tminterface package

Submodules

tminterface.client module

class tminterface.client.Client

Bases: object

on_bruteforce_evaluate(iface, info: BFEvaluationInfo) → BFEvaluationResponse

Called on each bruteforce physics step iteration. This method will only be called when the bruteforce script is enabled in TMInterface. Used for implementing custom evaluation strategies. For greater control over the simulation, use the Client.on_simulation_step method instead.

Parameters

• iface (TMInterface) – the TMInterface object

• info (BFEvaluationInfo) – the info about the current bruteforce settings and race time

Returns

None if the bruteforce script should continue its builtin evaluation or a BFEvaluationResponse that signifies what the script should do.

on_checkpoint_count_changed(iface, current: int, target: int)

Called when the current checkpoint count changed (a new checkpoint has been passed by the vehicle). The current and target parameters account for the total amount of checkpoints to be collected, taking lap count into consideration.

Parameters

• iface (TMInterface) – the TMInterface object

• current (int) – the current amount of checkpoints passed

• target (int) – the total amount of checkpoints on the map (including finish)

on_client_exception(iface, exception: Exception)

Called when a client exception is thrown. This can happen if opening the shared file fails, or reading from it fails.

Parameters

• iface (TMInterface) – the TMInterface object

• exception (Exception) – the exception being thrown

on_custom_command(iface, time_from: int, time_to: int, command: str, args: list)

Called when a custom command has been executed by the user.

Parameters

• iface (TMInterface) – the TMInterface object

• time_from (int) – if provided by the user, the starting time of the command, otherwise -1

• time_to (int) – if provided by the user, the ending time of the command, otherwise -1

• command (str) – the command name being executed

• args (list) – the argument list provided by the user

on_deregistered(iface)

A callback that the client has been deregistered from a TMInterface instance. This can be emitted when the game closes, the client does not respond in the timeout window, or the user manually deregisters the client with the deregister command.

Parameters

iface (TMInterface) – the TMInterface object that has been deregistered

on_laps_count_changed(iface, current: int)

Called when the current lap count changed (a new lap has been passed).

Parameters

• iface (TMInterface) – the TMInterface object

• current (int) – the current amount of laps passed

on_registered(iface)

A callback that the client has registered to a TMInterface instance.

Parameters

iface (TMInterface) – the TMInterface object that has been registered

on_run_step(iface, _time: int)

Called on each “run” step (physics tick). This method will be called only in normal races and not when validating a replay.

Parameters

iface (TMInterface) – the TMInterface object

on_shutdown(iface)

A callback that the TMInterface server is shutting down. This is emitted when the game is closed.

Parameters

iface (TMInterface) – the TMInterface object that has been closed

on_simulation_begin(iface)

Called when a new simulation session is started (when validating a replay).

Parameters

iface (TMInterface) – the TMInterface object

on_simulation_end(iface, result: int)

Called when a new simulation session is ended (when validating a replay).

Parameters

iface (TMInterface) – the TMInterface object

on_simulation_step(iface, _time: int)

Called on each simulation step (physics tick). This method will be called only when validating a replay.

Parameters

iface (TMInterface) – the TMInterface object

tminterface.client.run_client(client: Client, server_name: str = 'TMInterface0', buffer_size=65535)

Connects to a server with the specified server name and registers the client instance. The function closes the connection on SIGBREAK and SIGINT signals and will block until the client is deregistered in any way. You can set the buffer size yourself to use for the connection, by specifying the buffer_size parameter. Using a custom size requires launching TMInterface with the /serversize command line parameter: TMInterface.exe /serversize=size.

Parameters

• client (Client) – the client instance to register

• server_name (str) – the server name to connect to, TMInterface0 by default

• buffer_size (int) – the buffer size to use, the default size is defined by tminterface.constants.DEFAULT_SERVER_SIZE

tminterface.commandlist module

class tminterface.commandlist.BaseCommand

Bases: object

The BaseCommand class is a base class for all command classes such as Command, TimedCommand and InputCommand.

to_script() → str

Converts a command to a script string line.

Returns

the script snippet that represents this command

Return type

str

class tminterface.commandlist.Command(args: list)

Bases: BaseCommand

A Command represents an immediate command that is executed immediately by TMInterface whenever it’s encountered.

args: list

to_script() → str

Converts a command to a script string line.

Returns

the script snippet that represents this command

Return type

str

class tminterface.commandlist.CommandList(obj=None)

Bases: object

A CommandList represents a list of TMInterface commands usually forming a script which can contain immediate and timed commands.

A CommandList can be loaded by providing a file handle to an existing script file or from a string. You can also construct an empty CommandList to add your own commands to then convert them into a valid TMInterface script.

If a resource is provided, the class will attempt to parse all of its contents into immediate and timed commands. You can use CommandList.to_script() to convert all the commands back into a valid TMInterface script. If any command cannot be converted, it will be commented out.

