Introduction

A collection of programs for working with various games in the BEMANI series. This could be untangled quite a bit into various modules that provide simpler pieces. However, this is how it ended up evolving over time. This repository includes utilities for unpacking (and sometimes repacking) various file formats, emulating network services for various games, utilities for sniffing, redirecting and reconstructing network packets, utilities for gathering information about various game music databases and associated tooling that makes developing the previous utilities easier. It is meant to be a complete ecosystem for somebody looking to provide hobby network services to themselves in order to preserve a particular era of gaming that is no longer officially supported.

Thanks to Tau for the great writeup on the binary network format. Thanks to some rando on stack overflow for RC4 code for Python. Thanks to some other rando on stack overflow for sample sniffer code for Python. Thanks to Tau again for the great logging in easerver to compare my original output to. Thanks to PKGINGO for encouragement and well-received excitement about progress. Thanks to Sarah and Alice for being good RE partners for several games, and sharing good finds. Thanks to helvetica for helping with game RE and retweeting cute anime ladies onto my feed every goddamn night.

2dxutils

A utility for unpacking and repacking .2dx files. This isn't the best utility and I think there are more complete and more accurate programs out there. However, they all lack source as far as I could tell, so I developed this. Run it like ./2dxutils --help to see help output and determine how to use this.

afputils

Utilities for working with several animation formats found across a vast range of games. This includes a TXP2 container parser, a GE2D shape parser and an AFP/BSI parser. Together, they make a set of utilities that attempts to work with AFP, the fork of SWF that handles animations in various games. This utility is capable of rendering animations out of IFS and TXP2 files as well as providing decompiled pseudocode for the flash-like bytecode found in many animation files. Note that this format is similar to SWF and thus very complicated. Therefore, it is unlikely that these tools will correctly handle all animations from all games that it encounters. Run it like ./afputils --help to see help output and determine how to use it.

api

Development version of this repository's BEMAPI implementation. Run it like ./api --help to see help output and determine how to use this. Much like "services" and "frontend", this should be pointed at the development version of your services config file, which holds information about the MySQL database that this should connect to as well as what game series are supported. See config/server.yaml for an example file that you can modify.

Do not use this utility to serve production traffic. Instead, see bemani/wsgi/api.wsgi for a ready-to-go WSGI file that can be used with a Python virtualenv containing this project and its dependencies, uWSGI and nginx.

arcutils

A utility for unpacking .arc files. This does not currently repack files. However, the format is so trivial that adding such a feature would be fairly easy. Run it like ./arcutils --help to see help output and determine how to use this.

bemanishark

A wire sniffer that can decode eAmuse packets and print them. Run it on a computer that can sniff traffic between an eAmusement server and a supported game and it will spit out the requests and responses XML-formatted identically to the legacy easerver XML output. This works on both binary and XML traffic. Note that it does not have the capability to sniff SSL-encrypted traffic, so don't even bother trying to run this at an arcade with official support.

Run it like sudo ./bemanishark to invoke. Will run indefinitely until killed (Ctrl-C will suffice). Run like `./bemanishark --help for options. Without options, it assumes you want to sniff port 80 for all addresses. Note that it doesn't support the Base64 binary blob formats found in SN1 and 2. Note also that over time it will start to lose packets. This is a bug that I never figured out, and it appears to be the OS failing to send over some packets resulting in a failure to reassemble the TCP stream.

This utility might be better if rewritten to be a plugin for Wireshark instead of a standalone sniffing utility, but I don't have the time.

binutils

A utility for unpacking raw binxml data (files that use the same encoding scheme as the binary network protocol) to their XML representation. This is useful for examining raw binary blobs or digging into unknown file formats that contain binxml. Run it like ./binutils --help to see help and learn how to use this.

bootstrap

A utility for quickly bootstrapping a local setup's music database from an already running BEMAPI-compatible server. This is better documented in the below "Database Initialization" section.

cardconvert

A command-line utility for converting between card numbers written on the back of a card and the card ID stored in the RFID of the card. Run it like ./cardconvert --help to see how to use this. This will sanitize input, so you can feed it card numbers with or without spaces, and you can mix up 1 and I as well as 0 and O, and it will properly handle decoding. This supports both new and old style e-Amusement cards but does not support the cross-play network cards with five groups of digits on the back of the card.

