initial side-by-side impl for linux temmp and process as a poc

src/data_collection/batteries.rs

@@ -4,7 +4,14 @@
 //! the battery crate.
 
 cfg_if::cfg_if! {
-    if #[cfg(any(target_os = "windows", target_os = "macos", target_os = "linux", target_os = "freebsd", target_os = "dragonfly", target_os = "ios"))] {
+    if #[cfg(any(
+        target_os = "windows",
+        target_os = "macos",
+        target_os = "linux",
+        target_os = "freebsd",
+        target_os = "dragonfly",
+        target_os = "ios",
+    ))] {
         pub mod battery;
         pub use self::battery::*;
     }

src/new_data_collection/README.md

@@ -1,5 +1,8 @@
 # Data Collection
 
+**Note:** This information is really only useful to _developers_ of bottom,
+and can be ignored by users.
+
 Data collection in bottom has two main components: **sources** and **collectors**.
 
 **Sources** are either libraries or system APIs that actually extract the data.
@@ -8,7 +11,7 @@ or `libc` bindings, or Linux-specific code.
 
 **Collectors** are _platform-specific_ (typically OS-specific), and can pull from
 different sources to get all the data needed, with some glue code in between. As
-such, sources should be written to be per-"job", and be divisible such that
+such, sources should be written to be per-"job", and be divided such that
 collectors can import specific code as needed.
 
 We can kinda visualize this with a quick-and-dirty diagram (note this is not accurate or up-to-date):
@@ -42,3 +45,11 @@ flowchart TB
     freebsd -..-> FreeBSD
     sysinfo -..-> FreeBSD
 ```
+
+## Sources
+
+As mentioned above, sources should be written in a way where collectors can easily pick the necessary code required.
+
+## Collectors
+
+Each platform should implement the `DataCollector` trait in `collectors/common.rs`. The trait has default implementations where essentially no work is done, which is used as fallback behaviour.

src/new_data_collection/collectors/common.rs

@@ -0,0 +1,27 @@
+//! Common code amongst all data collectors.
+
+use crate::new_data_collection::{
+    error::CollectionResult,
+    sources::common::{processes::ProcessHarvest, temperature::TemperatureData},
+};
+
+/// The trait representing what a per-platform data collector should implement.
+pub(crate) trait DataCollector {
+    /// Refresh inner data sources to prepare them for gathering data.
+    ///
+    /// Note that depending on the implementation, this may
+    /// not actually need to do anything.
+    fn refresh_data(&mut self) -> CollectionResult<()> {
+        Ok(())
+    }
+
+    /// Return temperature data.
+    fn get_temperature_data(&mut self) -> CollectionResult<Vec<TemperatureData>> {
+        Ok(vec![])
+    }
+
+    /// Return process data.
+    fn get_process_data(&mut self) -> CollectionResult<Vec<ProcessHarvest>> {
+        Ok(vec![])
+    }
+}

src/new_data_collection/collectors/fallback.rs

@@ -0,0 +1,7 @@
+use super::common::DataCollector;
+
+/// A fallback [`DataCollector`] for unsupported systems
+/// that does nothing.
+pub struct FallbackDataCollector {}
+
+impl DataCollector for FallbackDataCollector {}

src/new_data_collection/collectors/freebsd.rs

@@ -0,0 +1,36 @@
+//! The data collector for FreeBSD.
+
+use crate::{
+    app::filter::Filter,
+    new_data_collection::{
+        error::CollectionResult,
+        sources::{
+            common::temperature::{TemperatureData, TemperatureType},
+            sysinfo::temperature::get_temperature_data,
+        },
+    },
+};
+
+use super::common::DataCollector;
+
+/// The [`DataCollector`] for FreeBSD.
+pub struct FreeBsdDataCollector {
+    temp_type: TemperatureType,
+    temp_filters: Option<Filter>,
+}
+
+impl DataCollector for FreeBsdDataCollector {
+    fn refresh_data(&mut self) -> CollectionResult<()> {
+        Ok(())
+    }
+
+    fn get_temperature_data(&self) -> CollectionResult<Option<Vec<TemperatureData>>> {
+        let mut results = get_temperature_data(&self.temp_type, &self.temp_filters);
+
+        for entry in sysctl_temp_iter(&self.temp_type, &self.temp_filters) {
+            results.push(entry);
+        }
+
+        Ok(Some(results))
+    }
+}

src/new_data_collection/collectors/linux.rs

@@ -0,0 +1,71 @@
+//! The data collector for Linux.
+
+use std::time::Instant;
+
+use starship_battery::{Battery, Manager};
+
+use crate::{
+    app::filter::Filter,
+    new_data_collection::{
+        error::CollectionResult,
+        sources::{
+            common::{
+                processes::ProcessHarvest,
+                temperature::{TemperatureData, TemperatureType},
+            },
+            linux::{
+                processes::{linux_process_data, ProcessCollector},
+                temperature::get_temperature_data,
+            },
+        },
+    },
+};
+
+use super::common::DataCollector;
+
+/// The [`DataCollector`] for Linux.
+pub struct LinuxDataCollector {
+    current_collection_time: Instant,
+    last_collection_time: Instant,
+
+    temp_type: TemperatureType,
+    temp_filters: Option<Filter>,
+
+    proc_collector: ProcessCollector,
+
+    system: sysinfo::System,
+    network: sysinfo::Networks,
+
+    #[cfg(feature = "battery")]
+    battery_manager: Option<Manager>,
+    #[cfg(feature = "battery")]
+    battery_list: Option<Vec<Battery>>,
+
+    #[cfg(feature = "gpu")]
+    gpus_total_mem: Option<u64>,
+}
+
+impl DataCollector for LinuxDataCollector {
+    fn refresh_data(&mut self) -> CollectionResult<()> {
+        Ok(())
+    }
+
+    fn get_temperature_data(&mut self) -> CollectionResult<Vec<TemperatureData>> {
+        Ok(get_temperature_data(&self.temp_type, &self.temp_filters))
+    }
+
+    fn get_process_data(&mut self) -> CollectionResult<Vec<ProcessHarvest>> {
+        let time_diff = self
+            .current_collection_time
+            .duration_since(self.last_collection_time)
+            .as_secs();
+
+        linux_process_data(
+            &self.system,
+            time_diff,
+            &mut self.proc_collector,
+            #[cfg(feature = "gpu")]
+            self.gpus_total_mem,
+        )
+    }
+}

src/new_data_collection/collectors/macos.rs

@@ -0,0 +1,33 @@
+//! The data collector for macOS.
