csv_parser

This library provides a CSV parser to be used in an event loop for processing large amounts of streaming data. The parser itself does not use any internal buffers. Whenever a field value is available, a user's specified callback will be invoked with the field data and its CSV location (row and column).

Building & Installation

Prerequisites:

• gcc
• libtool
• autoconf
• automake
• check (to build tests)

To build and install:

$ sh autogen.sh
$ ./configure
$ make
$ make install

You can optionally build and run tests:

$ make test

Usage

Once installed, you can link your own projects against the library using pkg-config:

$ gcc my_program.c -o my_program $(pkg-config --cflags --libs csv_parser)

Quick Example

Here is a minimal example demonstrating how to initialize the parser and process data:

#include <stdio.h>
#include <string.h>
#include <csv_parser.h>

int on_field(csv_parser_t *parser, const char *data, size_t length, int row, int col) {
    printf("Row: %d, Col: %d, Value: %.*s\n", row, col, (int)length, data);
    return 0;
}

int main() {
    csv_parser_t parser;
    csv_parser_init(&parser);

    csv_parser_settings_t settings = {
        .delimiter = ',',
        .field_cb = on_field
    };

    const char *csv_data = "name,age\nAlice,30\nBob,25";
    csv_parser_execute(&parser, &settings, csv_data, strlen(csv_data));
    csv_parser_finish(&parser, &settings);

    return 0;
}

See the examples/ directory for more advanced use cases.

Benchmarks

The library is highly optimized for throughput, avoiding internal buffering and memory allocations while parsing. We include a benchmarking tool (examples/bench) and several test datasets to measure parsing performance.

Recent benchmark results (run locally on an Apple M4 Pro machine):

Dataset	File Size	Description	Throughput
iris.csv	4 KB	Small, standard dataset	~440 MB/s
large.csv	7.6 MB	100,000 rows, 10 columns (simple text/numbers)	~1.3 GB/s
wide.csv	6.6 MB	1,000 rows, 1,000 columns	~1.3 GB/s
quoted.csv	23.8 MB	100,000 rows, 10 columns with quoted strings & newlines	~1.4 GB/s

You can run these benchmarks yourself using:

$ ./examples/bench examples/large.csv 100