Use & (ampersand) in single line bash loop

Question

I have been using this command successfully, which changes a variable in a config file and then executes a Python script within a loop:

for((i=114;i<=255;i+=1)); do echo $i > numbers.txt; python DoMyScript.py; done

As each DoMyScript.py instance takes about 30 seconds to run before terminating, I'd like to relegate them to the background while the next one can be spawned.

I have tried what I am familiar with, by adding in an ampersand as below:

for((i=114;i<=255;i+=1)); do echo $i > numbers.txt; python DoMyScript.py &; done

However, this results in the below error:

-bash: syntax error near unexpected token `;'

Solution 1

Drop the ; after &. This is a syntactic requirement

for((i=114;i<=255;i+=1)); do echo $i > numbers.txt;python DoMyScript.py & done

Solution 2

Given Stephane's comment on 1_CR's answer, you probably want:

for i in {114..255}; do { echo $i > numbers.txt && python DoMyScript.py; } & done

Solution 3

Lose the ;:

for((i=114;i<=255;i+=1)); do echo $i > numbers.txt;python DoMyScript.py & done

Solution 4

As stated by the other answers:

• The problem is the ; after the &.
• Deleting the ; after the & will result in a command that will run without shell syntax errors.  However, it is unlikely to function correctly, because it creates a race condition.

The OP stated that he resolved the race condition by adding a 3 second delay.

Nobody mentioned:

• A 3 second delay might work today.  Tomorrow, the computer might be more sluggish, and it could fail again.  Next year, the script might be modified to be more time-consuming, and it could fail again.  The probability of failure goes down if you increase the delay by an order of magnitude, by which I mean 30 seconds.  Of course this means that the script processes will (probably / presumably) be running essentially sequentially, and not concurrently (in parallel), thus defeating the purpose of making the processes asynchronous.
• If running multiple instances of the script is a requirement, it’s probably best to modify the script to accept the number parameter on the command line:

  for ((i=114;i<=255;i++)); do python DoMyScript.py --number="$i" & done
  

  and do away with the file.
• If changing the script is not an option, this solution should work:

  for i in {114..125}; do ( subdir="dir.$i" && mkdir "$subdir" && cd "$subdir" && 
            echo "$i" > numbers.txt && python ../DoMyScript.py; cd .. && rm -r "$subdir" ) & done
  

  This gives each script process a separate numbers.txt file, by giving it a separate directory to run in.  The directory is created and deleted, and the script is run, in a subshell.

  • If DoMyScript.py accesses any files other than numbers.txt by relative pathnames, the above command will need to be adjusted to accommodate that.

  • The above was inspired by glenn jackman’s answer (although I believe that mine will work and his won’t).

  • If command B depends on the success of command A, then A && B is better defensive programming than A; B.  But it’s probably not the ideal way to handle this situation.  If mkdir dir.114 fails, then the next 111 attempts are likely to fail also, and you’ll get 112 error messages.  It would be better to abort the loop if a fatal error occurs.

    The fact that the action is happening in an asynchronous subshell makes this somewhat tricky.

    for i in {114..125}; do
            { subdir="dir.$i" && mkdir "$subdir" && echo "$i" > "$subdir"/numbers.txt; } || break;
            ( cd "$subdir" && python ../DoMyScript.py; cd .. && rm -r "$subdir" ) & done
    

    will cause the loop to abort if a mkdir or an echo value > file command fails.

  • It might be better to work in a directory that is ‘‘guaranteed’’ to be writable, like /tmp.  However, this increases the risk that your command will collide (interfere) with some other process.

    • You can mitigate this by adding $$ to the directory name; e.g., dir.$$.$i or even dir.$BASHPID.
    • This would not eliminate the risk that the mkdir or the file creation might fail because the filesystem is full.

  • Note that the above code will keep on going and remove the temporary directory even if the script fails.  You might want to do something else in that situation.

    • Oops.  If, for some reason, the mkdir "$subdir" succeeds but the cd "$subdir" fails, this will go ahead and do cd .. && rm -r "$subdir".  If there is a directory with a name like dir.114 in your parent directory (i.e., parallel to your current directory), it will be removed.  I believe that you can (at least somewhat) fix this by changing the last line to

      ( cd "$subdir" && { python ../DoMyScript.py; cd ..; } && rm -r "$subdir" ) & done
      

      or by using absolute paths; e.g., subdir="$PWD/dir.$i".

View more solutions

Comments

[Martin von Wittich]

Damn it, 43 seconds too slow :)

[Mathias Begert]

@MartinvonWittich, a +1 to your answer by way of commiseration ;-)

[Stéphane Chazelas]

While that answers the question, it's probably not what the OP wants as by the time all the python instances have started and initialises they'll all see a numbers.txt containing 255.

[user]

Agreed with @StephaneChazelas, but am not sure what we can really do about it with the information given in the question. A proper solution to that probably needs more context.

[ljs.dev]

That worked great. OP likes the outcome, the file is changed and then read each time, so each iteration of the script gets a different variable. I did need to add in a 3 second delay just to avoid some race-condition/oddities.

[Angel Todorov]

Here's the bash manual reference: gnu.org/software/bash/manual/bashref.html#Lists

[ljs.dev]

Thanks Glenn. I tried a few variations of your code which I like due to the increased readability, but perhaps due to the extra sleep I needed, it was not working out for me. It would also repeat the last number in the for range... Here's what I am happy with at the moment: for((i=1;i<100;i+=1)); do echo $i > numbers.txt;sleep 3; python DoMyScript.py & done The main question point of getting the process to execute in the background is solved.

[ljs.dev]

Thanks @glennjackman, I would not have thought to search under lists, based on my previous usage of & limited to just appending it to commands to shush them. Time to bash my head against bash properly!

[ljs.dev]

*of course, I had <= in the original question, so the range you supplied was totally suitable based on the info I supplied. Again, thanks for the different approach

[G-Man Says 'Reinstate Monica']

Changing the ; after echo $i > numbers.txt to && is an improvement.  OTOH, if you can’t write a three-digit number to a file, you should probably abort the whole operation, and if you can write to the file on some iterations but not others, you have a problem that needs to be investigated and resolved. … (Cont’d)

[G-Man Says 'Reinstate Monica']

(Cont’d) … But, aside from that, I don’t see how this offers any real improvement.  You still have 112 python DoMyScript.py processes being spawned milliseconds apart.  Unless DoMyScript.py starts very quickly, the shell loop will have gone on and written the next value to numbers.txt before DoMyScript.py has read the value it’s supposed to read.