+
+use crate::{
+    app::filter::Filter,
+    new_data_collection::{
+        error::CollectionResult,
+        sources::{
+            common::temperature::{TemperatureData, TemperatureType},
+            sysinfo::temperature::get_temperature_data,
+        },
+    },
+};
+
+use super::common::DataCollector;
+
+/// The [`DataCollector`] for macOS.
+pub struct MacOsDataCollector {
+    temp_type: TemperatureType,
+    temp_filters: Option<Filter>,
+}
+
+impl DataCollector for MacOsDataCollector {
+    fn refresh_data(&mut self) -> CollectionResult<()> {
+        Ok(())
+    }
+
+    fn get_temperature_data(&self) -> CollectionResult<Option<Vec<TemperatureData>>> {
+        Ok(Some(get_temperature_data(
+            &self.temp_type,
+            &self.temp_filters,
+        )))
+    }
+}

src/new_data_collection/collectors/windows.rs

@@ -0,0 +1,33 @@
+//! The data collector for Windows.
+
+use crate::{
+    app::filter::Filter,
+    new_data_collection::{
+        error::CollectionResult,
+        sources::{
+            common::temperature::{TemperatureData, TemperatureType},
+            sysinfo::temperature::get_temperature_data,
+        },
+    },
+};
+
+use super::common::DataCollector;
+
+/// The [`DataCollector`] for Windows.
+pub struct WindowsDataCollector {
+    temp_type: TemperatureType,
+    temp_filters: Option<Filter>,
+}
+
+impl DataCollector for WindowsDataCollector {
+    fn refresh_data(&mut self) -> CollectionResult<()> {
+        Ok(())
+    }
+
+    fn get_temperature_data(&self) -> CollectionResult<Option<Vec<TemperatureData>>> {
+        Ok(Some(get_temperature_data(
+            &self.temp_type,
+            &self.temp_filters,
+        )))
+    }
+}

src/new_data_collection/error.rs

@@ -0,0 +1,42 @@
+use anyhow::anyhow;
+
+/// An error to do with data collection.
+#[derive(Debug)]
+pub enum CollectionError {
+    /// A general error to propagate back up. A wrapper around [`anyhow::Error`].
+    General(anyhow::Error),
+
+    /// The collection is unsupported.
+    Unsupported,
+}
+
+impl std::fmt::Display for CollectionError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        match self {
+            CollectionError::General(err) => err.fmt(f),
+            CollectionError::Unsupported => {
+                write!(
+                    f,
+                    "bottom does not support this type of data collection for this platform."
+                )
+            }
+        }
+    }
+}
+
+impl std::error::Error for CollectionError {}
+
+/// A [`Result`] with the error type being a [`DataCollectionError`].
+pub(crate) type CollectionResult<T> = Result<T, CollectionError>;
+
+impl From<std::io::Error> for CollectionError {
+    fn from(err: std::io::Error) -> Self {
+        Self::General(err.into())
+    }
+}
+
+impl From<&'static str> for CollectionError {
+    fn from(msg: &'static str) -> Self {
+        Self::General(anyhow!(msg))
+    }
+}

src/new_data_collection/mod.rs

@@ -0,0 +1,28 @@
+//! Module that just re-exports the right data collector for a given platform.
+
+pub mod error;
+
+mod collectors {
+    pub mod common;
+
+    cfg_if::cfg_if! {
+        if #[cfg(target_os = "linux")] {
+            pub mod linux;
+            pub use linux::LinuxDataCollector as DataCollectorImpl;
+        } else if #[cfg(target_os = "macos")] {
+            pub mod macos;
+            pub use macos::MacOsDataCollector as DataCollectorImpl;
+        } else if #[cfg(target_os = "windows")] {
+            pub mod windows;
+            pub use windows::WindowsDataCollector as DataCollectorImpl;
+        } else if #[cfg(target_os = "freebsd")] {
+            pub mod freebsd;
+            pub use freebsd::FreeBsdDataCollector as DataCollectorImpl;
+        } else {
+            pub mod fallback;
+            pub use fallback::FallbackDataCollector as DataCollectorImpl;
+        }
+    }
+}
+
+pub mod sources;

src/new_data_collection/sources/common/mod.rs

@@ -0,0 +1,2 @@
+pub mod processes;
+pub mod temperature;

src/new_data_collection/sources/common/processes.rs

@@ -0,0 +1,87 @@
+use std::{borrow::Cow, time::Duration};
+
+use crate::new_data_collection::sources::Pid;
+
+#[derive(Debug, Clone, Default)]
+pub struct ProcessHarvest {
+    /// The pid of the process.
+    pub pid: Pid,
+
+    /// The parent PID of the process. A `parent_pid` of 0 is usually the root.
+    pub parent_pid: Option<Pid>,
+
+    /// CPU usage as a percentage.
+    pub cpu_usage_percent: f32,
+
+    /// Memory usage as a percentage.
+    pub mem_usage_percent: f32,
+
+    /// Memory usage as bytes.
+    pub mem_usage_bytes: u64,
+
+    /// The name of the process.
+    pub name: String,
+
+    /// The exact command for the process.
+    pub command: String,
+
+    /// Bytes read per second.
+    pub read_bytes_per_sec: u64,
+
+    /// Bytes written per second.
+    pub write_bytes_per_sec: u64,
+
+    /// The total number of bytes read by the process.
+    pub total_read_bytes: u64,
+
+    /// The total number of bytes written by the process.
+    pub total_write_bytes: u64,
+
+    /// The current state of the process (e.g. zombie, asleep).
+    pub process_state: (String, char),
+
+    /// Cumulative process uptime.
+    pub time: Duration,
+
+    /// This is the *effective* user ID of the process. This is only used on
+    /// Unix platforms.
+    #[cfg(target_family = "unix")]
+    pub uid: Option<libc::uid_t>,
+
+    /// This is the process' user.
+    pub user: Cow<'static, str>,
+
+    /// GPU memory usage as bytes.
+    #[cfg(feature = "gpu")]
+    pub gpu_mem: u64,
+
+    /// GPU memory usage as percentage.
+    #[cfg(feature = "gpu")]
+    pub gpu_mem_percent: f32,
+
+    /// GPU utilization as a percentage.
