phpserialize-rs 0.1.2

High-performance PHP serialize/unserialize parser written in Rust with Python bindings

phpserialize-rs

High-performance PHP serialize/unserialize parser written in Rust with Python bindings.

Features

• Zero-copy parsing - Minimal memory allocations for maximum performance
• Full PHP serialize support - All types including objects, references, and PHP 8.1 enums
• UTF-8 aware - Proper handling of multi-byte characters (Korean, Chinese, etc.)
• Auto-unescape - Automatic detection and handling of DB-escaped strings
• Auto-fallback - Automatic recovery from encoding mismatches (e.g., EUC-KR to UTF-8)
• Error recovery - Configurable error handling for malformed data
• PySpark integration - Ready-to-use Arrow-optimized UDFs for Databricks/Spark workloads

Installation

Python

pip install phpserialize-rs

Rust

[dependencies]
php-deserialize-core = "0.1"

Quick Start

Python

from php_deserialize import loads, loads_json

# Basic usage
data = b'a:2:{s:4:"name";s:5:"Alice";s:3:"age";i:30;}'
result = loads(data)
print(result)  # {'name': 'Alice', 'age': 30}

# Direct JSON conversion (optimized for Databricks)
json_str = loads_json(data)
print(json_str)  # {"name":"Alice","age":30}

# Handle DB-escaped strings automatically
escaped = b'"a:1:{s:4:""key"";s:5:""value"";}"'
result = loads(escaped)  # Auto-unescapes
print(result)  # {'key': 'value'}

# Auto-fallback for encoding mismatches (no option needed!)
# Handles data serialized with EUC-KR but stored as UTF-8
mismatch = b's:4:"\xed\x95\x9c\xea\xb8\x80";'  # "한글" with wrong length
result = loads(mismatch)  # Automatically recovers
print(result)  # '한글'

# Strict mode (disable auto-fallback)
result = loads(data, strict=True)  # Fails on length mismatch

# Error handling options
result = loads(data, errors="replace")  # Replace invalid UTF-8
result = loads(data, errors="bytes")    # Return bytes for invalid UTF-8

PySpark / Databricks

from php_deserialize.spark import php_to_json
from pyspark.sql.functions import get_json_object

# Convert PHP serialize to JSON (Arrow-optimized UDF)
df = spark.table("bronze.my_table")
df = df.withColumn("data_json", php_to_json("serialized_column"))

# Extract fields from JSON
df = df.withColumn("name", get_json_object("data_json", "$.name"))
df = df.withColumn("age", get_json_object("data_json", "$.age"))

df.display()

For Databricks installation:

%pip install phpserialize-rs

Rust

use php_deserialize_core::{from_bytes, PhpValue};

fn main() -> Result<(), Box<dyn std::error::Error>> {
    let data = br#"a:2:{s:4:"name";s:5:"Alice";s:3:"age";i:30;}"#;
    let value = from_bytes(data)?;

    if let PhpValue::Array(items) = value {
        for (key, val) in items {
            println!("{:?} => {:?}", key, val);
        }
    }

    Ok(())
}

Supported PHP Types

Type	PHP Format	Example
Null	N;	N;
Boolean	b:0; / b:1;	b:1;
Integer	i:<value>;	i:42;
Float	d:<value>;	d:3.14;
String	s:<len>:"<data>";	s:5:"hello";
Array	a:<count>:{...}	a:1:{i:0;s:3:"foo";}
Object	O:<len>:"<class>":<count>:{...}	Object with properties
Reference	R:<index>; / r:<index>;	Circular references
Enum (PHP 8.1+)	E:<len>:"<Class:Case>";	E:10:"Status:Active";

Performance

Benchmarked on Apple M1 Pro:

Operation	Throughput
Simple array	~1.5 GB/s
Nested structure	~800 MB/s
Large string	~2.0 GB/s

Compared to php2json (Python):

• 10-50x faster for typical workloads
• 100x faster for large arrays

Error Handling

The library provides detailed error messages for debugging:

from php_deserialize import loads, PhpDeserializeError

try:
    loads(b"invalid data")
except PhpDeserializeError as e:
    print(f"Parse error at position {e.position}: {e.message}")

DB Escape Handling

When data is exported from databases (MySQL, PostgreSQL), strings may be double-quoted and escaped:

Original: a:1:{s:4:"key";s:5:"value";}
DB Export: "a:1:{s:4:""key"";s:5:""value"";}"

The library automatically detects and handles this format:

# Both work identically
loads(b'a:1:{s:4:"key";s:5:"value";}')
loads(b'"a:1:{s:4:""key"";s:5:""value"";}"')

Contributing

Contributions are welcome! Please see CONTRIBUTING.md for guidelines.

License

Licensed under either of:

• Apache License, Version 2.0 (LICENSE or http://www.apache.org/licenses/LICENSE-2.0)
• MIT license (LICENSE or http://opensource.org/licenses/MIT)

at your option.