the HT once via zend_new_array(N) and write directly into arPacked with ZVAL_* macros. Skips N zend_hash_next_index_insert calls, including their hash computation, growth checks, and capacity tuning.
2. Tagged scalar runs. [1, 2, 3, ...] (1000 longs) emits as a single PACKED_LONGS header + N zigzag varints, not 1000 (tag, varint) pairs. Decode is one tight loop with no per-element tag dispatch.
3. O(1) pointer-hash intern. Open-addressed zend_string* → slot hash, grown without eviction. Hit rate near 100% on literal rowset shapes (PHP interns literals; the same "id" zend_string pointer flows through every row), and unique value strings (names, emails) hit a single-probe miss instead of a linear scan. Separately allocated low-cardinality table columns use a bounded content scan, which puts encode ahead of igbinary on every measured shape. Pointer hits skip the byte-hash entirely.
4. Eager dict materialization with warm hashes. All dict slots are resolved up front during header parse, against the engine's interned-string table first. Property names, class names, and hot literals come back as the engine's own interned strings (no allocation, no refcount traffic, pointer-equality hash lookups), with a regular allocation as the fallback. Hashes are set on both paths; zend_hash_add_new reuses the cached hash.
5. Invariant-gated add_new on assoc decode. Wire-controlled duplicate keys must collapse to last-write-wins rather than produce phantom buckets (count($arr) != count(array_unique(array_keys($arr)))), and canonical integer strings must coerce to integer keys. Authentication proves key possession, not that the bytes came from phpser's encoder: a key holder can sign a handcrafted frame. TAG_ASSOC therefore always uses update semantics; schema-based paths use add_new only after validating distinct, non-numeric keys once at schema read.

6. Inline-short-string tag with upgrade-on-second-encounter. TAG_STR_INLINE (0x0c) and KEY_STR_INLINE (0x02) are emitted on a string's first occurrence; the next occurrence triggers an in-place upgrade to a dict entry, and all subsequent ones emit TAG_STR_DICT. Singletons (e.g. row_X values in a rowset) never hit the upgrade branch. They cost nothing in the dict header. The intern cache doubles as the "seen once?" signal: high bit of idx distinguishes INLINE_EMITTED from DICT_IDX. No pre-pass; single walk of the zval tree as before.

   A count-then-emit pre-pass (tag occurrences first, then emit inline for singletons and dict for repeats) cost ~200 ns per string and ate the per-singleton savings. The single-walk upgrade moved rowset_1000 encode from 8% to 25% faster than igbinary; the later columnar TAG_TABLE format and delta id columns reached -45% size and -58% encode versus igbinary.

7. Skip refcount machinery during build. All zvals built during decode are fresh and unshared until handed back to PHP. Internal writes can skip Z_TRY_ADDREF guards.

Local dev build

The hand-rolled Makefile.dev builds against an in-tree ~/php-src-8.4-opt checkout without phpize/autoconf, for working on the extension and PHP at the same time:

It is kept separate from the Makefile generated by phpize/configure, so a standard source build cannot overwrite the developer build rules. Override PHP_SRC= to target a different in-tree PHP checkout. Load alongside igbinary for the A/B bench:

Session integration is compiled when the target PHP headers define HAVE_PHP_SESSION, with either a phpize or Makefile.dev build; a session-less PHP build omits it. Registration is attempted only when phpser is loaded at startup, not through dl(). php --ri phpser reports whether registration succeeded.

