feat(fx): wipe from top/bottom

.clippy.toml

@@ -12,6 +12,7 @@ allowed-duplicate-crates = [
     "hyper",
     "itertools",
     "mio",
+    "nu-ansi-term",
     "pbkdf2",
     "regex-automata",
     "regex-syntax",

.envrc.example

@@ -1,6 +1,9 @@
 #!/bin/bash
 
+export RUST_BACKTRACE=full
 export RUST_LOG=none,kty=debug
+# export TRACING_TREE=none,kty=trace
+
 export LOCAL_REGISTRY="kty:5432"
 
 # export GHCR_USER="me"

Cargo.toml

@@ -96,7 +96,7 @@ warp = "0.3.7"
 workspace = true
 
 [profile.dev]
-debug = 0
+debug = 1
 split-debuginfo = "unpacked"
 
 [profile.release]

DEVELOPMENT.md

@@ -22,7 +22,8 @@ git rebase -ff branch-name && git push origin branch-name
 
 On PR, CI produces a darwin-arm64 binary and helm chart. Click on any step from
 the PR and then `Summary` on the left sidebar to see the uploaded artifacts.
-Docker images are not currently build/uploaded on PR runs.
+Docker images are built to verify that the `linux-amd64` binary can be built but
+they are not pushed or uploaded anywhere.
 