The class fully supports parsing commands with quoted arguments and inline comments and can be used to genereate new script files.

Parameters

obj –

the resource that needs to be parsed, either:

a file handle opened with open()

a string containing the command list

None to create an empty list

commands

the list containing all immediate commands

Type

list

timed_commands

the list containing all timed commands, including input commands

Type

list

content

the script string that was used to construct the CommandList

Type

str

add_command(command: BaseCommand)

Adds a command to the CommandList, converting it to an InputCommand if possible.

The command will be added to the commands list if it is of type Command. If the command is a TimedCommand, it will first be attempted to convert it to an InputCommand. If the conversion fails, it is added without any conversions. If the command is an InputCommnad, it is added to the timed_commands list.

Parameters

command (BaseCommand) – the command to be added

clear()

Clears all commands from the command list.

static parse_time(time_str: str) → int

Parses a singular timestamp which is either a number or a formatted time.

Parses a string like “947120” or “15:47.12” to an integer time in milliseconds.

Parameters

time_str (str) – the time string to parse

Returns

the time representing the string, -1 if parsing fails

Return type

int

static parse_time_range(range_str: str) → tuple

Parses a time range.

Parses a time range or a single timestamp, returning a tuple with two elements (from, to).

If the parsed time range consists only of one timestamp, to is set to -1. If from > to, the two integers are swapped.

Parameters

range_str (str) – the time range to parse

Returns

a tuple of two int’s (from, to)

Return type

tuple

sorted_timed_commands() → list

Returns all timed commands sorted in ascending order (stable).

Returns

timed commands sorted in ascending order

Return type

list

to_script() → str

Converts all immediate and timed commands to a valid TMInterface script.

Returns

the string representing the TMInterface script, one command per line

Return type

str

class tminterface.commandlist.InputCommand(timestamp: int, input_type: InputType, state: int)

Bases: BaseCommand

The InputCommand class specifically represents a command that is used to inject any kind of input into the game.

An input does not contain any arguments. Instead, the class defines an input with its type and state. InputCommand’s can be converted from an instance of TimedCommand.

InputCommand’s do not need to be stored to describe a TMInterface script, they are however automatically added by the CommandList for an easy access & manipulation of the input sequence.

input_type: InputType

state: int

timestamp: int

to_script() → str

Converts a command to a script string line.

Returns

the script snippet that represents this command

Return type

str

class tminterface.commandlist.InputType(value)

Bases: IntEnum

The InputType enum represents an input type that is coupled with its state.

DOWN = 1

GAS = 8

HORN = 6

LEFT = 2

RESET = 5

RESPAWN = 4

RIGHT = 3

STEER = 7

UNKNOWN = 9

UP = 0

static from_str(s: str)

Converts a script action string into an InputType.

If the action is invalid, InputType.UNKNOWN will be returned.

Parameters

s (str) – the string to convert

Returns

the converted input type

Return type

InputType

to_str() → str

class tminterface.commandlist.TimedCommand(args: list, timestamp: int, is_ending: bool)

Bases: Command

A TimedCommand describes any command that is executed at a specific timestamp.

The TimedCommand can represent any command, including any input commands. A command with a ranged timestamp will be always converted to two TimedCommand instances, where the earliest command will have is_ending set to False and the latest, to True.

is_ending: bool

timestamp: int

to_input_command() → InputCommand

Converts a TimedCommand to an InputCommand if possible.

If the conversion fails or the TimedCommand is not a valid input command, None is returned.

Returns

the converted InputCommand, or None if the conversion failed

Return type

InputCommand

to_script() → str

Converts a command to a script string line.

Returns

the script snippet that represents this command

Return type

str

tminterface.constants module

tminterface.eventbuffer module

class tminterface.eventbuffer.Event(*args: Any, **kwargs: Any)

Bases: ByteStruct

The Event class represents a game event (or input) with its respective time.

Such event is stored in 8 bytes internally by the game. The first 4 bytes is the time of the event. This time is a stored time, which means it is offset by 100000ms.

The last 4 bytes contain the event data. This data contains the actual event type (e.g. whether it was acceleration, braking, steering etc.) and the value of the event.

The event type is 1 byte long and it signifes an index into an array of available event types. This array is variable based on the information contained within the replay file. As such, it is required to get index of the desired event type dynamically. You can easily get/set this property through the name_index accessors.

The event value depends on the event type and is 3 bytes long. If the event type is binary (e.g. accelerate, brake, steer left), the value can be either 0 or 1. For managing this value type, use the binary_value getter/setter.

If the event type is analog, the value is stored in a custom format. You can convert between this format and the format TMInterface uses by using util.data_to_analog_value and utils.analog_value_to_data. You can avoid using these functions simply by using the analog_value getter/setter.

