Write a program that prints the RGB color of a given pixel on the screen.

The program should read a line from stdin in the format x,y, where x and y are unsigned integers. The pixel (0,0) is the top left pixel in the screen.

The program should then output a line to stdout in the format RRR,GGG,BBB, which represents the color of the pixel at (x,y).

RRR, GGG, and BBB should be floating-point numbers between 0 and 1, OR integers in the range [0, 255]. Leading zeros do not matter, for example: both 074 and 74 are acceptable.

Undefined Behavior

It is okay for the program to exhibit undefined behavior (crash, give incorrect output, etc) in the following conditions:

  • The user inputs a string, signed integer, float, or any other invalid data type
  • The user input is not in the format x,y
  • The specified pixel is off-screen

Notes:

  • The program cannot modify the pixel at the given coordinates.
  • If you have multiple monitors, it does not matter which monitor the pixel is on, as long as the same monitor is used each time the program runs.

The answer with the fewest bytes wins.

  • 22
    Why should RRR, GGG, and BBB be floating points between 0 and 1? Usually they're integers in the range [0,255]. I would suggest to allow both. – Kevin Cruijssen Dec 5 at 13:48
  • 4
    Can we limit our solutions to the window they are being run in, for the benefit of languages that can't access anything on screen outside of that window? – Shaggy Dec 5 at 17:03
  • 7
    If you allow only the 0 — 1 format you should not use RRR, GGG, and BBB in the spec, but R, G, B – sergiol Dec 5 at 17:21
  • 6
    What if we have multiple monitors? – tsh Dec 6 at 2:42
  • 7
    It could be worth to specify that the program can't modify the pixel at the given coordinate. Otherwise it might be more efficient to black out the entire screen and return 0, 0, 0, defeating the purpose of the challenge. – maxb Dec 6 at 10:15

11 Answers 11

Java 10 (lambda function), 105 75 bytes

x->y->(new java.awt.Robot().getPixelColor(x,y)+"").replaceAll("[^\\d,]","")

A function taking two integer parameters and returning a String RRR,GGG,BBB, where the colors are integers in the range [0, 255].

-30 bytes thanks to @LukeStevens by using java.awt.Color's default toString() output and modifying it so only the digits and commas remain.

Explanation:

x->y->                      // Method with two integer parameters and String return-type
  (new java.awt.Robot()     //  Create a AWT-Robot instance
    .getPixelColor(x,y)     //  Get the pixel AWT-Color at position x,y
    +"")                    //  Call the default toString() of the AWT-Color
                            //   i.e. "java.awt.Color[r=213,g=230,b=245]"
  .replaceAll("[^\\d,]","") //  Remove everything except for digits and commas

NOTE: The default toString() implementation of the java.awt.Color has always been the same across JVM versions as far as I know, but can potentially change in the future. I tested it in both Java 8 and Java 10 and it returned "java.awt.Color[r=#,g=#,b=#]".


But, since the challenge states:

  • Full program
  • Taking input in the format x,y from STDIN
  • Output in the format R,G,B to STDOUT
  • Have the R, G, B as floating points in the range [0.0, 1.0]

The code becomes A LOT more verbose:

Java 10 (full program), 266 bytes

interface M{static void main(String[]a)throws Exception{var s=new java.util.Scanner(System.in).next().split(",");var c=new java.awt.Robot().getPixelColor(new Short(s[0]),new Short(s[1]));System.out.print(c.getRed()/255d+","+c.getGreen()/255d+","+c.getBlue()/255d);}}

Explanation:

interface M{          // Class
  static void main(String[]a)
                      //  Mandatory main method
    throws Exception{ //    Mandatory throws clause for `new Robot()`
  var s=new java.util.Scanner(System.in)
                      //   Create a Scanner for STDIN
         .next()      //   Get the String user input
         .split(","); //   Split it on "," and save it as String-array in `s`
  var c=new java.awt.Robot()
                      //   Create a AWT-Robot instance
         .getPixelColor(
                      //   And get the pixel AWT-Color at position:
           new Short( //    Convert String to Short (and implicitly to int):
            s[0]),    //     x-coordinate input by user from String-array `s`
           new Short( //    Convert String to Short (and implicitly to int):
            s[1]));   //     y-coordinate input by user from String-array `s`
  System.out.print(   //   Print to STDOUT:
   c.getRed()         //    The red part of RGB as integer in the range [0,255]
    /255d             //     Converted to a floating point in the range [0.0, 1.0]
   +","               //    Appended with a comma delimiter
   +c.getGreen()      //    Appended with the green part of RGB as integer in the range [0,255]
     /255d            //     Converted to a floating point in the range [0.0, 1.0]
   +","               //    Appended with a comma delimiter
   +c.getBlue()       //    Appended with the blue part of RGB as integer in the range [0,255]
     /255d);}}        //     Converted to a floating point in the range [0.0, 1.0]
  • 2
    For the first, loose-rules answer, you can make use of Colors String conversion and simply have x->y->(new java.awt.Robot().getPixelColor(x,y)+"").replaceAll("[^\\d,]","") for 75 bytes – Luke Stevens Dec 5 at 14:31
  • @LukeStevens Smart use of the default toString() of java.awt.Color. Thanks! – Kevin Cruijssen Dec 5 at 14:39
  • Good job guys, you beat me to it: I was busy implementing the toString() way as well... ;-) Except I had to download an IDE to do all that: I only use TIO now :( – Olivier Grégoire Dec 5 at 14:45
  • 3
    Note that the Javadoc is very clear that Color::toString is implementation-discretionary. So it might be good to specify which JVM was used for the tests (it works as is with OpenJDK and Oracle JDK). – Olivier Grégoire Dec 5 at 14:50
  • 1
    Finally, you should replace " " with "," to comply with the strict RRR,GGG,BBB format (not RRR GGG BBB). – Olivier Grégoire Dec 5 at 14:57

6502 machine code (C64), 280 260 bytes

00 C0 A2 00 20 CF FF C9 0D 9D 02 C1 F0 03 E8 D0 F3 20 D2 FF A5 7A 85 FB A5 7B
85 FC A9 01 85 7A A9 C1 85 7B 20 73 00 20 6B A9 A5 14 85 FD A5 15 85 FE 20 73
00 20 6B A9 A5 FB 85 7A A5 FC 85 7B A5 14 4A 4A 4A AA 20 F0 E9 A5 FD 46 FE 6A
4A 4A A8 A5 14 29 07 AA A9 00 38 6A CA 10 FC 85 FE B1 D1 0A 26 FC 0A 26 FC 0A
26 FC 85 FB A5 D2 29 03 09 D8 85 D2 B1 D1 85 02 A9 02 18 2D 18 D0 0A 0A 65 FC
29 0F 09 D0 85 FC A9 33 85 01 A5 FD 29 07 A8 B1 FB A2 37 86 01 25 FE D0 05 AD
21 D0 85 02 A6 D6 20 F0 E9 A5 02 29 0F AA BC D2 C0 20 BC C0 BC E2 C0 20 B7 C0
BC F2 C0 A9 2C 20 D2 FF 98 0A A9 30 90 02 A9 31 20 D2 FF A9 2E 20 D2 FF 98 29
7F 4C D2 FF 30 B0 35 35 36 33 32 39 36 33 38 33 35 37 34 37 30 B0 32 38 32 37
32 39 33 32 34 33 35 B0 34 37 30 B0 32 38 36 33 36 34 32 30 34 33 35 36 39 37

I expected this to be possible in a lot fewer bytes, but unfortunately... well, since I finished it, posting it now anyways. At least, the restrictive format helped with one thing: Something similar to stdin (input from a current device) only exists on the C64 in text mode, because the OS only supports this mode -- so no need to consider other modes of the graphics chip.

Note on the output of the color values: the C64 graphics chip doesn't use RGB colors but directly generates a video signal with YUV colors, with a fixed 16 colors palette. I used rounded values of the colodore conversion to RGB with "default" monitor settings here.

-20 bytes: better output routine, encoding the 3 character output per color channel in a single byte.

