various: deprecate `%cpu-usage`; add new variables by NotAShelf · Pull Request #56 · NotAShelf/watt

NotAShelf · 2026-02-17T06:24:25Z

Adds a few new system variables per #45. New variables:

`$cpu-scaling-maximum - CPU scaling maximum frequency in MHz
%cpu-core-count - Number of CPU cores
$load-average-since System load average over a given duration
?lid-closed - Boolean for lid state
$hour-of-day - Current hour (0-23) based on local time
$battery-cycles - Battery cycle count
%battery-health - Battery health percentage

Notes:

%cpu-usage is now fully deprecated. Updated example config with best-case examples I can think of. Could be optimized further if we decide to actually benchmark power consumption on a live system.
Like we do for CPU usage, it's worth adding support for measuring load average over certain durations rather than just fixed 1m, 5m and 15m averages.
- A system with many I/O-bound processes can show high load average but low CPU usage. Conversely, a few CPU-intensive tasks can show high CPU usage but moderate load. In this case we have a good reason to distinguish the two.
$cpu-scaling-minimum is not added, because minimum should be always 0. Maximum is what we'll want to use to infer system "capability". I cannot think of a single case where it will not be 0.
?$hour-of-day is added despite my belief that it'll be obsoleted by the D-Bus interface. It's a trivial addition, and I'd rather not curse people with D-Bus for something so trivial. It's good for some basic rules, and of trivial cost.
I'm not particularly proud of the battery aggregation. This'll need to be polished to support per-battery supports.

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Id8593744f5c1b4a862ff9d516570a1986a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ie65f50fae6ca27a115b12c3d04c7032f6a6a6964

…bles Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Idf55162fa903329d659caac19bbe92e86a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I3d5a06963de7b1f485861ce77607d2226a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Id94264310cd7f9ccde4a7d1edd162e336a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ia2809ab13363fa591132631f3956579a6a6a6964

Adds `%cpu-core-count`, `?lid-closed`, `$battery-cycles` and `%battery-health` variables. Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I513545b3c54aa7d725423896a295532a6a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ic792e0ccffd71f3229c03f91572c72006a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ifd02e5852dcfff8720ba3f631c47d00d6a6a6964

watt/system.rs

NotAShelf · 2026-02-17T10:13:12Z

watt/system.rs

+          // SW_LID is bit 0 in the capabilities bitmask
+          // The state file shows the current state of switches as a hex bitmask
+          // If bit 0 is set, the lid is closed
+          if let Ok(caps) = u64::from_str_radix(sw_caps.trim(), 16) {
+            self.lid_closed = (caps & 0x1) != 0;
+            log::debug!(
+              "lid state from input device {path}: {state}",
+              path = entry_path.display(),
+              state = if self.lid_closed { "closed" } else { "open" }
+            );


I don't know how this holds up against different systems. Will need to investigate.

NotAShelf · 2026-02-17T10:14:29Z

watt/power_supply.rs

+      // Battery health as a percentage (0-100)
+      // Some systems report this as capacity_level or health
+      self.health = if let Some(health) =
+        fs::read_n::<u64>(self.path.join("health"))
+          .with_context(|| format!("failed to read {self} health"))?
+      {
+        Some(health as f64 / 100.0)
+      } else {
+        // Try to calculate health from energy_full vs energy_full_design
+        let energy_full = fs::read_n::<u64>(self.path.join("energy_full"))
+          .with_context(|| format!("failed to read {self} energy_full"))?;
+
+        let energy_full_design =
+          fs::read_n::<u64>(self.path.join("energy_full_design"))
+            .with_context(|| {
+              format!("failed to read {self} energy_full_design")
+            })?;
+
+        match (energy_full, energy_full_design) {
+          (Some(full), Some(design)) if design > 0 => {
+            Some(full as f64 / design as f64)
+          },
+          _ => None,
+        }
+      };


Like we do for paths, we might want to classify those paths depending on the vendor. It's really difficult to support various vendors without diving into the kernel code, which is very time consuming. Though, I don't like assuming things either.

watt/power_supply.rs

watt/system.rs

watt/config.rs

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ie8f136a922eebace444a8b40cebd7eac6a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: If2d8e55a400a7a52903752fb480728056a6a6964

Now computed on-demand from `state.cpus`` during expression evaluation rather than stored as pre-computed fields in `EvalState`. Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ifeb78acbb2f8c28468b6bcc912f88b556a6a6964

I guess this was bound to happen and now I feel stupid for wasting my time with the static expressions earlier. This commit adds `load-average-since` that computers average load from the CPU log over a configurable duration, similiar to `cpu-usage-since`. Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I95a9e5dca7a34da5395ade1d6deb25a76a6a6964

I've changed cycles to cycle paths via search & replace previously so it accidentally regressed the path. This fixes the incorrect sysfs path from "cycles" to "cycle_count" for reading battery cycle count on Linux systems. Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ia801fe84abb87cbfbf5e14be14eeab566a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I99b2146dfff44569fd90436a836692476a6a6964

Adds new expression variants: - `BatteryCyclesFor { name }` - cycle count for specific battery - `BatteryHealthFor { name }` - health for specific battery - `IsBatteryAvailable { value }` - predicate to check battery existence Those eplace the aggregated-only `battery_cycles` and `battery_health`. Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I736448f3e8a73a6a90538cd99af188996a6a6964

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ibaeb1235a15aebe0450f6f056c6a9c966a6a6964