dbutils

A command-line utility for working with the DB used by "api", "services" and "frontend". This utility includes options for creating tables in a newly-created DB, granting and revoking admin rights to the frontend, generating migration scripts for live DBs, and upgrading live DBs based on previously created migration scripts. Its driven by alembic under the hood. You will use create on initial setup to generate a working MySQL database. If you change the schema in code, you can use this again with the generate option to generate a migration sript. Whenever you run an upgrade to your production instance, you should run this against your production DB with the upgrade option to bring your production DB up to sync with the code you are deploying. Run it like ./dbutils --help to see all options. The config file that this works on is the same that is given to "api", "services" and "frontend".

frontend

Development version of a frontend server allowing for account and server administration as well as score viewing and profile editing. Run it like ./frontend --help to see help output and determine how to use this. Much like "services" and "api", this should be pointed at the development version of your services config file, which holds information about the MySQL database that this should connect to as well as what game series are supported. See config/server.yaml for an example file that you can modify.

Do not use this utility to serve production traffic. Instead, see bemani/wsgi/api.wsgi for a ready-to-go WSGI file that can be used with a Python virtualenv containing this project and its dependencies, uWSGI and nginx.

ifsutils

A mediocre utility that can extract .ifs files. This has a lot of baked in assumptions and is not nearly as good as other open-source utilities for extracting files. It also cannot repack files yet. This is included for posterity, and because some bootstrapping code requires it in order to fully start a production server. Run it like ./ifsutils --help to see help output and learn how to use it.

iidxutils

A utility for patching IIDX music database files. Note that this currently can only apply a "hide leggendarias from normal folders" patch, although its probable that it can be extended for other uses.

jsx

A utility which takes the existing JSX files in the repository and builds them for you. This is offered purely as a way to serve JSX files in a production setup from nginx or similar instead of compiling them on-the-fly when they are requested. You can use this to lower cold-start load times of your frontend.

proxy

A utility to MITM an eAmuse session. Point a game at the port this listens on, and point it at another network to see the packets flowing between the two. Takes care of rewriting the facility message to MITM all messages. Has the ability to rewrite a request/response on the fly which is not currently used except for facility rewriting. Its possible that this could be used to on-the-fly patch packets coming back from a network which you don't control to do things such as enable paseli and adjust other settings that you cannot normally access. Logs in an identical format to bemanishark. Useful for black-box RE of other networks. Note that this does not have the ability to MITM SSL-encrypted traffic, so don't bother trying to use this on an official network.

This also has the ability to route a packet to one of several known networks based on the PCBID, so this can also be used as a proxy for switching networks on the fly. With a config file, this can be used as a VIP of sorts, allowing you to point all of your games at a single server that runs this proxy, and forward games on a per-PCBID basis to various networks behind the scenes. For an example config file to use "proxy" as a VIP, see config/proxy.yaml. For a more reliable proxy, use the wsgi version of this utility located at bemani/wsgi/proxy.wsgi along with uWSGI and nginx.

Run it like ./proxy --help to see how to use this utility.

psmap

A utility to take an offset from a DLL/EXE file and produce python code that would generate a suitable response that said DLL/EXE will properly parse. Essentially, if you are reversing a new game and they use the psmap function to decode all or part of a packet, you can grab either the physical offset into the DLL or the virtual address of the data and use this utility to generate the code necessary to service that request. Run it like ./psmap --help to see how to use this utility.

read

A utility to read music DB information out of game files and populate a database. This should be given the same config file as "api", "services" or "frontend" and assumes that "dbutils" has already been used to instantiate a valid MySQL DB. It also assumes you have the correct game files to read out of. Run it like ./read --help to see how to use it. This utility's uses are extensively documented below in the "Installation" section.

replay

A utility to take a packet as logged by proxy, services, trafficgen or bemanishark, and replay that packet against a particular server. Useful for quickly grabbing packets that caused a crash and debugging the crash (and verifying the fix). It is also compatible with the packet logs found on exception and unsupported packet messages in the Admin Event Logs page on the frontend. It also lets you replay that packet against your production instance once you fix the issue in case that packet was a score or profile update that you care about.