Limitations / known gaps

• The three entry points signal corrupt input differently. phpser_unserialize() silently returns null on malformed, truncated, or over-deep input, indistinguishable from a successfully decoded null. Trailing bytes after a complete top-level value are tolerated on this path (prefix read). phpser_unserialize_signed() throws instead (since 0.4.0), so a legitimately-signed null still decodes cleanly, and both the signed and session paths require exact consumption of the frame. The session handler reports decode failure to the engine (warning; a scalar-root payload fails the read rather than becoming an empty session). When a cache miss must be told apart from a cached null, use the signed entry point. SECURITY.md carries the full contract.
• Recursion depth is capped at 512 on both encode and decode. On decode, anything deeper than 512 nested containers / refs is rejected (returns null) to bound stack consumption against adversarial wire payloads. On encode, input deeper than 512 throws an Exception rather than silently shipping a truncated payload. Object cycles go through the id table and don't count against the cap; only genuinely deep trees hit it. Cache workloads typically nest 5-10 deep.
• Closures, other non-serializable objects, and resources encode as NULL. Native serialize() throws for Closure, Generator, Fiber, and other classes marked not serializable, and writes every resource as i:0;, so native unserialize() returns int(0) where phpser returns NULL. Neither serializer stores the resource id or can restore an open stream or other resource handle. phpser treats all of these as unsupported cache values and writes NULL.
• Unknown classes at decode become __PHP_Incomplete_Class with the original name preserved, matching PHP native unserialize(). This holds for plain objects and __serialize-based objects, and for any shape whose class the allowed_classes filter denies. Two wire shapes differ when the class is merely absent (rather than denied): a legacy Serializable value (TAG_OBJECT_LEGACY) whose class is gone decodes as null in place, and an enum (TAG_ENUM) whose class is gone fails the whole decode, because neither carries a property schema to install into. The positional DTO (TAG_OBJECT_SLOTS, used with or without a built properties table) decodes to an incomplete object that still claims its id for later back-refs; when its class is already loaded the values are installed onto the incomplete, and only when the class is unavailable are the unnamed slot values dropped (phpser does not autoload a class for the sole purpose of recovering property names). Evolve classes append-only if cached payloads must outlive a schema change.
• Enum cases are filtered by allowed_classes. Native unserialize() does not consult the allowlist on its enum path, so a serialized enum is always resurrected there. phpser applies the filter: a disallowed enum decodes to __PHP_Incomplete_Class. Enum cases are inert singletons, so the security delta is small, but list their class names in the allowlist if you rely on enums round-tripping through a filtered decode.
• Strings and blobs are capped at 4 GiB (UINT32_MAX) on the wire. Encoding a longer value fails loud (throws for the userland API, an E_WARNING for the session handler) rather than emitting bytes the decoder would reject. No single cache value realistically approaches this.
• Encode memory is O(distinct strings), uncapped. The encoder interns every distinct string it emits into a cache that grows without eviction for the duration of the call (pre-sized from the top-level value, so the common shapes never regrow mid-encode). A value with N unique strings holds N entries until the encode returns; nothing is retained between calls. Unusually string-diverse values cost temporary memory proportional to their diversity; chunk the value if that matters.
• Wire integers outside the target zend_long range are rejected on 32-bit builds. A payload written by a 64-bit process can carry values a 32-bit zend_long cannot hold. Decode returns null instead of narrowing them, so a cache shared between 32-bit and 64-bit processes has to stay inside the 32-bit range. This covers scalars, packed runs, table columns, and array keys; encoding is unaffected.
• TAG_OBJECT_SLOTS is positional. Eligible typed objects encode their declared properties as values in properties_info_table (declaration) order with no per-property names; decode installs them back in that order. This holds whether or not the object's properties table has been built (get_object_vars(), foreach, var_dump()); only a dynamic property or an uninitialized or unset slot moves an otherwise eligible object to the keyed form. An older payload carrying a prefix of the current effective slot table is accepted and appended properties retain their class defaults. A payload with more slots than the current class is rejected. Nothing else is detectable. The decoder cannot tell an append-at-end from an insertion or a reorder, so any other schema change silently lands values in the wrong slots, including a change that alters the slot count. Inserting $mid between $a and $b, for example, decodes an old two-slot payload as a=1, mid=2, b=<default> with no error. Type declarations catch only the subset of mismatches that fail a coercion. For payloads that outlive a deploy, append properties only at the end of the effective properties_info_table order. Adding a parent property can insert before child slots and break append-only order.

• session.serialize_handler=phpser is shipped (compiled in when the target PHP headers define HAVE_PHP_SESSION, with either a phpize or Makefile.dev build; omitted on session-less PHP builds). Registration requires a startup load, not dl(). phpredis integration is not yet wired; call phpser_serialize_keyed() / phpser_unserialize_keyed() (or the unsigned pair) directly when using the extension with phpredis. The handler serializes $_SESSION as a single phpser array value; it is not byte-compatible with the built-in php, php_binary, or igbinary session formats, so switching handlers doesn't read back sessions written by another. It decodes with all classes allowed and magic methods enabled. Without phpser.session_key, treat the session store as trusted, exactly as with the native php_serialize handler; with it, only blobs this key signed for this session ID are decoded. Setting, changing, or removing the key invalidates every existing session. If session encode fails on coder limits (over-deep or oversized input), the handler persists a 29-byte phpser:session-not-serialized tombstone marker instead of the session, so the next read fails loudly (engine warning, session destroyed) rather than silently resuming an empty session. When a serialization hook throws, the exception propagates and the handler returns no data: PHP 8.6+ then skips the write and the previous session survives, while PHP 8.5 and earlier write an empty session, which the next request reads as new.