Parameters

time (int) – the stored time of the event

time

the stored time of the event

Type

int

data

the final data that is written into game’s memory

Type

int

property analog_value: int

property binary_value: bool

property name_index: int

class tminterface.eventbuffer.EventBufferData(events_duration: int)

Bases: object

The internal event buffer used to hold player inputs in run or simulation mode.

While simulating a race, the game loads the inputs from a replay file into an internal buffer and begins to apply “events” (inputs) from this buffer. The buffer itself consists of 8 byte values, the first 4 bytes is used for the event time and the last 4 is used for event data. See the Event class for more information.

The event time is so called a “stored” time. The stored time is defined as 100000 + race time. The stored time is saved in the replay file and is also used in the internal buffer itself.

The buffer itself is stored in decreasing order. That means that the event at index 0 in the list is the last one simulated in the race. The start and end of the race is marked by special “_FakeIsRaceRunning” and “_FakeFinishLine” events. These events mark the start and finish of the race. Note that without the presence of “_FakeIsRaceRunning” event, the race will not start at all. This event has a constant stored time of 100000.

Before the starting event, a “Respawn” event can be generated by the game, this event can also be saved in the replay file itself. The very first input that can be applied by the player happens at stored time of 100010.

Parameters

events_duration (int) – the duration of the events, equalling the finish time, mostly ignored and does not need to be set

events_duration

the duration of the events

Type

int

control_names

the list of supported event types by this buffer

Type

list

events

the list of events held by this buffer

Type

list

add(time: int, event_name: str, value: Union[int, bool])

Adds an event to the event buffer.

This is a wrapper function that provides easy API for adding new events. Depending on the event_name parameter, the method will interpret the value in different ways. If the event is an analog event, the value passed should be in the range of [-65536, 65536] where negative values represent left steering and postive, right steering.

If the event is binary, the value should be False for disabling the input and True for enabling it.

The time parameter is zero based, where 0 is the first human input that can be injected. Internally, 0 translates to stored time 100010, which is the first simulation step after the countdown.

Parameters

• time (int) – zero based timestamp when the input is injected

• event_name (str) – the event name that specifies the input type

• value (Union[int, bool]) – the value for the event, based on the event type

clear()

Removes all events in the current event buffer, leaving the race running event in the buffer.

A race running event should always be present in the buffer, to make the game start the race.

copy()

Copies the event buffer with all its events.

Returns

a deep copy of the original event buffer

find(**kwargs)

Finds matching events according to keyword arguments.

Any unspecified parameter will be skipped in the search and will not be compared. You may use this method to filter events based on time, event type and value.

Find all analog steering events with value -65536:

matching = event_buffer.find(event_name=structs.ANALOG_STEER_NAME, value=-65536)

Find all events that happened at input time 0:

matching = event_buffer.find(time=0)

Find the finish line event:

matching = event_buffer.find(event_name=structs.BINARY_RACE_FINISH_NAME, value=True)

Calling this method without any keyword arguments will return all events in the buffer.

Parameters

**kwargs – the keyword arguments

Keyword Arguments

• event_name (str) – match events with this event type

• time (int) – match events with this time (zero based)

• value (Union[int, bool]) – match events with this value, bool if the event type is binary, int if analog, this parameter can only be filtered if event_name is provided

Returns

the events that matched the query

Return type

list

sort()

Sorts the event buffer by time in decreasing order.

This is the order that events are stored internally by the game. Calling this is not needed, if you are calling set_event_buffer. The server will always take care of properly sorting the events.

to_commands_str(all_events=False)

Converts event buffer events and constructs a string consisting of commands compatible with TMInterface’s script syntax.

By default, only events that happened in the race (signified by the race running and finish events) will be converted and appended to the final string. If you want to convert all commands that are available in the buffer, call this method with all_events set to True.

Parameters

all_events (bool) – whether to convert all commands available in the buffer

Returns

the string containing commands compatible with TMInterface’s script syntax

Return type

str

tminterface.interface module

class tminterface.interface.Message(_type: int, error_code=0)

Bases: object

The Message class represents a binary buffer that contains useful methods to construct a message to send to the server. A message additionally contains its type, whether it is a response to a server call, or a normal client call. It also contains an error code, if there was any failure writing the message.

Parameters

• _type (int) – the message type

• error_code (int) – the error code of the message, 0 if none

_type

the message type

Type

int

error_code

the error code of the message, 0 if none

Type

int

data

the binary data

Type

bytearray

to_data() → bytearray

write_buffer(buffer: bytearray)

write_double(n: float)

write_int(n, size)

write_int16(n: int)

write_int32(n: int)

write_uint16(n: int)

write_uint32(n: int)

write_uint8(n)

write_zeros(n_bytes)

class tminterface.interface.MessageType(value)

Bases: IntEnum

An enumeration.