responsegen

A utility to take a packet as logged by proxy, services, trafficgen or bemanishark, and generate python code that would have generated that exact packet. Useful for quickly grabbing packets sniffed from another network and prototyping new game support. Think of this as a combination of "replay" and "psmap". This is also extremely useful when building new integration test clients. Run it like ./responsegen --help to see all information and usage.

scheduler

A command-line utility for kicking off scheduled work that must be performed against the DB. This includes picking new dailies/weeklies, new courses, and others depending on the game and any requirements that the server perform some actual calculation based on time. Essentially, any game backend that includes a run_scheduled_work override will be acted on by this utility. Note that this takes care of scheduling cadence and should be seen as a utility-specific cron handler. You can safely run this repeatedly and as frequently as desired. Run like ./scheduler --help to see how to ues this. This should be given the same config file as "api", "frontend" and "services".

services

Development version of an eAmusement protocol server using flask and the protocol libraries also used in "bemanishark" and "trafficgen". Currently it lets most modern BEMANI games boot and supports full profile and events for Beatmania IIDX 20-26, Pop'n Music 19-26, Jubeat Saucer, Saucer Fulfill, Prop, Qubell, Clan and Festo, Sound Voltex 1, 2, 3 Season 1/2 and 4, Dance Dance Revolution X2, X3, 2013, 2014 and Ace, MÚSECA 1, MÚSECA 1+1/2, MÚSECA Plus, Reflec Beat, Limelight, Colette, groovin'!! Upper, Volzza 1 and Volzza 2, Metal Gear Arcade, and finally The*BishiBashi.

Do not use this utility to serve production traffic. Instead, see bemani/wsgi/api.wsgi for a ready-to-go WSGI file that can be used with a Python virtualenv containing this project and its dependencies, uWSGI and nginx.

shell

A convenience wrapper to invoke a Python 3 shell that has paths set up to import the modules in this repository. If you want to tinker or write a quick one-off, this is probably the easiest way to do so.

struct

A convenience utility for helping reverse-engineer structures out of game DLLs/EXEs. You can give this a physical DLL offset or a virtual memory address for the start and end of the data as well as a python struct format (documentation at https://docs.python.org/3.6/library/struct.html) and this will print the decoded data to the screen one entry per line. It includes several enhancements for decoding pointers to sub-structures and pointers to C strings. Run it like ./struct --help to see how to use this.

trafficgen

A utility for simulating traffic to an eAmusement service. Given a particular game, this will run through and attempt to verify simple operation of that service. No guarantees are made on the accuracy of the emulation though I've strived to be correct. In some cases, I will verify the response, and in other cases I will simply verify that certain things exist so as not to crash a real client. This currently generates traffic emulating Beatmania IIDX 20-26, Pop'n Music 19-26, Jubeat Saucer, Fulfill, Prop, Qubell, Clan and Festo, Sound Voltex 1, 2, 3 Season 1/2 and 4, Dance Dance Revolution X2, X3, 2013, 2014 and Ace, The*BishiBashi, MÚSECA 1 and MÚSECA 1+1/2, Reflec Beat, Reflec Beat Limelight, Reflec Beat Colette, groovin'!! Upper, Volzza 1 and Volzza 2 and can verify card events and score events, as well as PASELI transactions.

verifylibs

Unit test frontend utility. This will invoke nosetests on the embarrasingly small collection of unit tests for this repository. If you are making modifications, it can be useful to write a test first (placed in the bemani/tests/ directory) and code from there. It is also useful when optimizing or profiling, and also to verify that you haven't regressed anything. Supports all options that nosetests does including filtering, verbose printing and such. Run it like ./verifylibs --help to see how to do these things. When submitting pull requests make sure to run this so you know that all tests pass.

verifylint

Lint invocation utility. This simply invokes flake8 with various options so that you can see you haven't introduced any lint errors. When submitting pull requests make sure to run this so you know you aren't introducing any lint errors into the codebase.

verifytraffic

A utility which attempts to call "trafficgen" for each supported game on the network. Think of this as a full integration test suite, as it will sweep through each supported game and verify that network services are actually working. This assumes that you are running "services". Do not point this at a production instance since it will submit bogus cards, scores, names and the like and mess up your network. This takes a config file which sets up how the client should behave. See config/trafficgen.yaml for a sample file that can be used. When submitting pull requests make sure to run this against a development version of your server so you know you haven't broken any existing game implementations.

verifytyping

Typing invocation utility. Since this repository is fully typed, this verifies that you haven't introduced any type errors and often catches bugs far faster than attemping to play a round only to see that you misused a class or misspelled a variable. When submitting pull requests make sure to run this so you know you aren't introducing any type errors into the codebase.

