Bash script interacting with application

Question

I am trying to automate a scanning proccess for a physical application. My script is

#!/bin/bash i=1 file="ATLASbins.txt" while IFS= read line do scan=$line cat test.sh | sed "s/ set vchi 5000/ set vchi $scan/g" > test2.sh chmod +x test2.sh bash -x /home/mario/Mine/test2.sh i=$((i + 1)) done <"$file" 

Where test2.sh is another script in which I launch the application in which the scans are made. An example of what goes in the second script is

#!/bin/bash /home/mario/mg5/bin/mg5_aMC "import model Implementation" "generate u++ > l+ l+" output firstscript$i set vchi 6500 launch firstscript$i 

Where 'import model', 'output' and 'launch' are commands of the application (which runs in the terminal).

What happens is that the commands (inside the application) don't work and I get lines like

PATH/test2.sh: line 5: import model Implementation: command not found

I have no idea how to do this (which is to write a script that can write the commands to the application) and have tried various different separators, running the test2 script in a different shell and to call it in a new terminal using terminal-gnome. How can I make this work?

Another observation is that I need the variable value of i to be written inside the application. I tried this writing, for instance, 'output firstscript$i', which wouldn't work even if the command worked, I imagine.

Answer 1

Since you run your test2.sh as a separate shell script, the shell variable i is not available in it. This means that the expansion of $i will be empty in the test2.sh script. You can solve this in two ways:

1. Make i an environment variable through export i in the first script. This is convenient, but not really a good solution in the general case, as the other script may well want to use its own i variable which may be independent of whatever value the variable is in the calling script.

2. Give $i on the command line of the test2.sh script when invoking it: test2.sh "$i". This would make it possible to access the value $i in test2.sh as "$1" (the first command line argument).

The second issue is that your mg5_aMC program expects to get input, i.e. the commands that you list in the script. But the way you have entered these in test2.sh means that they would be taken as shell commands. There is nothing in the script that passes the special control commands to the program.

As you can see from the error that you get, it's the shell that complains that the commands can't be found. They are not shell commands, so this is (when you know how things will be interpreted) no surprise.

---

Assuming that your mg5_aMC program reads from standard input, I would not write out a shell script for each run of the program, but instead provide an input control/command file for it:

#!/bin/bash file="ATLASbins.txt" i=1 while IFS= read -r scan do sed -e "s/@scan@/$scan/" \ -e "s/@i@/$i/" \ input-template.in >input.in /home/mario/mg5/bin/mg5_aMC <input.in i=$((i + 1)) done <"$file" 

Here, the file input-template.in may look something like

"import model Implementation" "generate u++ > l+ l+" output firstscript@i@ set vchi @scan@ launch firstscript@i@ 

I've chosen to use @thing@ for things I will replace with the sed call in the shell script.

This assumes that the value of $scan contains no characters that would disturb the sed command (such as /).

---

An alternative, if the control script for your program is reasonably short, is to use "here-document" to feed the control commands into your program:

#!/bin/bash file="ATLASbins.txt" i=1 while IFS= read -r scan do /home/mario/mg5/bin/mg5_aMC <<END_INPUT "import model Implementation" "generate u++ > l+ l+" output firstscript$i set vchi $scan launch firstscript$i END_INPUT i=$((i + 1)) done <"$file" 

A here-document is basically a type of redirection of a piece of text which is given, not in a file, but between <<TAG and the ending TAG. The variables in the text would (if written as above) be expanded by the shell before being fed into your command.