* refactor: more memory collection cleanup
* more cleanup
* clean up data_harvester, remove heim sensor flag
Separate out most individual components to separate functions. Also
remove Linux's usage of heim's sensors feature, since I wasn't using it
apparently.
* clean up GPU section
* fix cond
* fix feature flags
* more cleanup
* even more cleanup
* Upgrade clap to 3.2.2 to allow future fix warnings
* cargo fmt fix
* Replaced deprecated `value_of` and `is_present`, builds with no warnings, clap 3.2.2
* cargo fmt
* updated according to comments. builds fine, cargo test fine
* Match some versions
* Update Cargo.lock
* Fix typo, mb
---------
Co-authored-by: Clement Tsang <34804052+ClementTsang@users.noreply.github.com>
Denotes both usage and usage percentage. This also redoes the calculation for percentage to be based on the sum of avail + used, rather than on total, as otherwise we get potentially confusing percentages.
Pulls in the tui-rs grid logic so I can implement custom braille painting logic. We basically "flatten" the layering logic into a single layer by replacing resetting cells if the colour is different. This avoids the multiple allocations if we used multiple layers as intended with tui-rs.
This gives us chart results similar to the current stable version, but with a flamegraph similar to the current master branch.
* other: group all dataset draws in a time chart
We used to draw each data set separately as a new canvas. Now, in one
canvas, we draw all datasets.
Note that this changes how dataset lines are drawn - rather than
drawing one on top of another, it now draws kinda all at once. This
effect is *kinda* a bit better IMO, but it might also look a bit
more cluttered.
* other: optimize truncate_text
Flamegraphs showed that this area seems to be a bit heavy at times with
some inefficient use of iterators and collection. This change should
hopefully optimize this a bit by reducing some collections or
reallocations.
There can also be some further optimizations with less allocations from
callers.
* Reduce some redundant draws
* other: don't use manual map for color name mapping
I actually don't know why I was doing it like that before.
This commit removes the phf crate, as it's not needed anymore.
* update test
* refactor: move to new data table implementation
* more work towards refactor
* move disk and temp over, fix longstanding bug with disk and temp if removing the last value and selected
* work towards porting over CPU
work towards porting over CPU
fix typo
partially port over cpu, fix some potentially inefficient concat_string calls
more work towards cpu widget migration
some refactoring
* sortable data
sortable data
more refactoring
some sort refactoring
more refactoringgggg
column refactoring
renaming and reorganizing
more refactoring regarding column logic
add sort arrows again
* move over sort menu
* port over process
port over process
precommit
temp
temp two, remember to squash
work
fix broken ltr calculation and CPU hiding
add back row styling
temp
fix a bunch of issues, get proc working
more fixes around click
fix frozen issues
* fix dd process killing
* revert some of the persistent config changes from #257
* fix colouring for trees
* fix missing entries in tree
* keep columns if there is no data
* add and remove tests
* Fix ellipsis
* deps: update sysinfo to 0.26.2
This dependency update has some nice things in store for us:
- MacOS M1 temperature support
- Bevy of bug fixes
* update documentation
* some fixes
* ci: spring cleaning of completions autogen
This commit changes a few things/cleans up stuff:
- Completion and manpage generation now drops the files off in
`./target/tmp/bottom` rather than arbitrarily in the build directory.
This was originally done because I was lazy and just needed it to work
in CI, but it's kinda gross if you want to build the manpages in your
own directory.
- CI was updated to handle this.
- Only run if the `BTM_GENERATE` env var is actually non-empty.
* docs: update for manpage/completion gen
* ci: auto delete autogen comp/manpage dir
* ci: fix incorrect mv for autogen
The mv was too late, should be earlier in the workflow.
* ci: specify shell in autogen delete
* docs: more updates to manpage/comp docs
* ci: unify env vars
* ci: skip autogen on build-msi
* freebsd clippy
* add arc support
* Code Review: moved runtime cfg checks to compile time and formatting
* remove compile platform checks
* add zfs feature flag to get_arc_data
* WIP FreeBSD support
* Implement get_cpu_data_list for FreeBSD
* Implement disks for FreeBSD
It doesn't work though as sysinfo doesn't make the device name available.