Regarding the comment: It's theoretically possible to use even multicolor character mode of the VIC with the stock C64 OS, but it requires a custom font that's actually legible with only 4 horizontal pixels of doubled width. Not entirely impossible, but very unlikely. Likewise, extended color mode (or extended background mode, which is the same) could be used with the C64 OS, but requires to reconfigure the graphics chip directly. I opt to ignore all these possibilities in the sense of code golfing here: It's not the standard environment you find on a Commodore 64 with running stock OS. What's possible with stock OS is switching between two builtin fonts (shift + commodore key), the program does account for that.

Online demo

Usage: SYS49152 to start.

Commented disassembly:

         00 C0       .WORD $C000        ; load address
.C:c000  A2 00       LDX #$00           ; loop index for input
.C:c002   .input:
.C:c002  20 CF FF    JSR $FFCF          ; character from input device
.C:c005  C9 0D       CMP #$0D           ; compare with enter
.C:c007  9D 16 C1    STA .buf,X         ; store to buffer
.C:c00a  F0 03       BEQ .parse         ; was enter -> start parsing
.C:c00c  E8          INX                ; next character
.C:c00d  D0 F3       BNE .input         ; and repeat input loop
.C:c00f   .parse:
.C:c00f  20 D2 FF    JSR $FFD2          ; output the enter character
.C:c012  A5 7A       LDA $7A            ; save pointer of BASIC parser
.C:c014  85 FB       STA $FB
.C:c016  A5 7B       LDA $7B
.C:c018  85 FC       STA $FC
.C:c01a  A9 15       LDA #$15           ; set pointer of BASIC parser to
.C:c01c  85 7A       STA $7A            ; buffer-1
.C:c01e  A9 C1       LDA #$C1
.C:c020  85 7B       STA $7B
.C:c022  20 73 00    JSR $0073          ; get next character
.C:c025  20 6B A9    JSR $A96B          ; BASIC routine to parse number
.C:c028  A5 14       LDA $14            ; lowbyte of parsed number to $fd
.C:c02a  85 FD       STA $FD
.C:c02c  A5 15       LDA $15            ; highbyte to $fe
.C:c02e  85 FE       STA $FE
.C:c030  20 73 00    JSR $0073          ; get next character
.C:c033  20 6B A9    JSR $A96B          ; parse as number ...
.C:c036  A5 FB       LDA $FB            ; restore pointer of BASIC parser
.C:c038  85 7A       STA $7A
.C:c03a  A5 FC       LDA $FC
.C:c03c  85 7B       STA $7B
.C:c03e  A5 14       LDA $14            ; load y coordinate
.C:c040  4A          LSR A              ; divide by 8 for character row
.C:c041  4A          LSR A
.C:c042  4A          LSR A
.C:c043  AA          TAX                ; -> to X
.C:c044  20 F0 E9    JSR $E9F0          ; set pointer to character row
.C:c047  A5 FD       LDA $FD            ; divide x coordinate by 8
.C:c049  46 FE       LSR $FE
.C:c04b  6A          ROR A
.C:c04c  4A          LSR A
.C:c04d  4A          LSR A
.C:c04e  A8          TAY                ; -> to Y
.C:c04f  A5 14       LDA $14            ; load y coordinate
.C:c051  29 07       AND #$07           ; mask pixel position in character
.C:c053  AA          TAX                ; -> to X
.C:c054  A9 00       LDA #$00           ; initialize pixel mask to 0
.C:c056  38          SEC                ; set carry for bit to shift in
.C:c057   .bitnum:
.C:c057  6A          ROR A              ; shift bit in mask
.C:c058  CA          DEX                ; and repeat until
.C:c059  10 FC       BPL .bitnum        ; in correct position
.C:c05b  85 FE       STA $FE            ; store pixel mask to $fe
.C:c05d  B1 D1       LDA ($D1),Y        ; load character code
.C:c05f  0A          ASL A              ; multiply by 8
.C:c060  26 FC       ROL $FC
.C:c062  0A          ASL A
.C:c063  26 FC       ROL $FC
.C:c065  0A          ASL A
.C:c066  26 FC       ROL $FC
.C:c068  85 FB       STA $FB            ; and store to $fb/$fc
.C:c06a  A5 D2       LDA $D2            ; move pointer to position in color RAM
.C:c06c  29 03       AND #$03
.C:c06e  09 D8       ORA #$D8
.C:c070  85 D2       STA $D2
.C:c072  B1 D1       LDA ($D1),Y        ; load color of character
.C:c074  85 02       STA $02            ; and store to $2
.C:c076  A9 02       LDA #$02           ; check which charset is active