• APCu signing covers what APCu serializes. APCu stores null, booleans, integers, floats, and strings itself without calling a serializer, so those values are never signed or verified, and neither are arrays persisted natively by a process whose apc.serializer did not resolve. Entries from apc.preload_path are loaded before phpser registers: arrays and scalars come back unverified. Do not preload objects; APCu 5.1.28 crashed at startup on a preloaded object in testing, with or without phpser.
• Keyed values are bound to the key and context, not to the entry name. Neither APCu nor phpredis passes the cache key to a serializer, so a value validly signed for one entry verifies under any other entry of the same context. See SECURITY.md.

Wire format (V1 / V2)

Varints are LEB128 (unsigned); signed values use zigzag encoding. Tags 0x0a/0x0d/0x0e/0x0f/0x11/0x12 each claim the next id in encounter order, before their contents, so the decoder reconstructs back-refs by counting container tags as it parses. 0x18 SHARED_ARRAY claims its id once its array is complete, after every id claimed inside it; a REF from inside the array to its own id is out of range and rejects. 0x10 REF never claims an id; it is lookup-only.

The encoder emits SHARED_ARRAY only for a non-empty nested array that is pointer-shared (refcount above one, not counting the encoder's own snapshots, or immutable) and whose walk reached no PHP reference; a later visit of the same HashTable emits REF. The top-level value never claims. An array directly holding a reference stays by-value on every visit: COW separation keeps a reference in each copy, and one shared decoded array would lose the aliasing a later write through either copy depends on. The decoder enforces the same traversal rule and rejects a SHARED_ARRAY whose contents create or reuse a reference (NEW_REF, or a REF to a reference). An existing object's back-reference does not traverse its properties, which may hold references. The encoder retries arrays that were initially unshareable once their objects have registered identities, avoiding repeated expansion through those objects. Arrays are pinned exactly like tracked objects, so a freed table's reused address cannot masquerade as a back-reference. Pins activate at the first hook. Before that, the encoder releases its private copies of shared arrays and property tails without the possible-root check, so a full root buffer cannot start a cycle collection whose destructors are user code. A shared array is captured at its first visit; see SECURITY.md for the one case where that differs from native.

The version byte is emitted as 0x02 only when the body actually uses a v2-only tag (0x12–0x18); otherwise it stays 0x01. On decode it is a minimum-reader signal, not a gate: the tag dispatch is version-agnostic, so a hand-built frame carrying a v2 tag under a 0x01 header still decodes. This keeps the version byte additive. Don't rely on it alone to reject a future format; a backwards-incompatible change gets a new version constant and explicit tag rejection.

🔗 Native PHP extensions

Companion native PHP extensions:

• php_excel: native Excel I/O via LibXL. 7-10× faster than PhpSpreadsheet, full XLS/XLSX with formulas, formatting, and styling.
• mdparser: native CommonMark + GFM markdown parser via md4c. 15-30× faster than pure-PHP libraries.
• php_clickhouse: native ClickHouse client speaking the wire protocol directly. Picks up where SeasClick left off.
• pdo_duckdb: PDO driver for DuckDB, analytical SQL in your PHP stack.
• fastjson: drop-in faster ext/json, backed by yyjson. 6× encode, 2.7× decode, 5× validate.
• fast_uuid: high-throughput UUID generation (v1/v4/v7), batched CSPRNG and SIMD hex formatter, ramsey-compatible API.
• fastchart: native chart-rendering extension. 38 chart types behind one fluent OO API, SVG-canonical with PNG/JPG/WebP and optional PDF output.
• statgrab: system statistics (CPU, memory, disk, network) via libstatgrab, no parsing /proc by hand.
• phonetic: native phonetic name matching (Double Metaphone, Beider-Morse, Daitch-Mokotoff, NYSIIS, Match Rating), the encoders PHP core lacks.

Follow on X • Read the writeup • If this cut your cache decode CPU, ⭐ star it!