Skip to main content

Quick start

The default logger exists so that adding logging to a program is one include and one call. Everything else in these docs — named loggers, sinks, patterns, async — is about outgrowing the default logger once a program has more than one component or more than one destination for its logs.

The default logger​

spdlog::info(...), spdlog::warn(...), spdlog::error(...), and friends are free functions that log through spdlog's default logger. Out of the box, that logger writes to stdout, in color, at info level and above (trace and debug are filtered out until you raise the level).

A complete first program​

main.cpp
#include "spdlog/spdlog.h"

int main() {
    spdlog::set_level(spdlog::level::debug);          // now debug+ is visible
    spdlog::set_pattern("[%H:%M:%S] [%^%l%$] %v");     // short, colored pattern

    spdlog::debug("loading configuration from {}", "config.yaml");
    spdlog::info("server listening on port {}", 8080);
    spdlog::warn("{} of {} worker threads started", 3, 4);

    return 0;
}

Naming your own logger​

The default logger is fine for a small program; the moment you have more than one component, a named logger lets you tell them apart in output and configure them independently:

auto log = spdlog::stdout_color_mt("app");
log->info("this line is tagged with the 'app' logger name");

Named loggers are also what makes per-component level control possible — you can turn on debug for "db" while leaving everything else at info. See Creating loggers for the full set of factory helpers.

Logging to a file in three lines​

auto file_log = spdlog::basic_logger_mt("file", "logs/app.log");
file_log->info("this goes to logs/app.log, not stdout");

That's the simplest durable sink. For anything long-running, prefer a rotating or daily sink instead — see File sinks for why basic_logger_mt alone isn't enough for a service that runs for weeks.

Shutting down​

Call spdlog::shutdown() before your program exits. It drops every registered logger and, if you're using async logging, joins the background thread pool so queued messages aren't lost. It matters more than it looks like it should — see Logger lifecycle for what goes wrong if you skip it.

See also​