+    #[cfg(feature = "gpu")]
+    pub gpu_util: u32,
+    // TODO: Additional fields
+    // pub rss_kb: u64,
+    // pub virt_kb: u64,
+}
+
+impl ProcessHarvest {
+    pub(crate) fn add(&mut self, rhs: &ProcessHarvest) {
+        self.cpu_usage_percent += rhs.cpu_usage_percent;
+        self.mem_usage_bytes += rhs.mem_usage_bytes;
+        self.mem_usage_percent += rhs.mem_usage_percent;
+        self.read_bytes_per_sec += rhs.read_bytes_per_sec;
+        self.write_bytes_per_sec += rhs.write_bytes_per_sec;
+        self.total_read_bytes += rhs.total_read_bytes;
+        self.total_write_bytes += rhs.total_write_bytes;
+        self.time = self.time.max(rhs.time);
+        #[cfg(feature = "gpu")]
+        {
+            self.gpu_mem += rhs.gpu_mem;
+            self.gpu_util += rhs.gpu_util;
+            self.gpu_mem_percent += rhs.gpu_mem_percent;
+        }
+    }
+}

src/new_data_collection/sources/common/temperature.rs

@@ -0,0 +1,70 @@
+use std::str::FromStr;
+
+#[derive(Default, Debug, Clone)]
+pub struct TemperatureData {
+    pub name: String,
+    pub temperature: Option<f32>,
+}
+
+#[derive(Clone, Debug, Copy, PartialEq, Eq, Default)]
+pub enum TemperatureType {
+    #[default]
+    Celsius,
+    Kelvin,
+    Fahrenheit,
+}
+
+impl FromStr for TemperatureType {
+    type Err = String;
+
+    fn from_str(s: &str) -> Result<Self, Self::Err> {
+        match s {
+            "fahrenheit" | "f" => Ok(TemperatureType::Fahrenheit),
+            "kelvin" | "k" => Ok(TemperatureType::Kelvin),
+            "celsius" | "c" => Ok(TemperatureType::Celsius),
+            _ => Err(format!(
+                "'{s}' is an invalid temperature type, use one of: [kelvin, k, celsius, c, fahrenheit, f]."
+            )),
+        }
+    }
+}
+
+impl TemperatureType {
+    /// Given a temperature in Celsius, covert it if necessary for a different
+    /// unit.
+    pub fn convert_temp_unit(&self, temp_celsius: f32) -> f32 {
+        fn convert_celsius_to_kelvin(celsius: f32) -> f32 {
+            celsius + 273.15
+        }
+
+        fn convert_celsius_to_fahrenheit(celsius: f32) -> f32 {
+            (celsius * (9.0 / 5.0)) + 32.0
+        }
+
+        match self {
+            TemperatureType::Celsius => temp_celsius,
+            TemperatureType::Kelvin => convert_celsius_to_kelvin(temp_celsius),
+            TemperatureType::Fahrenheit => convert_celsius_to_fahrenheit(temp_celsius),
+        }
+    }
+}
+
+#[cfg(test)]
+mod test {
+    use crate::new_data_collection::sources::common::temperature::TemperatureType;
+
+    #[test]
+    fn temp_conversions() {
+        const TEMP: f32 = 100.0;
+
+        assert_eq!(
+            TemperatureType::Celsius.convert_temp_unit(TEMP),
+            TEMP,
+            "celsius to celsius is the same"
+        );
+
+        assert_eq!(TemperatureType::Kelvin.convert_temp_unit(TEMP), 373.15);
+
+        assert_eq!(TemperatureType::Fahrenheit.convert_temp_unit(TEMP), 212.0);
+    }
+}

src/new_data_collection/sources/freebsd/mod.rs

@@ -0,0 +1 @@
+mod temperature;

src/new_data_collection/sources/freebsd/temperature.rs

@@ -0,0 +1,35 @@
+//! FreeBSD-specific temperature extraction code.
+
+// For RockPro64 boards on FreeBSD, they apparently use "hw.temperature" for
+// sensors.
+use sysctl::Sysctl;
+
+/// Return an iterator of temperature data pulled from sysctl.
+pub(crate) fn sysctl_temp_iter(
+    temp_type: &TemperatureType, filter: &Option<Filter>,
+) -> impl Iterator<Item = TemperatureData> {
+    const KEY: &str = "hw.temperature";
+
+    if let Ok(root) = sysctl::Ctl::new(KEY) {
+        sysctl::CtlIter::below(root).flatten().filter_map(|ctl| {
+            if let (Ok(name), Ok(temp)) = (ctl.name(), ctl.value()) {
+                if let Some(temp) = temp.as_temperature() {
+                    if Filter::optional_should_keep(filter, &name) {
+                        return Some(TemperatureData {
+                            name,
+                            temperature: Some(match temp_type {
+                                TemperatureType::Celsius => temp.celsius(),
+                                TemperatureType::Kelvin => temp.kelvin(),
+                                TemperatureType::Fahrenheit => temp.fahrenheit(),
+                            }),
+                        });
+                    }
+                }
+            }
+
+            None
+        })
+    } else {
+        std::iter::empty()
+    }
+}

src/new_data_collection/sources/linux/mod.rs

@@ -0,0 +1,2 @@
+pub mod processes;
+pub mod temperature;

src/new_data_collection/sources/linux/processes/mod.rs

@@ -0,0 +1,450 @@
+//! Process data collection for Linux.
+
+mod process;
+
+use std::{
+    fs::{self, File},
+    io::{BufRead, BufReader},
+    time::Duration,
+};
+
+use hashbrown::{HashMap, HashSet};
+use process::*;
+use sysinfo::ProcessStatus;
+
+use crate::new_data_collection::{
+    error::CollectionResult,
+    sources::{common::processes::ProcessHarvest, unix::processes::user_table::UserTable, Pid},
+};
+
+/// Maximum character length of a `/proc/<PID>/stat`` process name.
+/// If it's equal or greater, then we instead refer to the command for the name.
+const MAX_STAT_NAME_LEN: usize = 15;
+
+#[derive(Debug, Clone, Default)]
+pub struct PrevProcDetails {
+    total_read_bytes: u64,
+    total_write_bytes: u64,
+    cpu_time: u64,
+}
+
+/// Given `/proc/stat` file contents, determine the idle and non-idle values of
+/// the CPU used to calculate CPU usage.
+fn fetch_cpu_usage(line: &str) -> (f64, f64) {
+    /// Converts a `Option<&str>` value to an f64. If it fails to parse or is
+    /// `None`, it will return `0_f64`.
