README

Motivation

It was recently identified that Composer consumes high CPU + memory on packages that have a lot of historical tags. See composer/composer#7577

This means the composer+packagist infrastructure has a scalability issue: as time passes, the list of tags per packages grows, and the "Composer experience" degrades. This is significant for symfony/* today, and will become also a pain for any other packages over time.

symfony/flex solves this issue with a patch from @nicolas-grekas using a new extra parameter extra.symfony.require: symfony/flex#378 and symfony/flex#409

This project extracts this patch into a separete composer plugin for legacy projects (PHP5 and Symony 2/3)

Installation

Step 1: Profile application without the plugin

Open a command console, enter your project directory and execute the following command to profile the current memory and CPU time usage.

$ composer update --profile --ignore-platform-reqs --dry-run
....
[833.9MB/199.98s] Memory usage: 833.86MB (peak: 2811.34MB), time: 199.98s

Write down it to compare with the final step.

Step 2: Download the Bundle

Execute the following command to installs the composer plugin:

$ composer require rubenrua/symfony-clean-tags-composer-plugin

or globally with:

$ composer global require rubenrua/symfony-clean-tags-composer-plugin

Step 3: Configure the new extra parameter

Configure extra.symfony.require with the same symfony version constraints used in the application. For instance, if you are using symfony 2.8, execute the following command to modify the config composer section:

$ composer config extra.symfony.require 2.8.*

Also the SYMFONY_REQUIRE environment variable can be used instead of extra.symfony.require. See symfony/symfony travis configuration for a example.

Step 4: Profile application with the plugin

Finally profile the current memory and CPU time usage. Execute again the following command:

$ composer update --profile --ignore-platform-reqs --dry-run
....
[230.7MB/31.02s] Memory usage: 230.67MB (peak: 387.3MB), time: 31.02s

Please, feel free to comment the issue #3 with your improvement.

Notes

• MIT license.
• Thank you @nicolasgrekas