liftright-data-server

Data collection server for gamified LiftRight study

Using this data

Data Model

There are 2 high level record schema, Repetition and ImuData The data is stored using the Bucket pattern

ObjectId

ObjectId is an internal Mongo data type. When represented as JSON it is an object with a single $oid member. For consistency a Rust style definition is provided:

pub struct ObjectId {
    pub $oid: String //a u64 in base 16
}

Session

A session represents a single continuous use of the application, from beginning workout to ending. The session schema has 2 fields

pub struct Session {
    pub device_id: Uuid,
    pub session_id: Uuid,
}

Buckets at a high level

Buckets provide a way to maximize write performance. LiftRight writes a large amount of data into MongoDb. Buckets are defined generically in query_selector.rs. They allow an arbitrary structure to be serialized and updated using the bucket pattern.

ImuData

There are a series of schemas for ImuData, they are nested. The schemas for these records are defined in imurecords.rs.

ImuRecordPair

An ImuRecordPair is a pair of ImuRecords, an accelerometer record and a gyroscope record.

pub struct ImuRecordPair {
    pub acc: ImuRecord,
    pub gyro: ImuRecord,
}

ImuRecord

Represents an individual IMU datapoint from either the accelerometer and the gyroscope.

The time member of this data does not represent any meaningful point in absolute time. TThe values are relative to the device generating them. They have nanosecond precision and can be used for ordering and computing duration only.

pub struct ImuRecord {
    pub x: f64,
    pub y: f64,
    pub z: f64,
    pub time: i64,
}

The ImuData bucket

This struct is not strictly defined in the code. It is built dynamically by the to_bucket_update method in query_selector.rs.

Unlike the time member of the ImuRecord struct, the timestamp appended onto the id member has a Wall time representable value. The value is in nanoseconds, but it can be read into most languages for analytics by convertng to milliseconds (see your preferred time library for API documentation).

pub struct ImuRecordBucket {
    pub _id: ObjectId
    pub device_id: Uuid,
    pub kind: "ImuData",
    pub session_id: Uuid,
    pub id: [device_id]_[timestamp of last insert(ns)],
    pub imu_data: Vec<ImuRecordPair>
    pub imu_record_count: u64
}

The data stored in the imu_data member is stored in order, oldest first.

Repetitions and Sets

Sets are stored as discreet objects. Bucket size is much smaller for Repetition data. the time member of Repetition is also nanosecond precision and not wall-time meaningful. It can be used for determining intervals and ordering. If a wall time meaningful time for an individual Repetition is needed, the timestamp of the set id can be used for the last repetitionand offsets can be computed.

Repetition

pub struct Repetition {
    number: i32,
    rom: f64,
    duration: f64,
    time: i64,
}

The Repetition bucket

Similar to the ImuData bucket, this record is not defined in the code.

pub struct RepetitionBucket {
    pub _id: ObjectId,
    pub device_id: Uuid,
    pub kind: "Repetition",
    pub session_id: Uuid,
    pub set_id: Uuid,
    pub id: [device_id]_[timestamp of last insert(ns)],
    pub rep_count: u64,
    pub repetitions: Vec<Repetition>
}