Send key code to command line program on OS X
I want to make a script that starts a program and then sends it key input. In psuedo-script:
#!/bin/bash
./program << (PRESS CONTROL-Z)
The program is running so if there were additional commands in the script they will not be reached unless say control-z terminates the program.
Is this possible? From what I've found I thought开发者_Go百科 it might require key codes but I could be wrong.
This I think is probably a better solution than "expect" since it can be executed in native bash script, I'll be interested to see what you think.
Use
`printf "character code here"`
note the backticks
So for instance I have written a script that controls a remote gnu screen session, the following line opens window 2 and issues the ctrl-c key combo
ssh -t user@$host screen -p 2 -X stuff `printf "\003"`
- The -t option simulates terminal input on the remote machine
- -p allows us to specify the name or number of the window we are connecting to within the screen session.
- \003 is the bash format of character code 0x03
See here for a complete reference of codes.
To find the code of some key input you can use
printf "%#x\n" "'X"
0x58
- Were X is the key you want to find the code of
- To find codes of non literals you can use ctrl-v (makes bash append the next key to the command line rather than intepret it) and then type the key combo, so if I wanted to find the key code for ctrl-c I would delete the X press ctrl-v and then press ctrl-c.
One last thing the ascii code reference mentioned above lists 0x13 as the carriage return, but in the screen manual they list 0x15 as the enter key code, does anyone know why? Ive tested in a local screen and when I press enter 0x13 is produced, but when sending commands via ssh to a remote screen 0x13 doesn't work but 0x15 does.
Hope that helps
Piers
You might be looking for expect
(from http://expect.nist.gov/). This deals with the complexities of pseudo-ttys that make it appear to the program that the input from the script (in this scenario, the expect
program) is coming from a terminal.
Alternatively, you might be able to use echo
or cat
and pipe the output of that into the program - it depends on the program.
If you just want the program to start in the background, just do
#!/bin/bash
./program&
If your intent is to background the program, use:
./program & # The & sends the command to the background
echo commands here are executed while program is in the background
…
wait # Wait for the completion of background commands
echo commands here are executed after background program has completed
Edit: If your intent is to stop the program (as ctrl-Z often does in *nix shells), you can send it the STOP signal:
kill -STOP pid
To resume the execution, send it the CONT signal:
kill -CONT pid
In each of these examples pid
is the process id of the program. If you launch it in a script, it's easy to get with the variable $!
, e.g.
./prog &
echo prog started in the background
pid_of_prog=$!
kill -STOP $pid_of_prog
echo prog stopped
kill -CONT $pid_of_prog
echo prog continues
wait
echo prog finished
Edit 2: If your program is one that exits when it receives a ctrl-Z character, then remember that the control characters have the numerical value of the position letter in the alphabet (i.e. Ctrl-A is 1, Ctrl-B is 2, etc.). To send this character to a program you can:
echo -e "\032" | ./prog
(032
is 26, i.e. ^Z, in octal. Of course you can produce the same character by any means, perhaps adding it to the end of other input like ( cat inputfile ; echo -e "\032" ) | ./prog
.
But this may not necessarily work; the program must be designed to recognise this character from the input (which it probably won't); usually the shell catches it. Then again, most programs reading input from stdin
just exit when the input ends, so redirecting any finite input (even </dev/null
) should cause it to terminate.
And, finally, if the intent was to stop the execution of the program when some other event (detected elsewhere in the script) has occurred, you can just kill
it…
精彩评论