Installation

Dependency Setup

The code contained here assumes Python 3.6 as the base. If you don't have or don't want to install Python 3.6 as your system python, it is recommended to use virtualenv to create a virtual environment. The rest of the installation will assume you have Python 3.6 working properly (and are in an activated virtual environment if this is the route you've chosen to go). If you have a newer version of python available this code should be compatible with that as well. This code is designed to run on Linux. However, it has been tested successfully on Windows and OSX as it doesn't use any system libraries and contains pure Python implementations of all necessary pieces. YMMV in this regard, however, since the whole suite is built and tested using a Debian-based derivative and several critical pieces of code have much faster Cython implementations.

To install the required libraries, run the following command out of the root of the repository. This should allow all of the programs to at least start, but it still requires a MySQL database for many of them to be useful. This step has a dependency on an isntalled MySQL server and client as well as MySQL client development libraries. It also assumes that you've installed the 'wheel' package already. In order to compile the mysql client libraries, you will need to have libssl and libcrypto on your system as well. To satisfy these requirements on a Debian-based install, run the following command:

sudo apt install libssl-dev zlib1g-dev mysql-server mysql-client libmysqlclient-dev

Once you have all of the above present, run the following command:

pip install -r requirements.txt

Installing MySQL is outside the scope of this readme, so it is assumed that you have a MySQL database with permission to create a new DB and tables within it. Note that this software requires MySQL version 5.7 or greater. This is due to the extensive use of the "json" column type added in 5.7. Create a database (the default database with this code is 'bemani') accessed by some user and password (the default user/pass for this code is 'bemani'/'bemani'). To create all of the required tables for the installation, run the following, substituting the config file for one that you've customized if you've done so. The config file that you use here should also be used with "api", "services", and "frontend" as well as various other utilities documented above.

./dbutils --config config/server.yaml create

In order to run the frontend, Python will need to find a javascript runtime. This is so it can precompile react components at render time so there doesn't need to be a compile step when developing. I found it absolutely bonkers that the backend could be on-the-fly reloaded but I had to go through an entire build process to produce interpreted JS code, so I went the route of self-contained services instead. Installing a JS runtime is also outside the scope of this document, but a quick way to get started is to install node.js.

The default configuration points the frontend/backend cache at /tmp. It is recommended to change to a different directory, as using /tmp can cause some items not to be cached. This is due to the way /tmp on Linux restricts file access to the creator only, so if you share your cache with multiple utilities running under different users, it will fail to reuse the cache and drastically slow down the frontend.

Database Initialization

At this point, games will boot when pointed at the network, but you won't be able to save scores. This is due to the missing song/chart -> score mapping. You will find default configuration files for the traffic generator and the services backend in the config/ directory. If you've customized your database setup, you will want to update the hostname/username/password/database in the configs. You will also want to update the server address and frontend URL to customize your instance.

To create the song/chart -> score mapping, you will want to run through the following section to import data from each game series. Be sure to substitute your own services config in place of the default if you've customized it. Note that if there have been updates to the files since you initially imported, you can run with the --update flag which forces the metadata to be overwritten in the DB instead of skipped. This won't normally happen, but if you make improvements to music DB parsing, you will want to do this to update your database.

Note that you'll see a lot of re-used song entries. That will happen when the import script finds an existing set of charts for the same song in a different game version and links the two game versions together. This is how scores can be shared across different versions of the same game.

If you happen to already be an authorized client of a BEMAPI-compatible server, you can fast-track initializing your server by pointing it at the remote server and using its existing database to seed your own. If this is the case, run the following command to perform a complete initialization. If you wish to update your initial setup with newer data, perhaps because a new supported game is available, you can run the following script and append the --update flag to it. Otherwise, run the following command like so:

./bootstrap --config config/server.yaml --server http://some-server.here/ --token some-token-here

If you do not have a BEMAPI-compatible server, you can initialize the server from the game files of the games you wish to run. See the following sections for how exactly to do that.