+    fn str_to_f64(val: Option<&str>) -> f64 {
+        val.and_then(|v| v.parse::<f64>().ok()).unwrap_or(0_f64)
+    }
+
+    let mut val = line.split_whitespace();
+    let user = str_to_f64(val.next());
+    let nice: f64 = str_to_f64(val.next());
+    let system: f64 = str_to_f64(val.next());
+    let idle: f64 = str_to_f64(val.next());
+    let iowait: f64 = str_to_f64(val.next());
+    let irq: f64 = str_to_f64(val.next());
+    let softirq: f64 = str_to_f64(val.next());
+    let steal: f64 = str_to_f64(val.next());
+
+    // Note we do not get guest/guest_nice, as they are calculated as part of
+    // user/nice respectively See https://github.com/htop-dev/htop/blob/main/linux/LinuxProcessList.c
+    let idle = idle + iowait;
+    let non_idle = user + nice + system + irq + softirq + steal;
+
+    (idle, non_idle)
+}
+
+struct CpuUsage {
+    /// Difference between the total delta and the idle delta.
+    cpu_usage: f64,
+
+    /// Overall CPU usage as a fraction.
+    cpu_fraction: f64,
+}
+
+fn cpu_usage_calculation(
+    prev_idle: &mut f64, prev_non_idle: &mut f64,
+) -> CollectionResult<CpuUsage> {
+    let (idle, non_idle) = {
+        // From SO answer: https://stackoverflow.com/a/23376195
+        let first_line = {
+            // We just need a single line from this file. Read it and return it.
+            let mut reader = BufReader::new(File::open("/proc/stat")?);
+            let mut buffer = String::new();
+            reader.read_line(&mut buffer)?;
+
+            buffer
+        };
+
+        fetch_cpu_usage(&first_line)
+    };
+
+    let total = idle + non_idle;
+    let prev_total = *prev_idle + *prev_non_idle;
+
+    let total_delta = total - prev_total;
+    let idle_delta = idle - *prev_idle;
+
+    *prev_idle = idle;
+    *prev_non_idle = non_idle;
+
+    // TODO: Should these return errors instead?
+    let cpu_usage = if total_delta - idle_delta != 0.0 {
+        total_delta - idle_delta
+    } else {
+        1.0
+    };
+
+    let cpu_fraction = if total_delta != 0.0 {
+        cpu_usage / total_delta
+    } else {
+        0.0
+    };
+
+    Ok(CpuUsage {
+        cpu_usage,
+        cpu_fraction,
+    })
+}
+
+/// Returns the usage and a new set of process times.
+///
+/// NB: cpu_fraction should be represented WITHOUT the x100 factor!
+fn get_linux_cpu_usage(
+    stat: &Stat, cpu_usage: f64, cpu_fraction: f64, prev_proc_times: u64,
+    use_current_cpu_total: bool,
+) -> (f32, u64) {
+    // Based heavily on https://stackoverflow.com/a/23376195 and https://stackoverflow.com/a/1424556
+    let new_proc_times = stat.utime + stat.stime;
+    let diff = (new_proc_times - prev_proc_times) as f64; // No try_from for u64 -> f64... oh well.
+
+    if cpu_usage == 0.0 {
+        (0.0, new_proc_times)
+    } else if use_current_cpu_total {
+        (((diff / cpu_usage) * 100.0) as f32, new_proc_times)
+    } else {
+        (
+            ((diff / cpu_usage) * 100.0 * cpu_fraction) as f32,
+            new_proc_times,
+        )
+    }
+}
+
+fn read_proc(
+    prev_proc: &PrevProcDetails, process: Process, args: ReadProcArgs, user_table: &mut UserTable,
+) -> CollectionResult<(ProcessHarvest, u64)> {
+    let Process {
+        pid: _,
+        uid,
+        stat,
+        io,
+        cmdline,
+    } = process;
+
+    let ReadProcArgs {
+        use_current_cpu_total,
+        cpu_usage,
+        cpu_fraction,
+        total_memory,
+        time_difference_in_secs,
+        uptime,
+    } = args;
+
+    let (command, name) = {
+        let truncated_name = stat.comm.as_str();
+        if let Ok(cmdline) = cmdline {
+            if cmdline.is_empty() {
+                (format!("[{truncated_name}]"), truncated_name.to_string())
+            } else {
+                (
+                    cmdline.join(" "),
+                    if truncated_name.len() >= MAX_STAT_NAME_LEN {
+                        if let Some(first_part) = cmdline.first() {
+                            // We're only interested in the executable part... not the file path.
+                            // That's for command.
+                            first_part
+                                .rsplit_once('/')
+                                .map(|(_prefix, suffix)| suffix)
+                                .unwrap_or(truncated_name)
+                                .to_string()
+                        } else {
+                            truncated_name.to_string()
+                        }
+                    } else {
+                        truncated_name.to_string()
+                    },
+                )
+            }
+        } else {
+            (truncated_name.to_string(), truncated_name.to_string())
+        }
+    };
+
+    let process_state_char = stat.state;
+    let process_state = (
+        ProcessStatus::from(process_state_char).to_string(),
+        process_state_char,
+    );
+    let (cpu_usage_percent, new_process_times) = get_linux_cpu_usage(
+        &stat,
+        cpu_usage,
+        cpu_fraction,
+        prev_proc.cpu_time,
+        use_current_cpu_total,
+    );
+    let parent_pid = Some(stat.ppid);
+    let mem_usage_bytes = stat.rss_bytes();
+    let mem_usage_percent = (mem_usage_bytes as f64 / total_memory as f64 * 100.0) as f32;
+
+    // This can fail if permission is denied!