.C:c078  18          CLC
.C:c079  2D 18 D0    AND $D018
.C:c07c  0A          ASL A              ; and calculate offset
.C:c07d  0A          ASL A
.C:c07e  65 FC       ADC $FC            ; add to (character code * 8)
.C:c080  29 0F       AND #$0F
.C:c082  09 D0       ORA #$D0           ; and add offset to character ROM
.C:c084  85 FC       STA $FC
.C:c086  A9 33       LDA #$33           ; bank in character ROM
.C:c088  85 01       STA $01
.C:c08a  A5 FD       LDA $FD            ; load y coordinate
.C:c08c  29 07       AND #$07           ; mask pixel-row number
.C:c08e  A8          TAY
.C:c08f  B1 FB       LDA ($FB),Y        ; load pixel row from character ROM
.C:c091  A2 37       LDX #$37           ; bank out character ROM
.C:c093  86 01       STX $01
.C:c095  25 FE       AND $FE            ; apply pixel mask
.C:c097  D0 05       BNE .pixelcol      ; not 0 -> pixel is set
.C:c099  AD 21 D0    LDA $D021          ; otherwise load background color
.C:c09c  85 02       STA $02            ; and store to $2
.C:c09e   .pixelcol:
.C:c09e  A6 D6       LDX $D6            ; restore screen row pointer for
.C:c0a0  20 F0 E9    JSR $E9F0          ; current cursor position
.C:c0a3  A5 02       LDA $02            ; load color
.C:c0a5  29 0F       AND #$0F           ; mask low nibble (only 16 colors)
.C:c0a7  AA          TAX                ; -> to X
.C:c0a8  BC D2 C0    LDY .red,X         ; load encoded output for red
.C:c0ab  20 BC C0    JSR .out2          ; call output without comma
.C:c0ae  BC E2 C0    LDY .green,X       ; load encoded output for green
.C:c0b1  20 B7 C0    JSR .out1          ; call output with comma
.C:c0b4  BC F2 C0    LDY .blue,X        ; load encoded output for blue
.C:c0b7   .out1:
.C:c0b7  A9 2C       LDA #$2C           ; load ","
.C:c0b9  20 D2 FF    JSR $FFD2          ; and output
.C:c0bc   .out2:
.C:c0bc  98          TYA                ; encoded output to A
.C:c0bd  0A          ASL A              ; shift top bit to carry
.C:c0be  A9 30       LDA #$30           ; load "0"
.C:c0c0  90 02       BCC .firstdig      ; carry clear -> to output
.C:c0c2  A9 31       LDA #$31           ; load "1"
.C:c0c4   .firstdig:
.C:c0c4  20 D2 FF    JSR $FFD2          ; and output
.C:c0c7  A9 2E       LDA #$2E           ; load "."
.C:c0c9  20 D2 FF    JSR $FFD2          ; and output
.C:c0cc  98          TYA                ; encoded output to A
.C:c0cd  29 7F       AND #$7F           ; mask out top bit
.C:c0cf  4C D2 FF    JMP $FFD2          ; to output and exit
.C:c0d2   .red:                                 ; encoded values for red
.C:c0d2  30 B0 35 35 .BYTE $30,$B0,$35,$35      ; ASCII digit ($30-$39) after
.C:c0d6  36 33 32 39 .BYTE $36,$33,$32,$39      ; decimal point, with bit 7
.C:c0da  36 33 38 33 .BYTE $36,$33,$38,$33      ; indicating 0 or 1 before
.C:c0de  35 37 34 37 .BYTE $35,$37,$34,$37      ; decimal point
.C:c0e2   .green:                               ; encoded values for green
.C:c0e2  30 B0 32 38 .BYTE $30,$B0,$32,$38      ; ...
.C:c0e6  32 37 32 39 .BYTE $32,$37,$32,$39
.C:c0ea  33 32 34 33 .BYTE $33,$32,$34,$33
.C:c0ee  35 B0 34 37 .BYTE $35,$B0,$34,$37
.C:c0f2   .blue:                                ; encoded values for blue
.C:c0f2  30 B0 32 38 .BYTE $30,$B0,$32,$38      ; ...
.C:c0f6  36 33 36 34 .BYTE $36,$33,$36,$34
.C:c0fa  32 30 34 33 .BYTE $32,$30,$34,$33
.C:c0fe  35 36 39 37 .BYTE $35,$36,$39,$37
.C:c102   .buf:                                 ; buffer for input ("stdin")
  • This doesn't seem to support multicolor mode. – nwellnhof Dec 5 at 17:34
  • Of course not, it works with the original OS. Technically, this OS works in any VIC mode, but there will be only garbage on the screen, so it isn't usable. But your comment made me remember "extended color mode", which works quite well with stock OS ... dammit – Felix Palmen Dec 5 at 18:03
  • @nwellnhof added some reasoning about it, I hope this is agreeable. – Felix Palmen Dec 6 at 9:14
  • Makes one wonder if writing this in C would be shorter? – Rogem Dec 6 at 19:15
  • @Rogem could try that with cc65 .. quite possible :) but at least, the compiled code will be larger :) – Felix Palmen Dec 6 at 22:50

