CSV Parser

C++ module for single-threaded CSV parsing.

Usage

A delimited file can be loaded into a csv::dataframe object by calling csv::parse(pathname, delimiter, buffersize). Alternatively, one can call csv::parse(pathname), which will default to delimiter= ',' and buffersize = 1 MB.

Example

import csv;
import std;

int main() {
    std::string filepath = "/path/to/file";
    char delimiter = ',';
    std::size_t buffersize = 1024*1024;

    auto parsed_data = csv::parse(filepath, delimiter, buffersize)
    if (!parsed_data.has_value()) {
        std::println("{}", parsed_data.error().message);
        return -1;
    }

    auto df = std::move(parsed_data.value());
    auto data_view = std::mdspan(df.data.data(), df.rows, df.columns);
}

Building

A build system is provided as a way to run tests and as an example on how to use the module. The project is setup to be built with Meson and requires C++23 support.

The project can be built with

• Linux

  meson setup build
  ninja -C build
  ./build/csv
  

  , which will default to the GCC compiler if available (GCC 16+ required). Alternatively, an LLVM (LLVM 22+ required) build path is provided, available with CXX=clang++ meson setup build.

• Windows

  meson setup build --vsenv
  meson compile -C build
  ./build/csv
  

Release builds are made available through Meson, with the command line argument --buildtype=release during the setup step: meson setup build --buildtype=release

Performance

These are the average duration of a csv::parse() call for each file.

• Setup:
  • Linux 7.0
  • Clang 22.1.6 (-O3 optimization level)
  • Ryzen 4800H
• Results
  • Electric Vehicle Population Data
    • ~110 ms
  • Motor Vehicle Collisions - Crashes
    • ~1240 ms
  • U.S. Chronic Disease Indicators
    • ~210 ms