+    let (total_read_bytes, total_write_bytes, read_bytes_per_sec, write_bytes_per_sec) =
+        if let Ok(io) = io {
+            let total_read_bytes = io.read_bytes;
+            let total_write_bytes = io.write_bytes;
+            let prev_total_read_bytes = prev_proc.total_read_bytes;
+            let prev_total_write_bytes = prev_proc.total_write_bytes;
+
+            let read_bytes_per_sec = total_read_bytes
+                .saturating_sub(prev_total_read_bytes)
+                .checked_div(time_difference_in_secs)
+                .unwrap_or(0);
+
+            let write_bytes_per_sec = total_write_bytes
+                .saturating_sub(prev_total_write_bytes)
+                .checked_div(time_difference_in_secs)
+                .unwrap_or(0);
+
+            (
+                total_read_bytes,
+                total_write_bytes,
+                read_bytes_per_sec,
+                write_bytes_per_sec,
+            )
+        } else {
+            (0, 0, 0, 0)
+        };
+
+    let user = uid
+        .and_then(|uid| {
+            user_table
+                .get_uid_to_username_mapping(uid)
+                .map(Into::into)
+                .ok()
+        })
+        .unwrap_or_else(|| "N/A".into());
+
+    let time = if let Ok(ticks_per_sec) = u32::try_from(rustix::param::clock_ticks_per_second()) {
+        if ticks_per_sec == 0 {
+            Duration::ZERO
+        } else {
+            Duration::from_secs(uptime.saturating_sub(stat.start_time / ticks_per_sec as u64))
+        }
+    } else {
+        Duration::ZERO
+    };
+
+    Ok((
+        ProcessHarvest {
+            pid: process.pid,
+            parent_pid,
+            cpu_usage_percent,
+            mem_usage_percent,
+            mem_usage_bytes,
+            name,
+            command,
+            read_bytes_per_sec,
+            write_bytes_per_sec,
+            total_read_bytes,
+            total_write_bytes,
+            process_state,
+            uid,
+            user,
+            time,
+            #[cfg(feature = "gpu")]
+            gpu_mem: 0,
+            #[cfg(feature = "gpu")]
+            gpu_mem_percent: 0.0,
+            #[cfg(feature = "gpu")]
+            gpu_util: 0,
+        },
+        new_process_times,
+    ))
+}
+
+pub(crate) struct PrevProc {
+    pub prev_idle: f64,
+    pub prev_non_idle: f64,
+}
+
+#[derive(Clone, Copy)]
+pub(crate) struct ProcHarvestOptions {
+    pub use_current_cpu_total: bool,
+    pub unnormalized_cpu: bool,
+}
+
+fn is_str_numeric(s: &str) -> bool {
+    s.chars().all(|c| c.is_ascii_digit())
+}
+
+/// General args to keep around for reading proc data.
+#[derive(Copy, Clone)]
+pub(crate) struct ReadProcArgs {
+    pub(crate) use_current_cpu_total: bool,
+    pub(crate) cpu_usage: f64,
+    pub(crate) cpu_fraction: f64,
+    pub(crate) total_memory: u64,
+    pub(crate) time_difference_in_secs: u64,
+    pub(crate) uptime: u64,
+}
+
+pub struct ProcessCollector {
+    pub options: ProcHarvestOptions,
+    pub prev_proc: PrevProc,
+    pub pid_mapping: HashMap<Pid, PrevProcDetails>,
+    pub user_table: UserTable,
+
+    #[cfg(feature = "gpu")]
+    pub gpu_pids: Option<Vec<HashMap<u32, (u64, u32)>>>,
+}
+
+pub(crate) fn linux_process_data(
+    system: &sysinfo::System, time_difference_in_secs: u64, collector: &mut ProcessCollector,
+    #[cfg(feature = "gpu")] gpus_total_mem: Option<u64>,
+) -> CollectionResult<Vec<ProcessHarvest>> {
+    let total_memory = system.total_memory();
+
+    let ProcHarvestOptions {
+        use_current_cpu_total,
+        unnormalized_cpu,
+    } = collector.options;
+
+    let PrevProc {
+        prev_idle,
+        prev_non_idle,
+    } = &mut collector.prev_proc;
+
+    // TODO: [PROC THREADS] Add threads
+
+    let CpuUsage {
+        mut cpu_usage,
+        cpu_fraction,
+    } = cpu_usage_calculation(prev_idle, prev_non_idle)?;
+
+    if unnormalized_cpu {
+        let num_processors = system.cpus().len() as f64;
+
+        // Note we *divide* here because the later calculation divides `cpu_usage` - in
+        // effect, multiplying over the number of cores.
+        cpu_usage /= num_processors;
+    }
+
+    let mut pids_to_clear: HashSet<Pid> = collector.pid_mapping.keys().cloned().collect();
+
+    let pids = fs::read_dir("/proc")?.flatten().filter_map(|dir| {
+        if is_str_numeric(dir.file_name().to_string_lossy().trim()) {
+            Some(dir.path())
+        } else {
+            None
+        }
+    });
+
+    let args = ReadProcArgs {
+        use_current_cpu_total,
+        cpu_usage,
+        cpu_fraction,
+        total_memory,
+        time_difference_in_secs,
+        uptime: sysinfo::System::uptime(),
+    };
+
+    let process_vector: Vec<ProcessHarvest> = pids
+        .filter_map(|pid_path| {
+            if let Ok(process) = Process::from_path(pid_path) {
+                let pid = process.pid;
+                let prev_proc_details = collector.pid_mapping.entry(pid).or_default();
+
+                #[allow(unused_mut)]
+                if let Ok((mut process_harvest, new_process_times)) =
+                    read_proc(prev_proc_details, process, args, &mut collector.user_table)
+                {
+                    #[cfg(feature = "gpu")]
+                    if let Some(gpus) = &collector.gpu_pids {
+                        gpus.iter().for_each(|gpu| {
+                            // add mem/util for all gpus to pid
+                            if let Some((mem, util)) = gpu.get(&(pid as u32)) {
+                                process_harvest.gpu_mem += mem;
+                                process_harvest.gpu_util += util;
+                            }
+                        });
+
+                        if let Some(gpu_total_mem) = gpus_total_mem {
+                            process_harvest.gpu_mem_percent =
+                                (process_harvest.gpu_mem as f64 / gpu_total_mem as f64 * 100.0)
+                                    as f32;
+                        }
+                    }
+
+                    prev_proc_details.cpu_time = new_process_times;
+                    prev_proc_details.total_read_bytes = process_harvest.total_read_bytes;
+                    prev_proc_details.total_write_bytes = process_harvest.total_write_bytes;
+
+                    pids_to_clear.remove(&pid);
+                    return Some(process_harvest);
+                }
+            }
+
+            None
+        })
+        .collect();
+
+    pids_to_clear.iter().for_each(|pid| {
+        collector.pid_mapping.remove(pid);
+    });
+
+    Ok(process_vector)
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    #[test]
+    fn test_proc_cpu_parse() {
+        assert_eq!(
+            (100_f64, 200_f64),
+            fetch_cpu_usage("100 0 100 100"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 4 values"
+        );
+        assert_eq!(
+            (120_f64, 200_f64),
+            fetch_cpu_usage("100 0 100 100 20"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 5 values"
+        );
+        assert_eq!(
+            (120_f64, 230_f64),
+            fetch_cpu_usage("100 0 100 100 20 30"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 6 values"
+        );
+        assert_eq!(
+            (120_f64, 270_f64),
+            fetch_cpu_usage("100 0 100 100 20 30 40"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 7 values"
+        );
+        assert_eq!(
+            (120_f64, 320_f64),
+            fetch_cpu_usage("100 0 100 100 20 30 40 50"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 8 values"
+        );
+        assert_eq!(
+            (120_f64, 320_f64),
+            fetch_cpu_usage("100 0 100 100 20 30 40 50 100"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 9 values"
+        );
+        assert_eq!(
+            (120_f64, 320_f64),
+            fetch_cpu_usage("100 0 100 100 20 30 40 50 100 200"),
+            "Failed to properly calculate idle/non-idle for /proc/stat CPU with 10 values"
+        );
+    }
+}

src/new_data_collection/sources/linux/processes/process.rs

@@ -0,0 +1,308 @@
+//! Linux process code for getting process data via `/proc/`.
