Notes from PHPers Summit 2023

High performance PHP8 at Scale - Max Małecki

• elastica bundle + doctrine - entity update listener problem
• try to include basic performance benchmarks into CI/CD pipeline
• phpbench.com microbenchmarks
• disk - access latency, cpi - io waits, net - dns
• xhprof - easier to use I thought and free
• websockets -> file descriptors problem
• load testing: siege, ab, gatling (programmable scenarios)
• Phil Karlton
• composer autoloader optimizations for prod
• php.ini configuration: opcache.preload, opcache.jit
• remember not to spam logs

Continuous Integration: Tips & Tricks - Paul Dragoonis

• run steps in parallel, always prepare for whats coming in later steps (for ex.: prepare base image for use in multiple later steps)
• if using Jenkins - configure perf-optimized mode, AWS EC2 plugin, use spot instances
• split work to Big (24CPU) and Small (4CPU) nodes - use labels to distribute steps
• git - do shallow clone (depth 1), skip fetching tags on feature branches (--no-tags)
• git - quick cleanup of local tree: git reset --hard [...] && git clean -fdx
• shellcheck, docker compose -f [file] config
• extract all commands / steps into bash scripts
• execute those bash scripts inside ci/cd configuration and any Dockerfile used
• gain repeatability and easier debuggability this way

API Anatomy - Tomasz Kowalczyk

• API is not the same thing as HTTP REST API, API can be exposed by multiple transport protocols
• main perspectives:
  • Representation: the way of interaction / usage - transport - protocol
  • Expression:
    • intent vs. effect
    • resource vs. procedure
    • fat event vs. thin event
  • Contract: naming, units, inability to express invalid state/intent, composition of methods
  • Evolution:
    • protecting entry-points, avoiding side-channels
    • test and verify only observable behavior (do not test internals)
• gherkin
• kierowniku poratujesz dziesiąteczką?

Czego boi sie TechLead? - Karol Kreft

• link to presentation
• give / create safe space for others to thrive and grow
• acknowledge in front of a team that you don't know all the things
• be curious, search for what interests you most
• fear is constant, try to face it
• look at your past experience - see how many fears you've already overcomed

Watch the Clock - Andreas Heigl

• Stella-Maris Solutions
• PSR-20 FIG - ClockInterface::now():
  • SystemClock
  • FixedClock
• common / shared way to parametrize time across all code - your own and external (non-editable) dependencies
• OS is awesome as long as you are not trying to achieve any political consensus :)
• DateTimeInterface instances can be compared and do not take TZ into account in the process (no need to convert to timestamps before comparing)

Asynchronous PHP - Florian Engelhardt

• bunny/bunny - async communication with RabbitMQ from PHP
• fully async creating and consuming RabbitMQ messages is possible in pure PHP
• ReactPHP - event driven and async programming in PHP

Architecture decisions and their impact - Denis Brumann

• Deptrac - configurable tool for visualising code architecture in form of a DAG

Advanced web layout management - Ivo Lukač

• Netgen - web agency
• Netgen Layouts - product, FOSS vs. enterprise
  • enterprise: previews, caching, translations, user roles, scheduling
• you include routes for page editor UI in your SF app
• it composes nicely with usual SF MVC pattern (you still have control over Controller / Twig template)
• built with idea of joining / fetching content from many sources
• data sources <- Netgen Layouts -> website (data source: local / custom / external)

Phel, a native LISP for PHP - Jose Maria Valera Reales

• Phel - comprehensive implementation of Lisp in PHP, transpiles to PHP code
• provides escape hatches / interop with PHP code / composer libraries
• Phel/LISP syntax - looks hard only on the first look but after few minutes is clear

Historia Pewnego Legacy - Adam Michna

• what to do not to die in legacy project
• pick your fights (leave some things as they are)
• work step by step, little by little
• make progress visible to all stakeholders:
  • measure stuff: performance, complexity, errors freq, publishing freq
  • improvements in measured values = better motivation for dev
  • make progress visible to product owner
• to understand how system works: observe how database changes - Debezium, AWS infra for DB streaming
• backup / revert / gradual deploy
• when proposing a change: justify it from product owner's standpoint