* Use libxo to read process cpu info on FreeBSD
* Populate get_io_usage with libxo too
Actual I/O stats still aren't populated though as there's not an
easy source for them.
* Share more processes code between macos and freebsd
* Extract function for deserializing libxo output on FreeBSD
* Implement filtering of disks in FreeBSD
* Clean up memory data collection
* Update module docs
This consolidates all the time graph drawing to one main location, as well
as some small improvements. This is helpful in that I don't have to
reimplement the same thing across three locations if I have to make one
change that in theory should affect them all. In particular, the CPU
graph, memory graph, and network graph are all now using the same,
generic implementation for drawing, which we call (for now) a component.
Note this only affects drawing - it accepts some parameters affecting style
and labels, as well as data points, and draw similarly to how it used to
before. Widget-specific actions, or things affecting widget state,
should all be handled by the widget-specific code instead. For example,
our current implementation of x-axis autohide is still controlled by the
widget, not the component, even if some of the code is shared. Components
are, again, only responsible for drawing (at least for now). For that
matter, the graph component does not have mutable access to any form of
state outside of tui-rs' `Frame`. Note this *might* change in the
future, where we might give the component state.
Note that while functionally, the graph behaviour for now is basically
the same, a few changes were made internally other than the move to
components. The big change is that rather than using tui-rs' `Chart`
for the underlying drawing, we now use a tweaked custom `TimeChart`
tui-rs widget, which also handles all interpolation steps and some extra
customization. Personally, I don't like having to deviate from the
library's implementation, but this gives us more flexibility and allows
greater control. For example, this allows me to move away from the old
hacks required to do interpolation (where I had to mutate the existing
list to avoid having to reallocate an extra vector just to insert one
extra interpolated point). I can also finally allow customizable
legends (which will be added in the future).
Adds the asset for the manpage to cargo deb config. Also moves the generated manpage file to a .1.gz file. Also, moves back to a build script since that was causing some issues for the automatic Cargo.toml fields detection for manpage and completion generation.
To prevent compilation from happening every time, and only in CI, we use an env var to avoid generation steps.
Adds manpage generation to the build process, as well as following the xtask concept of adding additional build scripts that only need to run on deploy/nightly as opposed to `build.rs`. Note this doesn't follow the recommended method of using workplaces because I don't really want to shift the entire repo structure just for this.
More on xtask: https://github.com/matklad/cargo-xtask
Due to a missing check, you could resize the window to a width that was too small, and it would trigger an endless while-loop for any table while trying to redistribute remaining space. This has been rectified with an explicit check, as well as a smarter method of redistributing remaining space borrowed from the rewrite.
This also adds explicit width checks for widgets that have borders; if the width is <2, before, it would panic.
Note that the rewrite I have kinda fixes all these issues already, so I don't want to invest too hard into this, but this should be fine as a patch for now.
Also note that minimal heights don't seem to be causing any issues, it just seems to be minimal widths.
Bumps up some dependencies and removes chrono, switching to the time crate instead.
One of side-effects of this change is that local time seems to not work (?)... so all logs are now in UTC. Oh well, this doesn't affect general user behaviour so I'm fine with it.
This is just a temp change, I wanted to remove it just for clarity's
sake among dependencies, and will probably add it back in the future.
For now I'll just stick to std's beef.
Refactor to split up data collection by OS and/or the backing library. The goal is to make it easier to work with and add new OS support, as opposed to how it was prior where we stored OS-independent implementations all in the same file.
Did not update crossterm (and tui-rs) since it seems to have resulted in a massive CPU usage increase. Also fix minor clippy error with a duplicated to_string call.
For the process widget, we now sum the resource usage of the child processes on the parent entry when collapsing in tree mode.