+//! Based on the [procfs](https://github.com/eminence/procfs) crate.
+
+use std::{
+    fs::File,
+    io::{self, BufRead, BufReader, Read},
+    path::PathBuf,
+    sync::OnceLock,
+};
+
+use anyhow::anyhow;
+use libc::uid_t;
+use rustix::{
+    fd::OwnedFd,
+    fs::{Mode, OFlags},
+    path::Arg,
+};
+
+use crate::data_collection::processes::Pid;
+
+static PAGESIZE: OnceLock<u64> = OnceLock::new();
+
+#[inline]
+fn next_part<'a>(iter: &mut impl Iterator<Item = &'a str>) -> Result<&'a str, io::Error> {
+    iter.next()
+        .ok_or_else(|| io::Error::from(io::ErrorKind::InvalidData))
+}
+
+/// A wrapper around the data in `/proc/<PID>/stat`. For documentation, see
+/// [here](https://man7.org/linux/man-pages/man5/proc.5.html).
+///
+/// Note this does not necessarily get all fields, only the ones we use in
+/// bottom.
+pub(crate) struct Stat {
+    /// The filename of the executable without parentheses.
+    pub comm: String,
+
+    /// The current process state, represented by a char.
+    pub state: char,
+
+    /// The parent process PID.
+    pub ppid: Pid,
+
+    /// The amount of time this process has been scheduled in user mode in clock
+    /// ticks.
+    pub utime: u64,
+
+    /// The amount of time this process has been scheduled in kernel mode in
+    /// clock ticks.
+    pub stime: u64,
+
+    /// The resident set size, or the number of pages the process has in real
+    /// memory.
+    pub rss: u64,
+
+    /// The start time of the process, represented in clock ticks.
+    pub start_time: u64,
+}
+
+impl Stat {
+    #[inline]
+    fn from_file(mut f: File, buffer: &mut String) -> anyhow::Result<Stat> {
+        // Since this is just one line, we can read it all at once. However, since it
+        // might have non-utf8 characters, we can't just use read_to_string.
+        f.read_to_end(unsafe { buffer.as_mut_vec() })?;
+
+        let line = buffer.to_string_lossy();
+        let line = line.trim();
+
+        let (comm, rest) = {
+            let start_paren = line
+                .find('(')
+                .ok_or_else(|| anyhow!("start paren missing"))?;
+            let end_paren = line.find(')').ok_or_else(|| anyhow!("end paren missing"))?;
+
+            (
+                line[start_paren + 1..end_paren].to_string(),
+                &line[end_paren + 2..],
+            )
+        };
+
+        let mut rest = rest.split(' ');
+        let state = next_part(&mut rest)?
+            .chars()
+            .next()
+            .ok_or_else(|| anyhow!("missing state"))?;
+        let ppid: Pid = next_part(&mut rest)?.parse()?;
+
+        // Skip 9 fields until utime (pgrp, session, tty_nr, tpgid, flags, minflt,
+        // cminflt, majflt, cmajflt).
+        let mut rest = rest.skip(9);
+        let utime: u64 = next_part(&mut rest)?.parse()?;
+        let stime: u64 = next_part(&mut rest)?.parse()?;
+
+        // Skip 6 fields until starttime (cutime, cstime, priority, nice, num_threads,
+        // itrealvalue).
+        let mut rest = rest.skip(6);
+        let start_time: u64 = next_part(&mut rest)?.parse()?;
+
+        // Skip one field until rss (vsize)
+        let mut rest = rest.skip(1);
+        let rss: u64 = next_part(&mut rest)?.parse()?;
+
+        Ok(Stat {
+            comm,
+            state,
+            ppid,
+            utime,
+            stime,
+            rss,
+            start_time,
+        })
+    }
+
+    /// Returns the Resident Set Size in bytes.
+    #[inline]
+    pub fn rss_bytes(&self) -> u64 {
+        self.rss * PAGESIZE.get_or_init(|| rustix::param::page_size() as u64)
+    }
+}
+
+/// A wrapper around the data in `/proc/<PID>/io`.
+///
+/// Note this does not necessarily get all fields, only the ones we use in
+/// bottom.
+pub(crate) struct Io {
+    pub read_bytes: u64,
+    pub write_bytes: u64,
+}
+
+impl Io {
+    #[inline]
+    fn from_file(f: File, buffer: &mut String) -> anyhow::Result<Io> {
+        const NUM_FIELDS: u16 = 0; // Make sure to update this if you want more fields!
+        enum Fields {
+            ReadBytes,
+            WriteBytes,
+        }
+
+        let mut read_fields = 0;
+        let mut reader = BufReader::new(f);
+
+        let mut read_bytes = 0;
+        let mut write_bytes = 0;
+
+        // This saves us from doing a string allocation on each iteration compared to
+        // `lines()`.
+        while let Ok(bytes) = reader.read_line(buffer) {
+            if bytes > 0 {
+                if buffer.is_empty() {
+                    // Empty, no need to clear.
+                    continue;
+                }
+
+                let mut parts = buffer.split_whitespace();
+
+                if let Some(field) = parts.next() {
+                    let curr_field = match field {
+                        "read_bytes:" => Fields::ReadBytes,
+                        "write_bytes:" => Fields::WriteBytes,
+                        _ => {
+                            buffer.clear();
+                            continue;
+                        }
+                    };
+
+                    if let Some(value) = parts.next() {
+                        let value = value.parse::<u64>()?;
+                        match curr_field {
+                            Fields::ReadBytes => {
+                                read_bytes = value;
+                                read_fields += 1;
+                            }
+                            Fields::WriteBytes => {
+                                write_bytes = value;
+                                read_fields += 1;
+                            }
+                        }
+                    }
+                }
+
+                // Quick short circuit if we have already read all the required fields.
+                if read_fields == NUM_FIELDS {
+                    break;
+                }
+
+                buffer.clear();
+            } else {
+                break;
+            }
+        }
+
+        Ok(Io {
+            read_bytes,
+            write_bytes,
+        })
+    }
+}
+
+/// A wrapper around a Linux process operations in `/proc/<PID>`.
