Rlplot

Small application that plots lines that are sent to the application's stdin.

Compile

Rlplot can be built using GNU make ( using gcc internally ) or using Zig ( zig is love ). Use whatever you like.

Using zig for linux

zig build -Doptimize=ReleaseFast

Using zig for windows

zig build -Dtarget=x86_64-windows-gnu -Doptimize=ReleaseSmall

Using make for linux ( using gcc internally )

make

Using make for windows ( but using zig internally... )

make windows

Using make for web

EMSCRIPTEN="path to emscripten" make www/index.wasm

Install

When built, rlplot is only one file and you can install it using install command. Here I'm installing it to /usr/bin directory, but this can be any other directory...

sudo install zig-out/bin/rlplot /usr/bin/rlplot

Clean

make clean

Uninstall

sudo rm /usr/bin/rlplot

Run

rlplot is designed in such a way that it plays nicely with other unix tools. You can just pipe the output of your program to rlplot and rlplot will do it's best to plot your data.

Examples

I think that more or less all the examples listed on ttyplot examples should work with rlplot ( just replace ttyplot with rlplot. ) But here are some more examples:

Plot the first 8 Fibonacci numbers from user input

rlplot
1 1 2 3 5 8 13 21

Nice Plot

# Plot numbers from 1 to 69
seq 69 | rlplot;

Square(Nice) Plot

# Plot numbers from 1 to 69
python -c "[print(x*x) for x in range(69)]" | rlplot;

Plot from data that is streamed to an UDP socket

nc -ulkp 42069 | rlplot;

Plot from data that is streamed to a TCP socket

nc -lkp 42069 | rlplot;

Plot random data

# This will most likely crash
cat /dev/random | rlplot;

Plot the temperature of core 0 on your CPU.

#Plot temeratur of core 0 on your cpu.

# Step 0: Get sensors value
sensors;
# output:
#...
#Package id 0:  +52.0°C  (high = +105.0°C, crit = +105.0°C)
#Core 0:        +52.0°C  (high = +105.0°C, crit = +105.0°C)
#Core 1:        +52.0°C  (high = +105.0°C, crit = +105.0°C)
#...

# Step 1: Grep Core 0:
sensors | grep 'Core 0';
#output:
#Core 0:        +52.0°C  (high = +105.0°C, crit = +105.0°C)

# Step 2: Get only temperature:
sensors | grep 'Core 0' | awk -p '{print $3}';
# output:
# +52.0°C 

# Step 3: Create a loop and pipe the value of core temperature to rlplot every 0.1 sec.
while :; do echo $(sensors | grep 'Core 0' | awk -p '{print $3}'); sleep 0.1; done | rlplot
# rlplot should not care about nonnumeric symbols So input `+52.0°C` should be fine.

Plot temperature of all CPU core 0.

while :; do echo $(sensors | grep 'Core' | awk -p '{print substr($3, 1, 4) ";" $2}'); sleep 0.1; done | rlplot
# substr is needed because "+52.0C;1" would be recogined as:
# Add 52 to group 0
# Add 1 to group 1
# substr transforms "+52.0C" to "52.0", so one awk will output lines like "52.0;0"

Plot ram usage.

while :; do echo $(free | grep Mem | awk -p '{print $3/1024}'); sleep 0.01; done | ./zig-out/bin/rlplot

UDP client in python

Write an UDP client in python:

import socket

client_socket = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
y_value = 69 # This can be any float32 value
group_id = 420 # This can be any int32 value
port_number = 42069 # This is a port number and it MUST be 42069
client_socket.sendto(f"{y_value};{group_id}".encode(), ("localhost", port_number))

Start rlplot that listens to UDP port 42069:

nc -ulkp 42069 | rlplot;

Controls

Right mouse button + Move mouse - Change offset
Mouse wheel - Change zoom
X + Mouse Wheel - Change zoom only in X axis
Y + Mouse Wheel - Change zoom only in Y axis
T - Add test points
C - Clear all points
R - Reset offset and zoom to (0, 0) and (1, 1)
D - Toggle quad tree debug view.

Todo

~~Make drawing lines use buffers ( Don't use DrawLineStrip function by raylib. ) Maybe use DrawMesh? It's ok for plots with ~1'000'000 points, but I want more!~~
- Implemented this now. For every line, 2 triangles are created. Old points are put in buffers and are drawn like that. Plotter can now handle 30'000'000 points, easy.
When having many points ( 30'000'000 ), a few probles ocure:
- Distant points start being rounded up/down to the closest float. It doesn't look right.
- ~~When zoomed out a lot. It becomes quite slow. ( I guess there is a lot of drawing of the same pixel.. )~~
  - Maybe combine few lines that are close when zoomed out... ( how to detect this ? )
  - This is partly fixed for plots where x values are sorted.
  - ~~Problem still remains if x values aren't sorted.~~
    - This is now solved by finding intervals in which numbers are sorted one way or the other.
    - But still there is a worst case when every line is in different interval, and it will cause it to once aging be slow.
- Maybe use geometry shader ( don't generate triangles on cpu. )
- ~~Gpu memory usage will be lower. Current gpu memory usage:~~
  - ~~(N lines)(2 triangles per line)(3vertices per triangle)((3 floats for position) + (3 float for tangents))*(4 bytes per float)~~
  - ~~If N = 64'000'000, gpu usage will be ~9GB. This seems high...~~
  - This is partly fixed. If plot values are sequential gpu memory usage can be constant with regard to number of points.
  - ~~Problem still remains if x values aren't sorted.~~
    - ~~This is now solved by using quad tree structure for storing data points.~~
    - ~~Still there is work to be done to make quad tree structure closer to optimal.~~
    - This is now solved by finding intervals in which numbers are sorted one way or the other.
~~I'm not happy with the thickness of the line when zooming in and out.~~
- ~~It's not that bad, but it's inconsistent.~~
- Made is consistent. And now it's smooth af.
~~Quad tree rectangles are not inside one another, bounds of the outer quad are smaller than those of the inner quad. Fix this...~~
- I deleted everything that had anything to do with quad trees so this is not a problem anymore.
~~Text looks like shit... I don't know how to fix it...~~
- Text doesn't look like shit any more. I found a way to fix it.
~~Values on x,y axis should be on each horizontal and vertical line. ( Not in corners. )~~
- Did this and it looks awesome.
Colors should be configurable. Black background is the best background, but maybe there will be a need for a white background.
- This will require having a configuration file ( Maybe )
- Or parse tty codes for changing colors... hmmm ( could be cool )
  - Does ttyplot do this ??
Add something to plot points. ( scatter plot )
- This will most likely require the use of quad tree, once again..
Add something for testing the UI.
- I want to record my actions and that play that back to see if something will segfault...
- This will require, I guess some kind of rework of input handling.
  - A structure will have to be introduced that stores two function pointers. One predicate and one action. Each frame call that predicate and if true call action.
- I saw pull request on raylib for something like that. But Ray answered that he has to look at the API more closely.
- ~~Maybe create something that does not depend on glfw and can be tested on headless servers. This would enable me to run those tests on github ci.~~
  - Implemented this. On its own, this feature is useful, already found 1 double free.
  - This is more or less now fuzz testing. I like it.
Nicer UI for setting color of a line, if it's show, maybe to export data to file or stdout.
Stack panel improvements
- make it more general. So that it accepts any kind of element, not just button
- add like a scroll bar on the left size of a stack
Zig build doesn't build tools/font_export.c... Make zig build that also, else default_font.h can't be created.

Screenshot

License

MIT

frisc98 / rlplot Goto Github PK

rlplot's Introduction