 ## Environment
 
@@ -60,6 +61,8 @@ The global debug level can be overly noisy. Instead of doing `-vvvv`, try:
 RUST_LOG=none,kty=debug
 ```
 
+If you'd like to see backtraces on panic, set `RUST_BACKTRACE`.
+
 ### Tracing Tree
 
 It can be a little difficult to reason about how events filter through the

TODO.md

@@ -29,38 +29,18 @@
 ## Server
 
 - Move over to axum for health instead of warp.
+- Terminate the session (but not the server) on panic.
 
 ## TUI
 
 - There needs to be some way to pre-flight permissions for a component so that
   an error is shown instead of letting the component fail.
 
-- Is it possible to get the kubelet logs?
-
-  - `kubectl get --raw "/api/v1/nodes/node-1.example/proxy/logs/` works as
-    `NodeQueryLog` which is beta in 1.30. It is a little weird though, k3s at
-    least returns html? And it doesn't contain the kubelet logs?
-
-- Use SSH forwarding to get into the nodes.
-
-  - Does it make sense to do the `nsenter` trick for some use cases? This
-    requires privileged mode to work.
-
-  - This is waiting on the next release of russh as `handle.open_channel_agent`
-    just landed.
-
-- There's some kind of lag happening when scrolling aggressively (aka, holding
-  down a cursor). It goes fine for ~10 items and then has a hitch in the
-  rendering.
-
 - Terminal resizing isn't wired up for the dev dashboard.
 
 - The way that layers work would be better served by something with ndarray. In
   particular, calculating what the area of a widget would be is ugly.
 
-- Add an error screen if the ready channel is closed with an error. See
-  widget/pod.rs.
-
 - Add routing back in.
 
 - Calculate visibility via areas + zindex to understand what needs to be
@@ -73,20 +53,32 @@
 - Dashboard as a struct doesn't really make sense anymore, it should likely be
   converted over to a simple function.
 
-- Move YAML over to viewport. Should viewport be doing syntax highlighting by
-  default? How do we do a viewport over a set of lines that require history to
-  do highlighting?
-
-- There's a bug somewhere in `log_stream`. My k3d cluster restarted and while I
-  could get all the logs, the stream wouldn't keep running - it'd terminate
-  immediately. `stern` seemed to be working fine. Recreating the cluster causedx
-  everything to work again.
-
 - Move over to something like
   [ratatui-textarea](https://github.com/rhysd/tui-textarea) for the inputs.
 
 - Add an editor to allow for creation of resources (should it just be pods?).
 
+- Animate the egress/ingress tunnel lines. In particular, it would be nice to
+  watch `Active` fade to `Listening` after a request goes through.
+
+### Nodes
+
+- Is it possible to get the kubelet logs?
+
+  - `kubectl get --raw "/api/v1/nodes/node-1.example/proxy/logs/` works as
+    `NodeQueryLog` which is beta in 1.30. It is a little weird though, k3s at
+    least returns html? And it doesn't contain the kubelet logs?
+
+- Use SSH forwarding to get into the nodes.
+
+  - Does it make sense to do the `nsenter` trick for some use cases? This
+    requires privileged mode to work.
+
+  - This is waiting on the next release of russh as `handle.open_channel_agent`
+    just landed.
+
+### Pods
+
 - Rethink the pod detail view. The yaml doesn't feel like the most important
   thing to look at, neither do logs. Shell feels the closest, but that's not
   great either. Maybe something like `kubectl describe`? It could be multi-panel
@@ -95,11 +87,15 @@
   _everything_ into memory (but doesn't try to render the entire thing every
   100ms).
 
-- Animate the egress/ingress tunnel lines. In particular, it would be nice to
-  watch `Active` fade to `Listening` after a request goes through.
+- Add `graph` to the pod view.
+
+### Table
+
+- Highlight filter matches in the list.
 
 ## SFTP
 
+- Get a watchman style demo working.
 - Allow globs in file paths, eg `/*/nginx**/etc/passwd`.
 - Return an error that is nicer than "no files found" when a container doesn't
   have cat/ls.
@@ -108,7 +104,6 @@
 ## SSH Functionality
 
 - Allow `ssh` directly into a pod without starting the dashboard.
-
 - How can `ssh` be shutdown without causing it to have a 1 as an exit code?
 
 ## Ingress Tunnel
@@ -126,9 +121,6 @@
 
 ## Build
 
-- Figure out why `git cliff` goes `0.2.0` -> `0.2.1` -> `0.3.0` instead of
-  `0.2.2`.
-
 - Add integration test for `kty resources install`.
 
 - Add integration test for `helm install`.
@@ -137,11 +129,6 @@
   be great for the development experience. Is there a way to have defaults but
   override them? Maybe with a dev instance of auth0?
 
-## Deployment
-
-- Add `kustomize` to allow for an easier getting started.
-- Make `helm install` work if someone's checked the repo out.
-
 ## Misc
 
 - Move to [bon](https://docs.rs/bon/latest/bon/) instead of `derive_builder`.

src/fx.rs

@@ -1,8 +1,10 @@
+pub mod wipe;
+
 use bon::{builder, Builder};
 use eyre::Result;
-use ndarray::{s, ArrayViewMut2};
 use ratatui::{buffer::Buffer, layout::Rect, Frame};
-use tachyonfx::{CellFilter, CellIterator, Duration, Effect, EffectRenderer, EffectTimer, Shader};
+use tachyonfx::{Effect, EffectRenderer, EffectTimer, Shader};
+pub use wipe::Start;
 
 use crate::{
     dashboard::render_interval,
@@ -46,124 +48,17 @@ impl Widget for Animated {
     }
 }
 
-#[derive(Clone, Default)]
-pub enum Start {
-    Left,
-    #[default]
-    Right,
-}
-
-/// Wipes content in from the start edge, uses `previous` as the source and then
-/// the current buffer as the destination.
-#[derive(Builder, Clone)]
-pub struct Wipe {
-    #[builder(into)]
-    timer: EffectTimer,
-    #[builder(default)]
-    start: Start,
-    previous: Buffer,
-    #[builder(default)]
-    done: bool,
-}
-
 #[builder]
-pub fn horizontal_wipe<T: Into<EffectTimer>>(
+pub fn wipe<T: Into<EffectTimer>>(
     timer: T,
     buffer: Buffer,
     #[builder(default)] start: Start,
 ) -> Effect {
     Effect::new(
-        Wipe::builder()
+        wipe::Wipe::builder()
             .timer(timer.into())
             .previous(buffer)
             .start(start)
             .build(),
     )
 }
-
-// This assumes that the area from the original buffer to the new one doesn't
-// change. If it is unable to create a slice correctly, it'll just give up.
-impl Shader for Wipe {
-    fn name(&self) -> &'static str {
-        "slider"
-    }
-
-    #[allow(
-        clippy::cast_possible_truncation,
-        clippy::cast_sign_loss,
-        clippy::cast_precision_loss
-    )]
-    fn process(&mut self, duration: Duration, buf: &mut Buffer, area: Rect) -> Option<Duration> {
-        let (overflow, alpha) = self
-            .timer_mut()
-            .map_or((None, 1.0), |t| (t.process(duration), t.alpha()));
-
-        let x = area.x as usize;
-        let y = area.y as usize;
-        let width = area.width as usize;
-        let height = area.height as usize;
-        let buffer_width = self.previous.area.width as usize;
-        let buffer_height = self.previous.area.height as usize;
-
-        let window = ((width as f32 * alpha).round() as usize).clamp(0, width);
-
-        if window == width {
-            self.done = true;
-        }
-
-        let Ok(previous) =
-            ArrayViewMut2::from_shape((buffer_height, buffer_width), &mut self.previous.content)
-        else {
-            tracing::debug!(area = ?area, "unable to create view from previous buffer");
-
-            self.done = true;
-            return overflow;
-        };
-
-        let Ok(mut next) =
-            ArrayViewMut2::from_shape((buffer_height, buffer_width), &mut buf.content)
-        else {
-            tracing::debug!(area = ?area, "unable to create view from next buffer");
-
-            self.done = true;
-            return overflow;
-        };
-
-        let slice = match self.start {
-            Start::Left => s![y..height, x + window..x + width],
-            Start::Right => s![y..y + height, x..x + width - window],
-        };
-
-        let previous_section = previous.slice(slice);
-
-        next.slice_mut(slice).assign(&previous_section);
-
-        overflow
-    }
-
-    fn execute(&mut self, _: f32, _: Rect, _: CellIterator<'_>) {}
-
-    fn done(&self) -> bool {
-        self.done
-    }
-
-    fn clone_box(&self) -> Box<dyn Shader> {
-        Box::new(self.clone())
-    }
-
-    fn area(&self) -> Option<Rect> {
-        None
-    }
-
-    fn set_area(&mut self, _area: Rect) {}
-
-    fn set_cell_selection(&mut self, _: CellFilter) {}
-
-    fn timer_mut(&mut self) -> Option<&mut EffectTimer> {
-        Some(&mut self.timer)
-    }
-
-    fn timer(&self) -> Option<EffectTimer> {
-        Some(self.timer)
-    }
-}