ANY = 35

C_DEREGISTER = 13

C_EXECUTE_COMMAND = 28

C_GET_CHECKPOINT_STATE = 25

C_GET_CONTEXT_MODE = 22

C_GIVE_UP = 17

C_HORN = 18

C_LOG = 34

C_PREVENT_SIMULATION_FINISH = 32

C_PROCESSED_CALL = 14

C_REGISTER = 12

C_REGISTER_CUSTOM_COMMAND = 33

C_REMOVE_STATE_VALIDATION = 31

C_RESPAWN = 16

C_SET_CHECKPOINT_STATE = 26

C_SET_EXECUTE_COMMANDS = 29

C_SET_GAME_SPEED = 27

C_SET_INPUT_STATES = 15

C_SET_TIMEOUT = 30

C_SIM_GET_EVENT_BUFFER = 21

C_SIM_GET_STATE = 20

C_SIM_REWIND_TO_STATE = 19

C_SIM_SET_EVENT_BUFFER = 23

C_SIM_SET_TIME_LIMIT = 24

S_ON_BRUTEFORCE_EVALUATE = 11

S_ON_CHECKPOINT_COUNT_CHANGED = 8

S_ON_CUSTOM_COMMAND = 10

S_ON_LAPS_COUNT_CHANGED = 9

S_ON_REGISTERED = 2

S_ON_RUN_STEP = 4

S_ON_SIM_BEGIN = 5

S_ON_SIM_END = 7

S_ON_SIM_STEP = 6

S_RESPONSE = 1

S_SHUTDOWN = 3

exception tminterface.interface.ServerException

Bases: Exception

An exception thrown when the server cannot perform requested operation.

class tminterface.interface.TMInterface(server_name='TMInterface0', buffer_size=65535)

Bases: object

TMInterface is the main class to communicate with the TMInterface server. The communication is done through memory mapping and a simple synchronous message system between the server and the client. A TMInterface server can only serve only one client at a time, it is however possible to connect to many servers from the same script.

The interface provides various functions to manipulate game state and hook to different periods of game execution.

Parameters

• server_name (str) – the server tag to connect to

• buffer_size (int) – the buffer size used by the server, the default is specified by tminterface.constants.DEFAULT_BUFFER_SIZE. Using a custom size requires launching TMInterface with the /serversize command line parameter: TMInterface.exe /serversize=size.

server_name

the server tag that’s used

Type

str

running

whether the client is running or not

Type

bool

registered

whether the client is registered

Type

bool

mfile

the internal mapped file used for communication

Type

mmap.mmap

buffer_size

the buffer size used for communication

Type

int

client

the registered client that’s controlling the server

Type

Client

clear_event_buffer()

Clears the current event buffer used for simulation, leaving the race running event in the buffer.

A race running event should always be present in the buffer, to make the game start the race.

close()

Closes the connection to the server by deregistering the current client and shutting down the thread for communication.

This method will send a message to the server to deregister the current client.

After a successful deregistration, Client.on_deregistered() will be called with the instance of the TMInterface class.

execute_command(command: str)

Adds an interface command to the internal command queue.

The command will not be immediately executed, rather, it may be executed when the current queue is processed on the next game frame.

Parameters

command (str) – the command to execute

get_checkpoint_state() → CheckpointData

Gets the current checkpoint state of the race.

See CheckpointData for more information.

Returns

the object that holds the two arrays representing checkpoint state

Return type

CheckpointData

get_context_mode() → int

Gets the context mode the TMInterface instance is currently in.

The context mode is determining if the current race is in “run” mode, that is a normal race or “simulation” mode, which is when a player validates a replay.

Returns

MODE_SIMULATION (0) if the player is in the simulation mode, MODE_RUN (1) if in a normal race

Return type

int

get_event_buffer() → EventBufferData

Gets the internal event buffer used to hold player inputs in run or simulation mode. If the server is in the run mode (that is, in a normal race controlled by the player), this method returns the inputs of the current race. Note that new inputs will be added to the buffer as the player or TMInterface injects inputs into the game.

See EventBufferData for more information.

Returns

the event buffer holding all the inputs of the current simulation

Return type

EventBufferData

get_simulation_state() → SimStateData

Gets the current simulation state of the race.

The method can be called in on_run_step or on_simulation_step calls. See SimStateData for more information.

Returns

the object holding the simulation state

Return type

SimStateData

give_up()

Restarts the current race.

This function does not do anything in a simulation context. To rewind to the start of the race in the simulation context, use simulation states.

horn(sim_clear_events: bool = True)

Queues a deterministic horn at next race tick. This function will not immediately call the game to horn, as TMInterface has to call the specific function at a specific place in the game loop.

In a simulation context, the server will add a new input event to the existing event buffer such that that the car horns at the next tick. By default, all other input events are cleared. If you want to preserve existing input events, pass sim_clear_events=False.

Parameters

sim_clear_events (bool) – whether to clear all other events in simulation mode

log(message: str, severity='log')

Prints a message in TMInterface’s console.

You can specify the severity of the command to highlight the line in a different color.