TI-BASIC (TI-83/84+), 22 bytes

Since the screen is black and white, we just need to test if the specified pixel is on or off and map that to black or white RGB. Also, the pixels are only accessible through row and column in that order, so that's why the coordinates are reversed.

Prompt X,Y
255
Ans-{Ans,Ans,Ans}pxl-Test(Y,X
  • 1
    Note that this doesn't work on color TI calculators (Like the TI-84 Plus CE). – pizzapants184 Dec 6 at 0:54
  • Are you sure about the byte count? I count 22. – lirtosiast Dec 7 at 7:17

bash, 103 / 86 bytes

With a strict interpretation of the spec (input from STDIN and output on STDOUT are both comma-separated):

read x
import -window root -crop 1x1+${x/,/+} ppm:-|od -dj13|awk 'NR<2{n=2^16;print$2/n","$3/n","$4/n}'

With a looser input format (plus-separated input as a command line argument, space-separated output:

import -window root -crop 1x1+$1 ppm:-|od -dj13|awk 'NR<2{n=2^16;print$2/n,$3/n,$4/n}'

Depends on imagemagick, awk, and coreutils.

Bash on Linux with xserver, 30 bytes

xset dpms force off;echo 0,0,0

Using the idea presented in my comment on the question, this solution should turn off the display output completely, and then echo that the screen is indeed black.

I was also thinking of using xbacklight =0, but that doesn't change the pixel colors.

  • +1 for finding the loophole, but I updated the constraints. – eeze Dec 9 at 2:05

Python 2 + PIL library, 96 91 bytes

import PIL.ImageGrab as i
print','.join('%.1f'%(x/255.)for x in i.grab().getpixel(input()))

Implements the specification literally as requested. Windows only though - doesn't work on Linux, and produces extra output (alpha value) on Mac.

  • from PIL import ImageGrab as ifrom PIL import*, i.grabImageGrab.grab. – Erik the Outgolfer Dec 5 at 14:59
  • @Erik this does not for PIL. PIL only imports the names explicitly listed. – ovs Dec 5 at 15:01
  • @ovs, Ah, that's probably the problem why I couldn't also make the __import__('...') way work. – Kirill L. Dec 5 at 15:03
  • Note that this outputs an alpha value on macOS, so it only adheres to the spec on Windows. – ovs Dec 5 at 15:09
  • @ovs, OK, good to know, thanks. – Kirill L. Dec 5 at 15:11

Mathematica, 69 Bytes

Just the function is 34 bytes.

CurrentScreenImage[]~PixelValue~#&

Takes input in the form {x,y}.

The image is the merging of images on all monitors. If you want a particular screen, use the integer index - e.g. CurrentScreenImage[1]

Full program exactly as specified is 69 Bytes CurrentScreenImage[]~PixelValue~ToExpression["{"<>InputString[]<>"}"]

AutoHotKey, 113 bytes

CoordMode,Pixel
InputBox,x
InputBox,y
PixelGetColor,c,x,y
MsgBox % c&255 . "," . c>>8&255 . "," . c>>16&255

Use dialog boxes instead of stdin/stdout.

