Current timestamp read code uses FPGA register reads. Through testing,
this read was slower then clock_gettime by about 4-5x. However, another
method of reading the FPGA time is available, using HMB. HMB
is memory mapped IO from RAM to the FPGA. So to code side,
reading the value is just a memory barrier and a memory read.
There is some latency on the write side, so a very small artifical delay
(5us) is added to avoid register reads such as interrupts being ahead
of current timestamps, which could cause issues.
Below is read times for 1000 calls to clock_gettime, register reads and
hmb reads.
```
Clock: Rise 1.72939400 s Fall 1.72990700 s Delta 0.00051300 s
FPGA : Rise 1.72999000 s Fall 1.73429300 s Delta 0.00430300 s
HMB : Rise 1.73466800 s Fall 1.73481900 s Delta 0.00015100 s
```
Also add full HMB struct to HAL for future usage.
15 m/s is about 50 ft/s, which is way above what FRC robots should be
able to achieve. This limit lets us catch user errors from bad unit
conversions immediately instead of the LUT generation in the LTV
controllers hanging for a really long time.
Fixes#5027.
This works around an exit race with wpi::Now() on Rio; it was overridden
to call HAL_GetFPGATime(), which calls chipobject, but on exit, because
there was not a library dependency, the chipobject could be destroyed
prior to wpiutil/wpinet being shut down.
# Background
Unit safety has always been a problem in WPILib. Any value corresponding to a physical measurement, such as current draw or distance traveled, is represented by a bare number with no unit tied to it; it's up to the programmer to know what units they're working and take care to remember that while working on their robot program. This leads to bugs when programmers accidentally mix units without knowing, or measure something (such as a wheel diameter) in one unit and program using another. `wpiunits` is intended to eliminate that class of bugs.
Another source of friction is the controllers and models in `wpimath` that expect all inputs to be in terms of SI units (meter, kilogram, and so on), while most FRC teams are US-based and most commonly use imperial units. wpimath does a good job of noting unit types in method names and argument names; however, it still relies on users properly converting values (and knowing they even have to do so).
# API
There are really only two core classes in this library: `Unit` and `Measure`. A `Unit` represents some dimension like distance or time. `Unit` is subclassed to define specific dimensions (eg `Distance` and `Time`) and those subclasses are instantiated to defined particular units in those dimensions, such as `Meters` and `Feet` being instances of the `Distance` class.
A `Measure` is a value tied to a particular dimension like distance and knows what unit that value is tied to. `Measure` has two implementations - one immutable and one mutable. The `Measure` interface only defines *read-only* operations; any API working with measurements should use the interface. The default implementation is `ImmutableMeasure`, which only implements those read-only operations and is useful for tracking constants. `MutableMeasure` also adds some methods that will allow for mutation of its internal state; this class is intended for use for things like sensors and controllers that track internal state and don't want to allocate new `Measure` objects every time something like `myEncoder.getDistance()` is called. However, the APIs for those methods should still only expose the read-only `Measure` interface so users can't (without casting or reflection) change the internal values.
A `Units` class provides convenient definitions for most of the commonly used unit types, such as `Meters`, `Feet`, and `Milliseconds`. I recommend static importing these units eg `import static edu.wpi.first.units.Units.Meters`) so they can be used like `Meters.of(1.234)` instead of `Units.Meters.of(1.234)`
# Examples
These examples are admittedly contrived. Users shouldn't be interacting much with measure objects themselves, since wpimath and wpilibj classes will be updated to support working with them; users will often just have to take a `Measure` output from one place (such as an encoder) and feed it as input to something else (such as a PID controller or kinematics model)
```java
// Using raw units
Encoder encoder = ...
int kPulsesPerRev = 2048;
double kWheelDiameterMeters = Units.inchesToMeters(6);
double kGearRatio = 10.86;
// always have to remember this encoder will output in meters!
encoder.setDistancePerPulse(kWheelDiameterMeters * Math.PI / (kGearRatio * kPulsesPerRev));
Command driveDistance(double distance) {
// have to know the distance argument needs to be in meters!
return run(this::driveStraight).until(() -> encoder.getDistance() >= distance);
}
// Oops! This will go 16 feet, not 5!
Command driveFiveFeet = driveDistance(5);
Command driveOneMeter = driveDistance(1);
```
```java
// Using wpiunits
Encoder encoder = ...
int kPulsesPerRev = 2048;
Measure<Distance> kWheelDiameter = Inches.of(6);
double kGearRatio = 10.86;
encoder.setDistancePerPulse(kWheelDiameter.times(Math.PI).divide(kGearRatio * kPulsesPerRev));
Command driveDistance(Measure<Distance> distance) {
// Measure#gte automatically handles unit conversions
return run(this::driveStraight).until(() -> encoder.getDistance().gte(distance));
}
// Users HAVE to be explicit about their units
Command driveFiveFeet = driveDistance(Feet.of(5));
Command driveOneMeter = driveDistance(Meters.of(1));
```
```java
SmartDashboard.putNumber("Temperature (C)", pdp.getTemperature().in(Celsius));
SmartDashboard.putNumber("Temperature (F)", pdp.getTemperature().in(Fahrenheit));
```
```java
var InchSecond = Inch.mult(Second); // new combined unit types can be user-defined
var InchPerSecond = Inch.per(Second);
PIDController<Distance, ElectricPotential> heightController = new PIDController<>(
/* kP */ Volts.of(0.2).per(Inch),
/* kI */ Volts.of(0.002).per(InchSecond),
/* kD */ Volts.of(0.008).per(InchPerSecond)
);
var elevatorTop = Feet.of(4).plus(Inches.of(6.125));
elevatorMotor.setVoltage(heightController.calculate(encoder.getDistance(), elevatorTop));
```
Many teams have issues trying to read the DS too early. By switching to an optional, we cause teams to check 2 things. Either 1) they explicitly check, and their code is correct, or 2) they just read .value() and their code reboots in a loop. However, because the DS will eventually connect, this 2nd case is ok, and should theoretically be undetectable on the field.
The problem is that you have to use a different pkg-config file if you want to use a static variant of libuv, but the buildsystem should not care which variant of libuv should be used. This is not a problem with the cmake config.
Accelerometer is hyper-specific to ADXL accelerometers, and Gyro is
less useful now that 3D IMUs are prevalent, and if those IMUs want to
support the Gyro interface, they also need to provide a way to set the
axis used for the Gyro interface, which is confusing. Higher-order
functions (e.g., lambdas) are a more flexible interface boundary than
interfaces, but they didn't exist when these interfaces were
created.