Pop'n Music

For Pop'n Music, get the game DLL from the version of the game you want to import and run a command like so. This network supports versions 19-26 so you will want to run this command once for every version, giving the correct DLL file:

./read --config config/server.yaml --series pnm --version 22 --bin popn22.dll

Jubeat

For Jubeat, get the music XML out of the data directory of the mix you are importing, and then use "read" with --series jubeat and --version corresponding to the following table:

• Saucer: saucer
• Saucer Fulfill: saucer-fulfill
• Prop: prop
• Qubell: qubell
• Clan: clan
• Festo: festo

An example is as follows:

./read --config config/server.yaml --series jubeat --version saucer --xml music_info.xml

You will also want to populate the Jubeat name database with the following command after importing all mixes:

./read --config config/server.yaml --series jubeat --version all --tsv data/jubeat.tsv

For Jubeat Prop and later versions, you will also need to import the emblem DB, or emblems will not work properly. An example is as follows:

./read --config config/server.yaml --series jubeat --version prop \
      --xml data/emblem-info/emblem-info.xml

IIDX

For IIDX, you will need the data directory of the mix you wish to support. The import script automatically scrapes the music DB as well as the song charts to determine difficulty, notecounts and BPM. For a normal mix, you will want to run the command like so. This network supports versions 20-26 so you will want to run this command once for every version, giving the correct bin file:

./read --config config/server.yaml --series iidx --version 22 --bin \
      gamedata/data/info/music_data.bin --assets gamedata/data/sound/

Note that for omnimix mixes, you will need to point at the omnimix version of music_data.bin, normally named music_omni.bin. For the version, prepend "omni-" to the number, like so:

./read --config config/server.yaml --series iidx --version omni-22 --bin \
      gamedata/data/info/music_omni.bin --assets gamedata/data/sound/

You will also want to update the IIDX name database with the following command after importing all mixes (this fixes some inconsistencies in names):

./read --config config/server.yaml --series iidx --version all --tsv \
      data/iidx.tsv

For Qpro editing to work properly, you will also need to import the Qpro database from the mix you wish to support. This does not need to be run separately for omnimix versions, the base version Qpros will be used for both that version and the omnimix of that version. This network supports editing Qpros for versions 20-26 so you will want to run this command once for every version, giving the correcct DLL file:

./read --config config/server.yaml --series iidx --version 22 --bin bm2dx.dll

DDR

For DDR, you will need the game DLL and musicdb.xml from the game you wish to import, and then run a command similar to the following. You will want to use the version corresponding to version in the following table:

• X2: 12
• X3 vs. 2ndMix: 13
• 2013: 14
• 2014: 15
• Ace: 16

./read --config config/server.yaml --series ddr --version 15 --bin ddr.dll --xml data/musicdb.xml

For DDR Ace, there is no musicdb.xml or game DLL needed. Instead, you will need the startup.arc file, like the following example:

./read --config config/server.yaml --series ddr --version 16 --bin data/arc/startup.arc

SDVX

For SDVX, you will need the game DLL and music_db.xml from the game you wish to import, and then run the following command, modifying the version parameter as required. Note that for SDVX 1, you want the music_db.xml file in data/others/music_db/ directory, but for SDVX 2 and onward, you will want the file in data/others/ instead.

./read --config config/server.yaml --series sdvx --version 1 \
      --xml data/others/music_db.xml

For SDVX 1, you will also need to import the item DB, or appeal cards will not work properly. To do so, run the following command.

./read --config config/server.yaml --series sdvx --version 1 \
      --bin soundvoltex.dll

For SDVX 2 and 3, you will also need to import the appeal message DB, or appeal cards will not work properly. To do so, run the following command, substituting the correct version number.

./read --config config/server.yaml --series sdvx --version 2 \
      --csv data/others/appealmessage.csv

For SDVX 4, you will also need to import the appeal card DB, or appeal cards will not work properly. To do so, run the following command.

./read --config config/server.yaml --series sdvx --version 4 \
      --xml data/others/appeal_card.xml

MÚSECA

For MÚSECA, you will need the music-info.xml file from the game you wish to import. Then, run the following command, modifying the version parameter as required.

