Awesome
demoinfocs-golang - CS2 Demo Parser
A blazing fast, feature complete and production ready Go library for parsing and analysing of Counter-Strike 2 and Counter-Strike: Global Offensive (CS:GO) demos (aka replays).
Discussions / Chat
You can use Discord or GitHub Discussions to ask questions and discuss ideas about this project.<br> For business inquiries please use the contact information found on the GitHub profile.
Go Get
Counter-Strike 2
go get -u github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs
CS:GO
go get -u github.com/markus-wa/demoinfocs-golang/v3/pkg/demoinfocs
Table of Contents
- Requirements
- Quickstart Guide
- Features
- Performance / Benchmarks
- Versioning
- Projects Using demoinfocs-golang
- Development
- Acknowledgements
- License (MIT)
Requirements
This library requires at least go 1.20
to run.
You can download the latest version of Go here.
Quickstart Guide
-
Download and install the latest version of Go from golang.org or via your favourite package manager
-
Create a new Go Modules project
mkdir my-project
cd my-project
go mod init my-project
go get -u github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs
-
Create a
main.go
file with the example below -
Run
go run main.go
Example
This is a simple example on how to handle game events using this library.
It prints all kills in a given demo (killer, weapon, victim, was it a wallbang/headshot?) by registering a handler for events.Kill
.
Check out the godoc of the events
package for some information about the other available events and their purpose.
package main
import (
"fmt"
"log"
"os"
dem "github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs"
events "github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs/events"
)
func main() {
f, err := os.Open("/path/to/demo.dem")
if err != nil {
log.Panic("failed to open demo file: ", err)
}
defer f.Close()
p := dem.NewParser(f)
defer p.Close()
// Register handler on kill events
p.RegisterEventHandler(func(e events.Kill) {
var hs string
if e.IsHeadshot {
hs = " (HS)"
}
var wallBang string
if e.PenetratedObjects > 0 {
wallBang = " (WB)"
}
fmt.Printf("%s <%v%s%s> %s\n", e.Killer, e.Weapon, hs, wallBang, e.Victim)
})
// Parse to end
err = p.ParseToEnd()
if err != nil {
log.Panic("failed to parse demo: ", err)
}
}
Sample Output
Running the code above will print something like this:
xms <AK-47 (HS)> crisby
tiziaN <USP-S (HS)> Ex6TenZ
tiziaN <USP-S> mistou
tiziaN <USP-S (HS)> ALEX
xms <Glock-18 (HS)> tiziaN
...
keev <AWP (HS) (WB)> to1nou
...
More Examples
Check out the examples folder for more examples, like how to generate heatmaps like this one:
<img alt="sample heatmap" src="https://raw.githubusercontent.com/markus-wa/demoinfocs-golang/master/examples/heatmap/heatmap.jpg" width="50%">Documentation
The full API documentation is available here on pkg.go.dev.
Features
- Game events (kills, shots, round starts/ends, footsteps etc.) - docs / example
- Tracking of game-state (players, teams, grenades, ConVars etc.) - docs
- Grenade projectiles / trajectories - docs / example
- Access to entities, server-classes & data-tables - docs / example
- Access to all net-messages - docs / example
- Chat & console messages <sup id="achat1">1</sup> - docs / example
- Matchmaking ranks (official MM demos only) - docs
- Full POV demo support
- Support for encrypted net-messages (if the decryption key is provided)
- JavaScript (browser / Node.js) support via WebAssembly - example
- Easy debugging via build-flags
- Built with performance & concurrency in mind
- <small id="f1">In MM demos the chat is encrypted, so
ParserConfig.NetMessageDecryptionKey
needs to be set - see alsoMatchInfoDecryptionKey()
.</small>
Performance / Benchmarks
Two of the top priorities of this parser are performance and concurrency.
Here are some benchmark results from a system with an Intel i7 6700k CPU and a SSD disk running Windows 10 and a demo with 85'000 frames.
Overview
Benchmark | Description | Average Duration | Speed |
---|---|---|---|
BenchmarkConcurrent | Read and parse 8 demos concurrently | 2.06 s (per 8 demos) | ~330'000 ticks / s |
BenchmarkDemoInfoCs | Read demo from drive and parse | 0.89 s | ~95'000 ticks / s |
BenchmarkInMemory | Read demo from memory and parse | 0.88 s | ~96'000 ticks / s |
That's almost 1.5 hours of gameplay per second when parsing in parallel (recorded at 64 ticks per second) - or 25 minues per second when only parsing a single demo at a time.
Raw Output
$ go test -run _NONE_ -bench . -benchtime 30s -benchmem -concurrentdemos 8
goos: windows
goarch: amd64
pkg: github.com/markus-wa/demoinfocs-golang
BenchmarkDemoInfoCs-8 50 894500010 ns/op 257610127 B/op 914355 allocs/op
BenchmarkInMemory-8 50 876279984 ns/op 257457271 B/op 914143 allocs/op
BenchmarkConcurrent-8 20 2058303680 ns/op 2059386582 B/op 7313145 allocs/op
--- BENCH: BenchmarkConcurrent-8
demoinfocs_test.go:315: Running concurrency benchmark with 8 demos
demoinfocs_test.go:315: Running concurrency benchmark with 8 demos
PASS
ok github.com/markus-wa/demoinfocs-golang 134.244s
Versioning
We use SemVer for versioning. For the versions available, see the tags on this repository. There is one caveat however: Beta features - which are marked as such via comments and in release notes - may change in minor releases.