TI-Nspire assembly - 112 bytes

50 52 47 00 30 40 2D E9 FF FF FF FA 00 F0 17 F8
04 1C 00 F0 14 F8 85 00 2D 18 AD 01 2D 19 6C 00
C0 21 09 06 09 69 08 5B 3F 25 42 09 2A 40 1F 25
03 1C 2B 40 C1 0A 29 40 0A A0 0A DF 30 BD 00 20
0A 23 07 49 10 25 8A 69 2A 42 FC D1 0A 68 FF 25
2A 40 30 3A 0B DB 85 00 2D 18 6D 00 A8 18 F1 E7
00 00 02 90 25 64 2C 25 64 2C 25 64 0A 00 70 47

This program outputs integers in the range 0-31 for R and B and 0-63 for G, because the device natively uses a RGB565 framebuffer. It uses serial for input and output.

Source:

.string "PRG"
push {r4, r5, lr}
blx main
.thumb
main:
@ read x and y from serial into r4 and r0
bl read_int
mov r4, r0
bl read_int

@ turn x and y into framebuffer offset
@ r4 = ((r0 * 320) + r4) * 2
lsl r5, r0, #2
add r5, r0
lsl r5, #6
add r5, r4
lsl r4, r5, #1

@ load pixel from framebuffer
@ r0 = ((uint16_t **)0xc0000000)[0x10][r4 / 2]
mov r1, #0xC0
lsl r1, #24
ldr r1, [r1, #0x10]
ldrh r0, [r1, r4]

@ unpack RGB565 value into r1, r2, r3
mov r5, #0x3f
lsr r2, r0, #5
and r2, r5
mov r5, #0x1f
mov r3, r0
and r3, r5
lsr r1, r0, #11
and r1, r5

@ call printf
adr r0, fmt
swi #10

@ return
pop {r4, r5, pc}

@ subroutine to read an integer from serial
read_int:
mov r0, #0
mov r3, #10
ldr r1, serial_base
@ loop until characters come in on serial
2:
mov r5, #(1<<4) 
1:
ldr r2, [r1, #0x18]
tst r2, r5
bne 1b
@ read character from serial and mask out status bits
ldr r2, [r1]
mov r5, #0xff
and r2, r5
@ subtract 48 ('0') from byte; if result is negative, return
sub r2, #48
blt 1f
@ multiply existing numbers by 10 and add new number to them
lsl r5, r0, #2
add r5, r0
lsl r5, #1
add r0, r5, r2
b 2b

serial_base:.word 0x90020000
fmt:.string "%d,%d,%d\n"
@ this instruction is over here because serial_base and fmt need to be word-aligned
1:bx lr

Bash + coreutils + scrot + netpbm, 90 bytes

scrot -e'pngtopnm $f'|(read a;read w h;read m;head -c$((3*(w*$2+$1)))>f;od -t u1 -N3 -An;)

Loose I/O version

Takes x and y as separate command-line arguments.

Prints r, g, b as ints from 0-255 on separate lines


Bash + coreutils + scrot + netpbm + bc + sed, 172 bytes

IFS=, read x y
scrot -e'pngtopnm $f'|(read a;read w h;read m;head -c$((3*(w*$y+$x)))>f;od -vw1 -tu1 -N3 -An|while read p;do bc<<<"scale=2;$p/$m"|tr '\n' ,;done;)|sed s/,$//

Strict I/O version

Input on stdin as x,y

Output on stdout as r.rr,g.gg,b.bb (no newline).

  • @Shaggy not anymore -- see the updates – eeze Dec 9 at 2:11

TI-BASIC (TI-83/84+), 15 bytes

Input Y
255not(rand(3)pxl-Test(Y,Ans

Takes one integer from Ans and one from the prompt. rand(3) creates a list of 3 nonzero random numbers, so the product is zero iff the pixel is dark.

Your Answer

By clicking "Post Your Answer", you acknowledge that you have read our updated terms of service, privacy policy and cookie policy, and that your continued use of the website is subject to these policies.

Not the answer you're looking for? Browse other questions tagged or ask your own question.