Parameters

• message (str) – the message to print

• severity (str) – one of: “log”, “success”, “warning”, “error”, the message severity

prevent_simulation_finish()

Prevents the game from stopping the simulation after a finished race.

Calling this method in the on_checkpoint_count_changed will invalidate checkpoint state so that the game does not stop simulating the race. Internally this is simply setting the last checkpoint time to -1 and can be also done manually in the client if additional handling is required.

register(client: Client) → bool

Registers a client on the server. The server can only register one client at a time, if the client is already registered, the method will return False.

This method will initially start a new thread and send a message to the server to register a new client. After a successful registration, Client.on_registered() will be called with the instance of the TMInterface class.

Parameters

client (Client) – a Client instance to register

Returns

True if registration was scheduled, False if client is already registered

register_custom_command(command: str)

Registers a custom command within the console.

This function allows you to implement a custom command that is registered within TMInterface’s console. When executing the command, the Client.on_custom_command() method will be called with additional arguments such as the time range and processed arguments list.

It is completely up to the command implementation to process the time range and additional arguments supplied in the on_custom_command hook. Quoted arguments such as filenames will be automatically joined into one argument, even if they contain spaces.

A console command is not immediately executed after submitting it to the console. TMInterface executes commands asynchronously, processing a fixed amount of commands each frame. This is done to prevent the game from hanging when loading scripts with 1000’s of commands.

Use the log() method to output any info about the execution of your command.

Parameters

command (str) – the command to register, the command cannot contain spaces

remove_state_validation()

Makes the game validate the replay without checking if the inputs match the states saved in the replay, as if it was validating a replay exported for validation.

Calling this method in the on_simulation_begin call will remove state validation from currently validated replay. After calling, TrackMania will not check if the simulation matches with saved states in the replay, therefore allowing for input modification without stopping the simulation prematurely.

respawn(sim_clear_events: bool = True)

Queues a deterministic respawn at the next race tick. This function will not immediately call the game to respawn the car, as TMInterface has to call the specific function at a specific place in the game loop.

In a simulation context, the server will add a new input event to the existing event buffer such that that the car respawns at the next tick. By default, all other input events are cleared. If you want to preserve existing input events, pass sim_clear_events=False.

The function will respawn the car to the nearest respawnable checkpoint or if there was no passed checkpoints, restart the race. The behaviour of this function also depends on the start_respawn console variable set within TMInterface. If start_respawn is set to true, respawning without any passed checkpoints will not restart the race, but only respawn the car on the start block, simulating online respawn behaviour.

Parameters

sim_clear_events (bool) – whether to clear all other events in simulation mode

rewind_to_state(state: SimStateData)

Rewinds to the provided simulation state.

The method of restoring the simulation state slightly varies depending on the context_mode field of the SimStateData class. Some buffers may not be restored at all but are replaced with native game function calls.

The simulation state is obtainable through the get_simulation_state method. This state can also be used to write a save state compatible file for TMInterface. Note that modifying important parts of the state invalidates the current race.

To provide state restoration across game instances, TMInterface uses memory masks to omit restoring instance specific fields such as pointers or arrays.

The method can be called in on_run_step or on_simulation_step calls. Note that rewinding to a state in any of these hooks will immediately simulate the next step after the hook. For example, rewinding to a state saved at race time 0, will result in the next call to on_run_step/on_simulation_step being at time 10. If you want to apply any immediate input state, make sure to apply it in the same physics step as the call to rewind_to_state.

Parameters

state (SimStateData) – the state to restore, obtained through get_simulation_state

set_checkpoint_state(data: CheckpointData)

Sets the checkpoint state of the game. See get_checkpoint_state to learn more about how the game stores checkpoint information.

Parameters

data (CheckpointData) – the checkpoint data

set_event_buffer(data: EventBufferData)

Replaces the internal event buffer used for simulation with a new one.

If you do not modify existing inputs or do not generate all events beforehand, use TMInterface.set_input_state for dynamic input injection. See EventBufferData for more information.

The events_duration and control_names fields are ignored in this call.

Parameters

data (EventBufferData) – the new event buffer

set_input_state(sim_clear_buffer: bool = True, **kwargs)

Sets the game input state of the vehicle.

Sets individual input states for the car. If successfully applied, key states are guaranteed to be applied at next physics tick. If you want to apply an input state that happens at 500ms, call send this message at 490ms (one step before).

Note that it is not guaranteed that the game will actually process the input in the RUN mode. This can happen when setting the game speed to high factors (such as >100). This does not affect the simulation context.

In a simulation context, the server will add new input events to the existing event buffer such that that the next tick has the desired input state. By default, all other input events are cleared. If you want to preserve existing input state & events, pass sim_clear_buffer=False.

Arguments left, right, accelerate and brake are binary events. To disable an action pass False and to enable it, pass True.