Projects Using demoinfocs-golang
- noesis.gg - A suite of explorative tools to help you analyze and improve your CS:GO performance
- esportal.com - An alternative Matchmaking service that aims to provide a friendly environment free from trolls and misbehaving individuals
- esportslab.gg - Building ML/AI tools for professional esports players
- scope.gg - Analytical and advisory service for advanced CS:GO players
- PureSkill.gg - An automated coach to help you get better at CS:GO.
- cs-demo-minifier - Converts demos to JSON, MessagePack and more
- csgo_spray_pattern_plotter - A tool to extract and plot spray patterns from CS:GO replays
- CS:GO Player Skill Prediction - Machine learning master thesis by @quancore about predicting player performance
- csgoverview - A 2D demo replay tool for CS:GO
- csgo-coach-bug-detector - Detects the abuse of an exploit used by some team coaches in professional matches
- megaclan3000 - A CS:GO stats page for clans with recent matches and player statistics
- awpy - A wrapper for the Golang parser in Python
If your project is using this library feel free to submit a PR or send a message via Discord to be included in the list.
Useful Tools & Libraries
- csgo-centrifuge - Get historic radar overview images to make 2D replays & heatmaps accurate on all map versions
- head-position-model - Approximate the player's head position (rather than just the camera position)
Development
Debugging
You can use the build tag debugdemoinfocs
to print out debugging information - such as game events or unhandled demo-messages - during the parsing process.<br>
e.g.
go run -tags debugdemoinfocs examples/print-events/print_events.go -demo example.dem
Side-note: The tag isn't called debug
to avoid naming conflicts with other libs (and underscores in tags don't work, apparently).
To change the default debugging behavior, Go's ldflags
parameter can be used. Example for additionally printing out all server-classes with their properties: -ldflags="-X 'github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs.debugServerClasses=YES'"
.
e.g.
go run -tags debugdemoinfocs -ldflags="-X 'github.com/markus-wa/demoinfocs-golang/v2/pkg/demoinfocs.debugServerClasses=YES'" examples/print-events/print_events.go -demo example.dem
Check out debug_on.go
for any other settings that can be changed.
Testing
Unit Tests
For any new features, Test Driven Development should be practiced where possible. However, due to some design flaws in some parts of the code it's currently not always practical to do so.
Running unit tests:
scripts/unit-tests.sh
# or (identical)
go test -short ./...
Regression Tests
For the full regression suite you will need to download the test demo-set.
Prerequisites:
- Git LFS must be installed
7z
must be in yourPATH
environment variable (p7zip
orp7zip-full
package on most Linux distros)
Downloading demos + running regression tests:
scripts/regression-tests.sh
Updating the default.golden
File
The file test/default.golden
file contains a serialized output of all expected game events in test/cs-demos/default.dem
.
If there is a change to game events (new fields etc.) it is necessary to update this file so the regression tests pass. To update it you can run the following command:
go test -run TestDemoInfoCs -update
Please don't update the .golden
file if you are not sure it's required. Maybe the failing CI is just pointing out a regression.
Generating Interfaces
We generate interfaces such as GameState
from structs to make it easier to keep docs in synch over structs and interfaces.
For this we use @vburenin's ifacemaker
tool.
You can download the latest version here.
After adding it to your PATH
you can use scripts/generate-interfaces.sh
to update interfaces.
Generating Protobuf Code
Should you need to re-generate the protobuf generated code in the msg
package, you will need the following tools:
go install google.golang.org/protobuf/cmd/protoc-gen-go@latest
Make sure both are inside your PATH
variable.
After installing these use go generate ./msg
to generate the protobuf code. If you're on Windows you'll need to run go generate from CMD, not Bash.
Git Hooks
To install some (optional, but quite handy) pre-commit
and pre-push
hooks, you can run the following script.
scripts/git-hooks/link-git-hooks.sh
pre-commit
:
- check if interfaces have been updated
- build the code
- run unit tests
pre-push
:
- run regression tests
Acknowledgements
This library was originally based on <a href="https://github.com/ValveSoftware/csgo-demoinfo" rel="external">Valve's demoinfogo</a> and <a href="https://github.com/StatsHelix/demoinfo" rel="external">SatsHelix's demoinfo</a> (although today it shares little resemblence with these two).
For Counter-Strike 2, dotabuff/manta was an amazing resource for how to parse Source 2 demos and CS2 support would not have been possible without it.<br> I would also like to specifically thank @akiver & @LaihoE for their brilliant help with CS2.
And a very special thanks goes out to all the ⭐contributors⭐️, be it in the form of PRs, issues or anything else.
Further shoutouts go to:
- @JuhaKiili for financial contributions in the form of bug bounties
- @esportalgroup, @refrag & @pureskillgg for offering consulting work that pays my bills
License
This project is licensed under the MIT license.