A blazing fast, feature complete and production ready Go library for parsing and analysing of Counter-Strike: Global Offensive (CS:GO) demos (aka replays).
Counter-Strike 2 support is experimental.
You can use Discord or GitHub Discussions to ask questions and discuss ideas about this project.
For business inquiries please use the contact information found on the GitHub profile.
go get -u github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs
go get -u github.com/markus-wa/demoinfocs-golang/v3/pkg/demoinfocs
- Requirements
- Quickstart Guide
- Features
- Performance / Benchmarks
- Versioning
- Projects Using demoinfocs-golang
- Development
- Acknowledgements
- License (MIT)
This library requires at least go 1.21 to run.
You can download the latest version of Go here.
-
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 github.com/<YOUR_GITHUB_USER>/my-project
# the module name (github.com/<YOUR_GITHUB_USER>/my-project) can always be changed later
# you can also put example.com/my-project or anything else if you don't plan to publish your project
go get -u github.com/markus-wa/demoinfocs-golang/v4/pkg/demoinfocs
-
Create a
main.gofile with the example below -
Run
go run main.go
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)
}
}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
...
Check out the examples folder for more examples, like how to generate heatmaps like this one:
The full API documentation is available here on pkg.go.dev.
- 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 1 - docs / example
- Matchmaking ranks (official MM demos only) - docs
- Full POV demo support 2
- 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
- In MM demos the chat is encrypted, so
ParserConfig.NetMessageDecryptionKeyneeds to be set - see alsoMatchInfoDecryptionKey(). - Better than some other parsers at the time of writing.
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.
| 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.
$ 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
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.
- 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.
- 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)
You can use the build tag debugdemoinfocs to print out debugging information - such as game events or unhandled demo-messages - during the parsing process.
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.
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 ./...
For the full regression suite you will need to download the test demo-set.
Prerequisites:
- Git LFS must be installed
7zmust be in yourPATHenvironment variable (p7ziporp7zip-fullpackage on most Linux distros)
Downloading demos + running regression tests:
scripts/regression-tests.sh
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.
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.
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.
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
- check if interfaces have been updated
- build the code
- run unit tests
- run regression tests
This library was originally based on Valve's demoinfogo and SatsHelix's demoinfo (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.
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.
This project is licensed under the MIT license.