Arguments steer and gas are analog events. Pass a value in the range of [-65536, 65536] to modify the state of these actions. You can also use the extended steer range of [-6553600, 6553600], note however that this range is not possible to achieve on physical hardware. This call is not affected by the extended_steer console variable.

Parameters

• sim_clear_buffer (bool) – whether to clear the event buffer when setting input state in simulation

• **kwargs – the keyword arguments

Keyword Arguments

• left (bool) – the left binary input, False = disabled, True = enabled

• right (bool) – the right binary input, False = disabled, True = enabled

• accelerate (bool) – the up binary input, False = disabled, True = enabled

• brake (bool) – the down binary input, False = disabled, True = enabled

• steer (int) – the steer analog input, in range of [-65536, 65536]

• gas (int) – the gas analog input, in range of [-65536, 65536]

set_simulation_time_limit(time: int)

Sets the time limit of the simulation.

This allows for setting an arbitrary time limit for the running simulation, making the game stop the simulation after the provided time limit is exhausted.

By default, this limit is set to the finish time of the original replay (taken from events duration found in the events buffer).

Note that setting the limit to a large value will extend the simulation to that limit, even after the race is finished. Make sure to manage finishing the race according to your application (e.g by rewinding to a state).

To reset the time to the original limit, pass -1. This call applies only to the simulation context.

Parameters

time (int) – the time at which the game stops simulating, pass -1 to reset to the original value

set_speed(speed: float)

Sets the global game speed, internally this simply sets the console variable “speed” in an TMInterface instance.

All characteristics of setting the global speed apply. It is not recommended to set the speed to high factors (such as >100), which could cause the game to skip running some subsystems such as the input subsystem.

This variable does not affect simulation contexts in which debug mode is disabled. When debug mode is disabled (default), the game runs only the simulation subsystem.

Parameters

speed (float) – the speed to set, 1 is the default normal game speed, factors <1 will slow down the game while factors >1 will speed it up

set_timeout(timeout_ms: int)

Sets the timeout window in which the client has to respond to server calls.

The timeout specifies how long will the server wait for a response from the client. If the response does not arrive in this time frame, the server will automatically deregister the client itself.

The timeout is specified in milliseconds, by default this is 2000ms (2 seconds). Set the timeout to -1 to have the server wait forever for the response.

Parameters

timeout_ms (int) – the timeout in milliseconds

tminterface.structs module

class tminterface.structs.BFEvaluationDecision(value)

Bases: IntEnum

The decision taken by the client in every bruteforce physics step. Returned in Client.on_bruteforce_evaluate() in an BFEvaluationResponse instance.

CONTINUE: run the default evaluation of the bruteforce script

DO_NOTHING: do not run any evaluation that could result in accepting or rejecting the evaluated solution

ACCEPT: accept the current solution as the new best one. Starts a new intial phase in the next physics step.

REJECT: rejects the current solution and generates a new one for next evaluation

STOP: stops the bruteforce script and lets the game simulate the race until the end

ACCEPT = 2

CONTINUE = 0

DO_NOTHING = 1

REJECT = 3

STOP = 4

class tminterface.structs.BFEvaluationInfo(*args: Any, **kwargs: Any)

Bases: ByteStruct

The bruteforce settings applied in the bruteforce process, including the current simulation race time.

class tminterface.structs.BFEvaluationResponse(*args: Any, **kwargs: Any)

Bases: ByteStruct

The response object sent by Client.on_bruteforce_evaluate().

If decision is set to BFEvaluationDecision.REJECT, you are allowed to change the inputs manually via the TMInterface.set_event_buffer() method and set the rewind_time to timestamp - 10 where timestamp is the first input that has been changed by your algorithm. Otherwise, TMInterface will automatically randomize the inputs according to the current settings itself.

class tminterface.structs.BFPhase(value)

Bases: IntEnum

The phase in which the bruteforce script is currently working.

The initial phase is executed at the beginning of the process and after each improvement. It is used primarily for collecting data about the race e.g: the race time, position of the car, checkpoint times etc. No state modification happens at this state and it is recommended to use this phase to collect information about the current solution.

The search phase is when TMInterface is searching for a new improvement. In this phase, the process changes inputs according to the user settings and evaluates the solution based on the current target.

INITIAL = 0

SEARCH = 1

class tminterface.structs.BFTarget(value)

Bases: IntEnum

The bruteforce metric that is being currently optimized.

CHECKPOINT_TIME = 1

DISTANCE_SPEED = 3

FINISH_TIME = 0

TRIGGER = 2

class tminterface.structs.CachedInput(*args: Any, **kwargs: Any)

Bases: ByteStruct

time

int

event

Event

event

alias of Event

class tminterface.structs.CheckpointData(*args: Any, **kwargs: Any)

Bases: ByteStruct

The CheckpointData object represents checkpoint state within the game.