./read --config config/server.yaml --series museca --version 1 \
      --xml data/museca/xml/music-info.xml

Reflec Beat

For Reflec Beat, get the game DLL from the version of the game you want to import and run a command like so. This network supports Reflec Beat up through Volzza 2, so you will want to run this with versions 1-6 to completely initialize:

./read --config config/server.yaml --series reflec --version 1 --bin reflecbeat.dll

Running Locally

Once you've set all of this up, you can start the network in debug mode using a command similar to:

./services --port 5730 --config config/server.yaml

You can start the frontend in debug mode using another similar command as such:

./frontend --port 8573 --config config/server.yaml

You can start the BEMAPI REST server in debug mode using a command similar to:

./api --port 18573 --config config/server.yaml

The network config for any particular game should look similar to the following, with the correct hostname or IP filled in for the services URL. No path is necessary. Note that if you wish to switch between an existing network and one you serve using the "proxy" utility, you can set up the services URL to include subdirectories as required by that network. This code does not examine nor care about anything after the initial slash, so it can be whatever.

<network>
    <timeout __type="u32">30000</timeout>
    <sz_xrpc_buf __type="u32">102400</sz_xrpc_buf>
    <ssl __type="bool">0</ssl>
    <services __type="str">http://127.0.0.1:5730/</services>
</network>

If you wish to verify the network's operation with some test traffic, feel free to point the traffic generator at your development network. You should run it similar to the command below, substituting the correct port to connect to your network and choosing one of the supported games. If you don't know a supported game, you can use the --list option to print them. If "Success!" is printed after all checks, you're good to go!

./trafficgen --config config/trafficgen.yaml --port 5730 --game pnm-22 && echo Success!

You will want to set up a cron job or similar scheduling agent to call "scheduler" on a regular basis. It is recommended to call it every five minutes since there are cache warming portions for the front-end that expire every 10 minutes. Game code will register with internal handlers to perform daily/weekly actions which are kicked off by this script. An example invocation of the tool is as follows:

./scheduler --config config/server.yaml

Once your network is up and running, if you pull new code down, the DB schema may have changed. For that, use the same DB util script detailed above in the following manner. This will walk through all migration scripts that you haven't applied and bring your DB up to spec. It is recommended to create a deploy script that knows how to install dependencies and install a new version of these utilities to your production virtualenv and then runs the following script to ensure that your production DB is kept in sync with upstream changes:

./dbutils --config config/server.yaml upgrade

Since the network provided is player-first, in order to promote an account to administrator you will have to create an account on a game first. Once you have done that, you can sign up for the front-end using those credentials (your card and PIN), and then use the dbutils script to promote yourself to admin, similar to this command:

./dbutils --config config/server.yaml add-admin --username <your-name-here>

Once you have create an admin account, you can use tools on the frontend to establish arcades and their owners. Any administrator can check system settings including event logs, configure news entries, help users recover passwords and change cards and the like. Arcade owners can choose how paseli is supported on the machines in their arcade, grant users credits and configure game options such as which events are active.

Troubleshooting

If you followed the above instructions, the network should "Just Work" for you. However there are several gotchas and caveats that might not be obvious to a first-time user of this software. If you run into trouble these troubleshooting steps may help.

Logs show that games only request the initial services packet. Additional packets are not sent and games do not go online.

The initial services packet is akin to a DNS request. The response tells games where to go for each service. The values sent by the server are controlled in config/server.yaml. Make sure the domain or IP in the server.address config entry is correct for the computer you're running services on. Make sure that the IP the DNS entry resolves to, or the literal IP you've typed in this setting is routable from the game's perspective. Make sure that the port setting in server.port is the same as you've specified in your command line if you are launching the debug program, or the same as your webserver config if you are setting up a production instance. Make sure that the specified port is unblocked in any firewall running on the computer you're running services on.

Games connect to the server, logs show successful exchanges, there are no exceptions and the game boots fine but freezes on the attract screen or refuses to mark itself as "online".