Note that if you search to filter, and collapse, it will not sum the pruned values (values that cannot be seen). This is partly because I'm a bit lazy, and partly because I think this behaviour makes sense.
For example, let's say I search for a process with 4 child processes "AA, AB, BA, BB", with CPU usage 0.1, 0.2, 0.3, 0.4 respectively. Assume the parent process has 0 usage.
- Without filter, it sums to 1.0
- With a filter on A, it would sum to just 0.3
- With a filter on AA, it would sum to 0.1
I think this is fine because I'm treating this as summing any child that is still *visible* somehow. Summing unseen values would probably be weird as it would look like it's not adding up.
Further note that if you had, say, a child "CC" with a usage of, say, 2.0, and its parent of "AB", and you searched for CC in our above example, you would get a sum of 2.2. This is because AB is still visible by the fact that CC was the searched process, and AB must still exist (albeit faded out) in the tree hierarchy, and as such will still be displayed.
Rewrite of the y-axis labeling and scaling for the network widget, along with more customization. This still has one step to be optimized (cache results so we don't have to recalculate the legend each time), but will be done in another PR for sake of this one being too large already.
Furthermore, this change adds linear interpolation at the 0 point in the case a data point shoots too far back - this seems to have lead to ugly gaps to the left of graphs in some cases, because the left hand limit was not big enough for the data point. We address this by grabbing values just outside the time range and linearly interpolating at the leftmost limit. This affects all graph widgets (CPU, mem, network).
This can be optimized, and will hopefully be prior to release in a separate change.
Adds users into the process widget (for Unix-based systems). This shows only in non-grouped modes, similar to state. Search is also supported.
In addition, a quick fix to prevent users from being in grouped mode when they tried to enter tree mode while grouped.
Fixes a bug where you could make the sorting arrow disappear in the mem column if you did:
1. Go to proc widget
2. Switch to memory values from %
3. Press `m`
Refactors to use only heim for Linux builds. This is now much easier to do since the 0.1 version of heim works fine for ARM. This is ideal since having to rely on two separate sources of data isn't the greatest if we can avoid it.
Sysinfo is still required for macOS and Windows, though. Temperature sensors do not work for those from heim, and for some reason, networks also don't work on Windows with heim...?
My personal CPU core calculation is also currently Linux-only, and as such, I'll still rely on sysinfo for Windows and macOS for now.
This isn't really a big optimization or anything btw. Just something I wanted to try.
Making some small changes that would hopefully improve performance a bit.
- Remove redundant string generations for CPU data conversion
- Switch to fnv for PID hashmap and hashsets
- Use buffered reading to avoid having to store too many lines as strings
Fixes some performance regressions and forgotten cleanup.
Changes to attempt to improve performance to match 0.4.x:
- Remove `trace!` and `--debug` for now. These were a significant hog. Removing this dropped initial memory usage by about half.
- Add additional cleaning step for `pid_mapping` during process harvesting. This should hopefully improve memory usage as time goes on.
- Slightly change how we do sorting to hopefully be a bit more optimal? This was just an easy change to make that I spotted.
- Fix broken cleaning child thread task.
Adds persistent search settings across runs, by saving to the config file. Each process widget keeps track of it's *own* behaviour. The previous flags/options are now for *global* behaviour.
The following new behaviour is:
- Relevant flags: `--case_sensitive`, `--whole_word`, and `--regex`, will *override* the current widget's default behaviour.
- Relevant options: `case_sensitive`, `whole_word`, and `regex`, will also *override* the current widget's default behaviour.
As per before, if you set, say, `--case_sensitive`and `case_sensitive=true`, the flag always overrides.
Documentation updates will be done in #248.
Refactors tui-rs usage to the new 0.11.0 release. This release also fixes the highlighting bug from #249, and now, expanding a widget no longer overrides the widget title colour.
This commit also introduces #255, but that seems to be easy to bandaid so hopefully it will get fixed soon?
Fix for an index out-of-bounds by resizing to a smaller terminal just after the program got the terminal size, but right before the terminal started drawing.
