Qs

qs

Quick serialization of R objects.

Basic Usage

library(qs)

# Save any R object
qsave(obj, "data.qs")

# Load
obj <- qread("data.qs")

Presets

# Fast preset (speed priority)
qsave(obj, "data.qs", preset = "fast")

# High preset (compression priority)
qsave(obj, "data.qs", preset = "high")

# Balanced preset (default)
qsave(obj, "data.qs", preset = "balanced")

# Archive preset (maximum compression)
qsave(obj, "data.qs", preset = "archive")

# Uncompressed
qsave(obj, "data.qs", preset = "uncompressed")

Custom Settings

# Custom compression
qsave(obj, "data.qs",
  algorithm = "zstd",      # or "lz4", "zstd_stream", "lz4_stream"
  compress_level = 4,      # 1-22 for zstd, 1-12 for lz4
  nthreads = 4
)

# Shuffle for better compression of numeric data
qsave(obj, "data.qs", shuffle_control = 15)

Supported Objects

# qs supports virtually all R objects:
# - Data frames, tibbles, data.tables
# - Lists, environments
# - Matrices, arrays
# - Functions, formulas
# - S3, S4, R6 objects
# - Factors with levels
# - Attributes preserved

# Complex nested structures
complex_obj <- list(
  df = data.frame(x = 1:100),
  model = lm(y ~ x, data = df),
  func = function(x) x^2,
  env = new.env()
)
qsave(complex_obj, "complex.qs")

Streaming

# Save to connection
con <- file("data.qs", "wb")
qsave(obj, con)
close(con)

# Read from connection
con <- file("data.qs", "rb")
obj <- qread(con)
close(con)

# Save to raw vector
raw_data <- qserialize(obj)

# Load from raw vector
obj <- qdeserialize(raw_data)

Performance

# qs is typically:
# - 3-10x faster than saveRDS
# - Better compression than RDS
# - Supports multithreading

# Benchmark
library(microbenchmark)
microbenchmark(
  qs = qsave(df, "test.qs"),
  rds = saveRDS(df, "test.rds"),
  times = 10
)

Thread Control

# Set threads for save/load
qsave(obj, "data.qs", nthreads = 4)
obj <- qread("data.qs", nthreads = 4)

# Check available threads
qs::qread_threads()

Strict Mode

# Strict mode for reproducibility
qsave(obj, "data.qs", strict = TRUE)

# Validates object integrity on read
obj <- qread("data.qs", strict = TRUE)

Hash Verification

# Save with hash
qsave(obj, "data.qs", check_hash = TRUE)

# Verify on read
obj <- qread("data.qs", validate_checksum = TRUE)

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: LeoLin990405
• Source: LeoLin990405/r-analytics-skill
• License: MIT

Install and usage instructions live in the source repository linked above.