The game keeps track of two arrays that contain checkpoint information. The first “state” array is an array of booleans (a boolean is 4 bytes long) and keeps track of which checkpoints were already passed. The length of the array represents the real count of the checkpoints on current the map (including finish). This does not mean that to finish the race the player has to pass through this exact count of checkpoints. A map with 3 laps and 5 checkpoints will still contain only 5 checkpoint states.

The second “times” array is an array of structures with 2 fields: time and an unknown field. This array holds the “logical” number of checkpoints that have to be passed (including finish). This means the total number of checkpoint passes, including the existence of laps.

Parameters

• cp_states (list) – the checkpoint states array

• cp_times (list) – the checkpoint times array, each element is a two element tuple of (time, flags)

read_from_file(file)

class tminterface.structs.CheckpointTime(*args: Any, **kwargs: Any)

Bases: ByteStruct

time

int

stunts_score

int

class tminterface.structs.ClassicString(*args: Any, **kwargs: Any)

Bases: ByteStruct

A string sent by the client to TMInterface.

command = None

class tminterface.structs.Engine(*args: Any, **kwargs: Any)

Bases: ByteStruct

max_rpm

float

braking_factor

float

clamped_rpm

float

actual_rpm

float

slide_factor

float

rear_gear

int

gear

int

class tminterface.structs.HmsDynaStateStruct(*args: Any, **kwargs: Any)

Bases: ByteStruct

quat

np.ndarray

rotation

np.ndarray

position

np.ndarray

linear_speed

np.ndarray

add_linear_speed

np.ndarray

angular_speed

np.ndarray

force

np.ndarray

torque

np.ndarray

inverse_inertia_tensor

np.ndarray

unknown

float

not_tweaked_linear_speed

np.ndarray

owner

int

property inverse_intertia_tensor

class tminterface.structs.HmsDynaStruct(*args: Any, **kwargs: Any)

Bases: ByteStruct

previous_state

HmsDynaStateStruct

current_state

HmsDynaStateStruct

temp_state

HmsDynaStateStruct

rest

bytearray

property prev_state

rest = 616

class tminterface.structs.PlayerInfoStruct(*args: Any, **kwargs: Any)

Bases: ByteStruct

team

int

prev_race_time

int

race_time

int

race_time

int

race_best_time

int

lap_start_time

int

lap_time

int

lap_best_time

int

min_respawns

int

nb_completed

int

max_completed

int

stunts_score

int

best_stunts_score

int

cur_checkpoint

int

average_rank

float

current_race_rank

int

current_round_rank

int

current_time

int

race_state

int

ready_enum

int

round_num

int

offset_current_cp

float

cur_lap_cp_count

int

cur_cp_count

int

cur_lap

int

race_finished

bool

display_speed

int

finish_not_passed

bool

countdown_time

int

rest

bytearray

rest = 32

class tminterface.structs.RealTimeState(*args: Any, **kwargs: Any)

Bases: ByteStruct

damper_absorb

float

field_4

float

field_8

float

field_12

np.ndarray

field_48

np.ndarray

field_84

np.ndarray

field_108

float

has_ground_contact

bool

contact_material_id

int

is_sliding

bool

relative_rotz_axis

np.ndarray

nb_ground_contacts

int

field_144

np.ndarray

rest

bytearray

rest = 12

class tminterface.structs.SceneVehicleCar(*args: Any, **kwargs: Any)

Bases: ByteStruct

is_update_async

bool

input_gas

float

input_brake

float

input_steer

float

is_light_trials_set

bool

horn_limit

int

quality

int

max_linear_speed

float

gearbox_state

int

block_flags

int

prev_sync_vehicle_state

SceneVehicleCarState

sync_vehicle_state

SceneVehicleCarState

async_vehicle_state

SceneVehicleCarState

prev_async_vehicle_state

SceneVehicleCarState

engine

Engine

has_any_lateral_contact

bool

last_has_any_lateral_contact_time

int

water_forces_applied

bool

turning_rate

float

turbo_boost_factor

float

last_turbo_type_change_time

int

last_turbo_time

int

turbo_type

int

roulette_value

int

is_freewheeling

bool

is_sliding

bool

wheel_contact_absorb_counter

int

burnout_state

int

current_local_speed

np.ndarray

total_central_force_added

np.ndarray

is_rubber_ball

bool

saved_state

np.ndarray

async_vehicle_state

alias of SceneVehicleCarState

engine

alias of Engine

prev_async_vehicle_state

alias of SceneVehicleCarState

prev_sync_vehicle_state

alias of SceneVehicleCarState

sync_vehicle_state

alias of SceneVehicleCarState

class tminterface.structs.SceneVehicleCarState(*args: Any, **kwargs: Any)

Bases: ByteStruct

speed_forward

float

speed_sideward

float

input_steer

float

input_gas

float

input_brake

float

is_turbo

bool

rpm

float

gearbox_state

int

rest

bytearray

rest = 28

class tminterface.structs.SimStateData(*args: Any, **kwargs: Any)

Bases: ByteStruct

The SimStateData object represents a full save state of the simulation state, including checkpoint and input information.