+///
+/// Core documentation based on [proc's manpages](https://man7.org/linux/man-pages/man5/proc.5.html).
+pub(crate) struct Process {
+    pub pid: Pid,
+    pub uid: Option<uid_t>,
+    pub stat: Stat,
+    pub io: anyhow::Result<Io>,
+    pub cmdline: anyhow::Result<Vec<String>>,
+}
+
+#[inline]
+fn reset(root: &mut PathBuf, buffer: &mut String) {
+    root.pop();
+    buffer.clear();
+}
+
+impl Process {
+    /// Creates a new [`Process`] given a `/proc/<PID>` path. This may fail if
+    /// the process no longer exists or there are permissions issues.
+    ///
+    /// Note that this pre-allocates fields on **creation**! As such, some data
+    /// might end up "outdated" depending on when you call some of the
+    /// methods. Therefore, this struct is only useful for either fields
+    /// that are unlikely to change, or are short-lived and
+    /// will be discarded quickly.
+    pub(crate) fn from_path(pid_path: PathBuf) -> anyhow::Result<Process> {
+        // TODO: Pass in a buffer vec/string to share?
+
+        let fd = rustix::fs::openat(
+            rustix::fs::CWD,
+            &pid_path,
+            OFlags::PATH | OFlags::DIRECTORY | OFlags::CLOEXEC,
+            Mode::empty(),
+        )?;
+
+        let pid = pid_path
+            .as_path()
+            .components()
+            .last()
+            .and_then(|s| s.to_string_lossy().parse::<Pid>().ok())
+            .or_else(|| {
+                rustix::fs::readlinkat(rustix::fs::CWD, &pid_path, vec![])
+                    .ok()
+                    .and_then(|s| s.to_string_lossy().parse::<Pid>().ok())
+            })
+            .ok_or_else(|| anyhow!("PID for {pid_path:?} was not found"))?;
+
+        let uid = {
+            let metadata = rustix::fs::fstat(&fd);
+            match metadata {
+                Ok(md) => Some(md.st_uid),
+                Err(_) => None,
+            }
+        };
+
+        let mut root = pid_path;
+        let mut buffer = String::new();
+
+        // NB: Whenever you add a new stat, make sure to pop the root and clear the
+        // buffer!
+        let stat =
+            open_at(&mut root, "stat", &fd).and_then(|file| Stat::from_file(file, &mut buffer))?;
+        reset(&mut root, &mut buffer);
+
+        let cmdline = cmdline(&mut root, &fd, &mut buffer);
+        reset(&mut root, &mut buffer);
+
+        let io = open_at(&mut root, "io", &fd).and_then(|file| Io::from_file(file, &mut buffer));
+
+        Ok(Process {
+            pid,
+            uid,
+            stat,
+            io,
+            cmdline,
+        })
+    }
+}
+
+#[inline]
+fn cmdline(root: &mut PathBuf, fd: &OwnedFd, buffer: &mut String) -> anyhow::Result<Vec<String>> {
+    open_at(root, "cmdline", fd)
+        .map(|mut file| file.read_to_string(buffer))
+        .map(|_| {
+            buffer
+                .split('\0')
+                .filter_map(|s| {
+                    if !s.is_empty() {
+                        Some(s.to_string())
+                    } else {
+                        None
+                    }
+                })
+                .collect::<Vec<_>>()
+        })
+        .map_err(Into::into)
+}
+
+/// Opens a path. Note that this function takes in a mutable root - this will
+/// mutate it to avoid allocations. You probably will want to pop the most
+/// recent child after if you need to use the buffer again.
+#[inline]
+fn open_at(root: &mut PathBuf, child: &str, fd: &OwnedFd) -> anyhow::Result<File> {
+    root.push(child);
+    let new_fd = rustix::fs::openat(fd, &*root, OFlags::RDONLY | OFlags::CLOEXEC, Mode::empty())?;
+
+    Ok(File::from(new_fd))
+}

src/new_data_collection/sources/linux/temperature.rs

@@ -8,15 +8,17 @@ use std::{
 use anyhow::Result;
 use hashbrown::{HashMap, HashSet};
 
-use super::{TempHarvest, TemperatureType};
-use crate::app::filter::Filter;
+use crate::{
+    app::filter::Filter,
+    new_data_collection::sources::common::temperature::{TemperatureData, TemperatureType},
+};
 
 const EMPTY_NAME: &str = "Unknown";
 
 /// Returned results from grabbing hwmon/coretemp temperature sensor
 /// values/names.
 struct HwmonResults {
-    temperatures: Vec<TempHarvest>,
+    temperatures: Vec<TemperatureData>,
     num_hwmon: usize,
 }
 
@@ -224,7 +226,7 @@ fn is_device_awake(path: &Path) -> bool {
 /// once this happens, the device will be *kept* on through the sensor
 /// reading, and not be able to re-enter ACPI D3cold.
 fn hwmon_temperatures(temp_type: &TemperatureType, filter: &Option<Filter>) -> HwmonResults {
-    let mut temperatures: Vec<TempHarvest> = vec![];
+    let mut temperatures: Vec<TemperatureData> = vec![];
     let mut seen_names: HashMap<String, u32> = HashMap::new();
 
     let (dirs, num_hwmon) = get_hwmon_candidates();
@@ -246,7 +248,7 @@ fn hwmon_temperatures(temp_type: &TemperatureType, filter: &Option<Filter>) -> H
 
         if !is_device_awake(&file_path) {
             let name = finalize_name(None, None, &sensor_name, &mut seen_names);
-            temperatures.push(TempHarvest {
+            temperatures.push(TemperatureData {
                 name,
                 temperature: None,
             });
@@ -329,7 +331,7 @@ fn hwmon_temperatures(temp_type: &TemperatureType, filter: &Option<Filter>) -> H
                 // probing hwmon if not needed?
                 if Filter::optional_should_keep(filter, &name) {
                     if let Ok(temp_celsius) = parse_temp(&temp_path) {
-                        temperatures.push(TempHarvest {
+                        temperatures.push(TemperatureData {
                             name,
                             temperature: Some(temp_type.convert_temp_unit(temp_celsius)),
                         });
@@ -352,7 +354,7 @@ fn hwmon_temperatures(temp_type: &TemperatureType, filter: &Option<Filter>) -> H
 /// See [the Linux kernel documentation](https://www.kernel.org/doc/Documentation/ABI/testing/sysfs-class-thermal)
 /// for more details.