Even if 100% of the network packets are responded to correctly, if the game itself can't ping the keepalive host it will refuse to enable online services. Verifiy the server.keepalive setting in config/server.yaml to make sure that it points at a computer that can be reached by the game. Make sure that that computer has enabled ICMP replies such as ping. Routers often block ping requests. It is recommended that you leave this as 127.0.0.1 as it will cause the game to ping itself and get a successful reply. This removes the usefulness of the network test screen outside of the IP setup but it is known to work.

pip install fails to compile MySQL on an ARM-based Mac

It appears that if you are installing this software on an ARM-based OSX machine and you have installed dependencies using brew, library paths are not correctly set up for MySQL to find the zstd library. As a result, pip install -r requirements.txt will fail with a cryptic error message including the line ld: library not found for -lzstd. The workaround is to specify the zstd library path manually in the pip install line. Try running the following (or a variation of the following if you've modified your pip install line already): LDFLAGS="-L$(brew --prefix zstd)/lib" pip install -r requirements.txt.

Production Setup

As alluded to several times in this README, the recommended way to run a production instance of this code is to set up uWSGI fronted by nginx. You should SSL-encrypt the frontend and the API services, and its recommended to use LetsEncrypt for a free certificate that you can manage easily. There are other ways to run this software and provide SSL credentials but I have no experience with or advice on them.

The easiest way to get up and running is to install MySQL 5.7, nginx and uWSGI along with Python 3.6 or higher. Create a directory where the services will live and place a virtualenv inside it (outside the scope of this document). Then, the wsgi files found in bemani/wsgi/ can be placed in the directory, uWSGI pointed at them and nginx set up. The setup for the top-level package will include all of the frontend templates, so you can set up a nginx directory to serve the static resources directly by pointing at the static directory inside your virtualenv.

For example configurations, an example install script, and an example script to back up your MySQL instance, see the examples/ directory.

Contributing

Contributions are welcome! Before submitting a pull request, ensure that your code is type-hint clean by running ./verifytyping and ensure that it hasn't broken basic libraries with ./verifylibs. Make sure that it is also lint-clean with ./verifylint. If you are changing code related to a particular game, it is nice to include a verification in the form of a game traffic emulator, so that basic functionality can be verified. To ensure you haven't broken another game with your changes, its recommended to run the traffic generator against your code with various games. For convenience, you can run ./verifytraffic --config config/trafficgen.yaml to run all supported games against your change. Remember that some games require you to run the scheduler to generate dailies/weeklies, and if you neglect to run this some of the integration tests will fail as they require full packet support! If possible, please also write a unit test for your changes. However, if the unit test is just a tautology and an integration/traffic test will suit better, then do that instead.

For documentation on how the protocol layer works, see "PROTOCOL". For documentation on how the eAmusement server is intended to work, see "BACKEND". Inside bemani/data/ the various DB model files have comments detailing the intended usage of each of the tables. For documentation on how the BEMAPI REST API should respond, please see the BEMAPI specification repository at https://github.com/DragonMinded/bemapi.

When updating DB schema in the various bemani/data/ python files, you will most-likely want to generate a migration for others to use. For that, we've integrated with alembic in order to provide robust migrations. The same DB utility script detailed above will create a migration script for you, given a message specifying the operation taking place. You should run this after making the code change to the schema in the relevant file under bemani/data/mysql. Alembic will automatically diff your development MySQL DB against the schema change you've made and generate an appropriate migration. Sometimes you will want to augment that migration with addtional data transformations. Various existing migrations do just that, so have a look at them under bemani/data/migrations/versions/. An example is as follows:

./dbutils --config config/server.yaml generate --message "Adding timestamp column to user."

Once the script finishes, check out the created migration script to be sure its correct and then check it in.

Development Tips

Several core components of this repo have a parallel C++ implementation for massive speed boosts. Several components are also cythonized to squeeze a bit more speed out of them as well. This project aims to provide a pure-python implementation of everything so it is not necessary to run with cythonized or compiled code either in development or production. However, if you want to benefit from the massive speed bumps provided by the equivalent implementations you can compile the code in-place for your development setup. The following will compile all needed libraries assuming you have a working C++ compiler and cython is set up to run on your computer:

python3 setup.py build_ext --inplace

If you are modifying files that have an equivalent C++ implementation and it changes their semantics, make sure to test both paths! If you are modifying code that is cythonized and you've compiled, make sure to re-run the above command or delete the compiled .so files, otherwise your changes will not show up when you test.