Benchmarks

A reproducible comparison of PHP TOML libraries for decode and encode throughput.

These are indicative numbers

The figures below come from a single machine and a single run on small documents. They are useful for relative comparison, not as absolute guarantees. Re-run them on your own hardware and workload before relying on any of it — see Reproduce.

Method

• The harness builds a throwaway Composer project that installs every library, then times each one over several rounds and reports the median.
• Two decode documents are used: a baseline TOML 1.0 document and a modern document exercising TOML 1.1 features (inline tables, offset/local date-times). Encoding uses the baseline PHP array.
• Metric: operations per second (higher is better), median wall-clock across 5 rounds.
• Environment for the numbers shown: PHP 8.4, php-collective/toml against the other current PHP libraries.

Results

Decode — baseline document

Library	Ops/s
internal	1031
petalbranch	1029
php-collective	1019
yosymfony	777
devium	435

Decode — modern document (TOML 1.1 features)

Library	Ops/s	Notes
internal	1418
petalbranch	954
php-collective	848
devium	505
yosymfony	—	does not parse the TOML 1.1 syntax used here

Encode — baseline document

Library	Ops/s	Notes
php-collective	17504
devium	15448
internal	5832
petalbranch	5653
yosymfony	—	fluent builder only; no array-to-TOML encode

Takeaways

Encoding

php-collective/toml is the fastest encoder in this comparison — roughly 3x the throughput of internal and petalbranch, and a little ahead of devium.

• Decoding is competitive, not the outright leader. On the baseline document it lands within about one percent of the fastest libraries. On the modern document the leaner pure-decoders (internal, petalbranch) pull ahead.
• That gap is the cost of what the decoder produces: a full AST with trivia, collected diagnostics, and strict validation — work the fastest decoders skip. For analysis, tooling, and source-aware editing this is a deliberate trade; for raw decode-only throughput a leaner library wins.

Reproduce

composer bench:compare

This installs each library into a temporary project, runs the benchmark, and prints a Markdown report. Results vary by host, PHP build, and document shape — always benchmark your own workload before drawing conclusions.