RGBCube · 2026-02-18T16:31:39Z

watt/config.rs

  *value == T::default()
 }

+fn find_battery<'a>(


This should really be inlined, or defined inside the fn that uses it

It can be moved to eval but I reckon we'll need it elsewhere if we end up adding more battery-specific expressions. For now I'm moving it but I strongly suspect it'll be moved out eventually.

watt/config.rs

RGBCube · 2026-02-18T16:33:17Z

watt/config.rs

  pub frequency_available: bool,
  pub turbo_available:     bool,

  pub cpu_usage:                  f64,


Remove this unused field

RGBCube · 2026-02-18T16:34:23Z

watt/config.rs

+        let max = state
+          .cpus
+          .iter()
+          .filter_map(|cpu| cpu.frequency_mhz_maximum)


I'm pretty sure this is the hardware limit. We have no field for software, add cpu.frequency_mhz_software_cap

RGBCube · 2026-02-18T16:36:00Z

watt/config.rs

+
+      CpuCoreCount => Number(state.cpus.len() as f64),
+
+      LoadAverageSince { duration } => {


duration should be an Expr so people can do ifs etc inside. We have .try_into_string as well.

RGBCube · 2026-02-18T16:38:49Z

watt/system.rs

+    let batteries: Vec<_> = self
+      .power_supplies
+      .iter()
+      .filter(|ps| ps.type_ == "Battery" && !ps.is_from_peripheral)
+      .collect();


Can't this use the find_batteries? Also, you don't seem to be checking is_from_peripheral in that fn def, I wonder if the disparity is a good idea

RGBCube · 2026-02-18T16:39:49Z

watt/system.rs

+        let entry = match entry {
+          Ok(entry) => entry,
+          Err(error) => {
+            log::debug!("failed to read input device entry: {error}");
+            continue;
+          },
+        };


This can be a let Ok(...) = ... else { log; continue; };

RGBCube · 2026-02-18T16:40:42Z

watt/system.rs

+        };
+
+        // Look for lid switch input device
+        if name.trim() == "Lid Switch" {


let "Lid Switch" = name.trim() else { log; continue; }; if body here

NotAShelf added 6 commits February 17, 2026 09:23

config: deprecate %cpu-usage

3cfbb99

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Id8593744f5c1b4a862ff9d516570a1986a6a6964

docs: update references to %cpu-usage with %cpu-usage-since

0ada459

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ie65f50fae6ca27a115b12c3d04c7032f6a6a6964

config: add $cpu-scaling-maximum and $load-average-* system varia…

dd013c7

…bles Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Idf55162fa903329d659caac19bbe92e86a6a6964

docs: mention new system variables

b7d784d

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I3d5a06963de7b1f485861ce77607d2226a6a6964

config: add $hour-of-day variable for scripting

70be1fa

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Id94264310cd7f9ccde4a7d1edd162e336a6a6964

docs: add $hour-of-day to README

0fea2d2

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ia2809ab13363fa591132631f3956579a6a6a6964

NotAShelf force-pushed the notashelf/push-ltukuzkpltnp branch from 7209726 to 0fea2d2 Compare February 17, 2026 09:04

NotAShelf changed the title ~~various: deprecatae %cpu-usage; add new variables~~ various: deprecate %cpu-usage; add new variables Feb 17, 2026

NotAShelf added 3 commits February 17, 2026 13:09

config: add new variables

f395cc8

Adds `%cpu-core-count`, `?lid-closed`, `$battery-cycles` and `%battery-health` variables. Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I513545b3c54aa7d725423896a295532a6a6a6964

power_supply: scan battery cycle count and health from sysfs

1bc639d

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ic792e0ccffd71f3229c03f91572c72006a6a6964

system: add lid state scanning & battery data aggregation

5152a61

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ifd02e5852dcfff8720ba3f631c47d00d6a6a6964

NotAShelf commented Feb 17, 2026

View reviewed changes

watt/system.rs Show resolved Hide resolved

NotAShelf commented Feb 17, 2026

View reviewed changes

NotAShelf requested a review from RGBCube February 17, 2026 10:14

RGBCube requested changes Feb 17, 2026

View reviewed changes

watt/power_supply.rs Outdated Show resolved Hide resolved

watt/system.rs Outdated Show resolved Hide resolved

watt/system.rs Outdated Show resolved Hide resolved

watt/system.rs Outdated Show resolved Hide resolved

watt/config.rs Show resolved Hide resolved

NotAShelf added 6 commits February 17, 2026 16:38

various: cycle_count -> cycles

b9056fe

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ie8f136a922eebace444a8b40cebd7eac6a6a6964

system: flip battery aggregation ifs; put failure cases first

23429e7

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: If2d8e55a400a7a52903752fb480728056a6a6964

config: fix serde rename placement on LoadAverageSince

2efb67c

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: I99b2146dfff44569fd90436a836692476a6a6964

NotAShelf marked this pull request as ready for review February 17, 2026 13:53

NotAShelf requested a review from RGBCube February 17, 2026 13:53

NotAShelf added 2 commits February 17, 2026 17:45

docs: document new DSL expressions and variables

340eed0

Signed-off-by: NotAShelf <raf@notashelf.dev> Change-Id: Ibaeb1235a15aebe0450f6f056c6a9c966a6a6964

RGBCube requested changes Feb 18, 2026

View reviewed changes


		CpuCoreCount => Number(state.cpus.len() as f64),

		LoadAverageSince { duration } => {

Conversation

NotAShelf commented Feb 17, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

NotAShelf commented Feb 17, 2026 •

edited

Loading