The simulation state consists of raw memory buffers representing various information about the race state. This includes the entirety of the vehicle state as well as the player info and other state variables such as current checkpoint count and input state.

The memory regions themselves are monitored by TMInterface itself and are used for functionality like save states or fast rewind in the bruteforce script. TMInterface may use additional native game methods to restore the state based on information present in some of these memory regions. It is important to note that the buffers contain instance specific fields such as pointers and array sizes. These are masked out automatically by TMInterface when restoring the state (and when calling TMInterface.rewind_to_state).

To query input state of the simulation state regardless of context, use input_* (input_accelerate, input_brake etc.) accessors.

version

int

context_mode

int

flags

int

timers

np.ndarray

dyna

HmsDynaStruct

scene_mobil

SceneVehicleCar

simulation_wheels

np.ndarray

plug_solid

bytes

cmd_buffer_core

bytes

player_info

PlayerInfoStruct

internal_input_state

np.ndarray

input_running_event

Event

input_finish_event

Event

input_accelerate_event

Event

input_brake_event

Event

input_left_event

Event

input_right_event

Event

input_steer_event

Event

input_gas_event

Event

cmd_buffer_core = 264

cp_data

alias of CheckpointData

property display_speed: int

dyna

alias of HmsDynaStruct

property input_accelerate: bool

input_accelerate_event

alias of Event

property input_brake: bool

input_brake_event

alias of Event

input_finish_event

alias of Event

property input_gas: int

input_gas_event

alias of Event

property input_left: bool

input_left_event

alias of Event

property input_right: bool

input_right_event

alias of Event

input_running_event

alias of Event

property input_steer: int

input_steer_event

alias of Event

player_info

alias of PlayerInfoStruct

plug_solid = 68

property position: list

property race_time: int

property rewind_time: int

property rotation_matrix: list

scene_mobil

alias of SceneVehicleCar

property time: int

property velocity: list

property yaw_pitch_roll: numpy.array

class tminterface.structs.SimulationWheel(*args: Any, **kwargs: Any)

Bases: ByteStruct

steerable

bool

field_8

int

surface_handler

SurfaceHandler

field_112

np.ndarray

field_160

int

field_164

int

offset_from_vehicle

np.ndarray

real_time_state

RealTimeState

field_348

int

contact_relative_local_distance

np.ndarray

prev_sync_wheel_state

WheelState

sync_wheel_state

WheelState

field_564

WheelState

async_wheel_state

WheelState

async_wheel_state

alias of WheelState

field_564

alias of WheelState

prev_sync_wheel_state

alias of WheelState

real_time_state

alias of RealTimeState

surface_handler

alias of SurfaceHandler

sync_wheel_state

alias of WheelState

class tminterface.structs.SurfaceHandler(*args: Any, **kwargs: Any)

Bases: ByteStruct

unknown

np.ndarray

rotation

np.ndarray

position

np.ndarray

class tminterface.structs.WheelState(*args: Any, **kwargs: Any)

Bases: ByteStruct

rest

bytearray

rest = 100

tminterface.util module

tminterface.util.analog_value_to_data(value: int) → int

Converts a value in [-65536, 65536] range to an internal analog state value.

The function supports values outside the normal range, that is you can convert values in the extended range as well.

Parameters

data (int) – the value to convert

Returns

the converted value

Return type

int

tminterface.util.data_to_analog_value(data: int) → int

Converts an internal analog state value to a [-65536, 65536] range.

The function supports values outside the normal range, that is you can convert values in the extended range as well.

Parameters

data (int) – the internal value, usually stored in an event buffer

Returns

the converted value

Return type

int

tminterface.util.mat3_to_quat(mat: numpy.array) → numpy.array

Converts a rotation matrix to a quaternion.

This function uses the internal implementation that the game uses to convert a rotation matrix to a quaternion.

The function itself is constructed from reverse engineering the code of the game. This particular implementation should be 100% compatible with the original method the game uses. Note however that this compatibility is not guaranteed.

Parameters

mat (np.array) – a 3x3 rotation matrix to convert

Returns

the quaternion consisting of 4 elements (x, y, z, w)

Return type

np.array

tminterface.util.quat_to_ypw(quat: numpy.array) → numpy.array

Converts a quaternion to yaw, pitch and roll values.

This function uses the internal implementation that the game uses to convert quaternions to yaw, pitch and roll.

The function itself is constructed from reverse engineering the code of the game. Note that this particular implementation is not 100% compatible with the actual assembly code, however it should produce the same results in most cases.

Parameters

quat (np.array) – the quaternion to convert (x, y, z, w)

Returns

an array containing 3 elements: yaw, pitch and roll

Return type

np.array

Module contents