 fn add_thermal_zone_temperatures(
-    temperatures: &mut Vec<TempHarvest>, temp_type: &TemperatureType, filter: &Option<Filter>,
+    temperatures: &mut Vec<TemperatureData>, temp_type: &TemperatureType, filter: &Option<Filter>,
 ) {
     let path = Path::new("/sys/class/thermal");
     let Ok(read_dir) = path.read_dir() else {
@@ -382,7 +384,7 @@ fn add_thermal_zone_temperatures(
                     if let Ok(temp_celsius) = parse_temp(&temp_path) {
                         let name = counted_name(&mut seen_names, name);
 
-                        temperatures.push(TempHarvest {
+                        temperatures.push(TemperatureData {
                             name,
                             temperature: Some(temp_type.convert_temp_unit(temp_celsius)),
                         });
@@ -396,14 +398,14 @@ fn add_thermal_zone_temperatures(
 /// Gets temperature sensors and data.
 pub fn get_temperature_data(
     temp_type: &TemperatureType, filter: &Option<Filter>,
-) -> Result<Option<Vec<TempHarvest>>> {
+) -> Vec<TemperatureData> {
     let mut results = hwmon_temperatures(temp_type, filter);
 
     if results.num_hwmon == 0 {
         add_thermal_zone_temperatures(&mut results.temperatures, temp_type, filter);
     }
 
-    Ok(Some(results.temperatures))
+    results.temperatures
 }
 
 #[cfg(test)]

src/new_data_collection/sources/mod.rs

@@ -1,14 +1,16 @@
-//! Re-exports all of the sources.
+pub mod common;
+pub mod linux;
+pub mod macos;
+#[cfg(feature = "gpu")]
+pub mod nvidia;
+pub mod sysinfo;
+pub mod unix;
+pub mod windows;
 
 cfg_if::cfg_if! {
-    if #[cfg(any(
-        target_os = "windows",
-        target_os = "macos",
-        target_os = "linux",
-        target_os = "freebsd",
-        target_os = "dragonfly",
-        target_os = "ios",
-    ))] {
-        pub mod starship_battery;
+    if #[cfg(target_family = "windows")] {
+        pub use windows::processes::Pid as Pid;
+    } else if #[cfg(target_family = "unix")] {
+        pub use unix::processes::Pid as Pid;
     }
 }

src/new_data_collection/sources/sysinfo/mod.rs

@@ -0,0 +1 @@
+pub mod temperature;

src/new_data_collection/sources/sysinfo/temperature.rs

@@ -1,51 +1,26 @@
 //! Gets temperature data via sysinfo.
 
-use anyhow::Result;
-
-use super::{TempHarvest, TemperatureType};
-use crate::app::filter::Filter;
+use crate::{
+    app::filter::Filter,
+    new_data_collection::sources::common::temperature::{TemperatureData, TemperatureType},
+};
 
 pub fn get_temperature_data(
     components: &sysinfo::Components, temp_type: &TemperatureType, filter: &Option<Filter>,
-) -> Result<Option<Vec<TempHarvest>>> {
-    let mut temperature_vec: Vec<TempHarvest> = Vec::new();
+) -> Vec<TemperatureData> {
+    let mut temperature_vec: Vec<TemperatureData> = Vec::new();
 
     for component in components {
         let name = component.label().to_string();
 
         if Filter::optional_should_keep(filter, &name) {
-            temperature_vec.push(TempHarvest {
+            temperature_vec.push(TemperatureData {
                 name,
                 temperature: Some(temp_type.convert_temp_unit(component.temperature())),
             });
         }
     }
 
-    // For RockPro64 boards on FreeBSD, they apparently use "hw.temperature" for
-    // sensors.
-    #[cfg(target_os = "freebsd")]
-    {
-        use sysctl::Sysctl;
-
-        const KEY: &str = "hw.temperature";
-        if let Ok(root) = sysctl::Ctl::new(KEY) {
-            for ctl in sysctl::CtlIter::below(root).flatten() {
-                if let (Ok(name), Ok(temp)) = (ctl.name(), ctl.value()) {
-                    if let Some(temp) = temp.as_temperature() {
-                        temperature_vec.push(TempHarvest {
-                            name,
-                            temperature: Some(match temp_type {
-                                TemperatureType::Celsius => temp.celsius(),
-                                TemperatureType::Kelvin => temp.kelvin(),
-                                TemperatureType::Fahrenheit => temp.fahrenheit(),
-                            }),
-                        });
-                    }
-                }
-            }
-        }
-    }
-
     // TODO: Should we instead use a hashmap -> vec to skip dupes?
-    Ok(Some(temperature_vec))
+    temperature_vec
 }

src/new_data_collection/sources/unix/mod.rs

@@ -0,0 +1 @@
+pub mod processes;

src/new_data_collection/sources/unix/processes/mod.rs

@@ -0,0 +1,5 @@
+pub mod user_table;
+
+/// A UNIX process ID.
+#[cfg(target_family = "unix")]
+pub type Pid = libc::pid_t;

src/new_data_collection/sources/unix/processes/user_table.rs

@@ -0,0 +1,33 @@
+use hashbrown::HashMap;
+
+use crate::data_collection::error::{CollectionError, CollectionResult};
+
+#[derive(Debug, Default)]
+pub struct UserTable {
+    pub uid_user_mapping: HashMap<libc::uid_t, String>,
+}
+
+impl UserTable {
+    pub fn get_uid_to_username_mapping(&mut self, uid: libc::uid_t) -> CollectionResult<String> {
+        if let Some(user) = self.uid_user_mapping.get(&uid) {
+            Ok(user.clone())
+        } else {
+            // SAFETY: getpwuid returns a null pointer if no passwd entry is found for the
+            // uid
+            let passwd = unsafe { libc::getpwuid(uid) };
+
+            if passwd.is_null() {
+                Err("passwd is inaccessible".into())
+            } else {
+                // SAFETY: We return early if passwd is null.
+                let username = unsafe { std::ffi::CStr::from_ptr((*passwd).pw_name) }
+                    .to_str()
+                    .map_err(|err| CollectionError::General(err.into()))?
+                    .to_string();
+                self.uid_user_mapping.insert(uid, username.clone());
+
+                Ok(username)
+            }
+        }
+    }
+}

src/new_data_collection/sources/windows/mod.rs

@@ -0,0 +1 @@
+pub mod processes;

src/new_data_collection/sources/windows/processes.rs

@@ -0,0 +1,3 @@
+/// A Windows process ID.
+#[cfg(target_family = "windows")]
+pub type Pid = usize;