Running commands

List techniques

A sequence of commands on a single command line.

Conditional lists

Each command depends on the success or failure of the previous command. This depends on the exit code. An exit code of 0 means success, and non-zero means failure:

echo $?             # get exit code

Use the && and || operators to create conditional lists:

&&: runs the following command only if the first succeeds. Good for chaining git commands.
||: runs the following command only if the first fails. Good if you want to exit a script if a command fails:

git add . && git commit -m 'commit message' && git push         # run only if all commands succeed
cd dir || mkdir dir                                             # if cd fails, make dir
cd dir || exit 1                                                # exit script with error if cannot access dir
cd dir || mkdir dir && cd dir || echo "cd dir failed"           # combined operators

Unconditional lists

Separated with a ; only–commands run one after the other, regardless of success or failure:

good for ad hoc commands
same as typing the commands individually and pressing enter
only the last exit code is assigned to the $? shell variable

sleep 7200; cp -a ~/important-files /mnt/backup-drive

Substitution techniques

Automatically replace the text of a command with other text.

Command substitution

A command is replaced by its output:

runs in a subshell, so accepts aliases
in scripts, commonly used to store output of command in variable

$(<command>)                                                # syntax
`command`                                                   # old syntax - DO NOT USE
mv $(grep -l "text match" *.txt) dest-dir                   # substitutes a list of files to move
echo Today is $(echo $(date +%A) | tr a-z A-Z)!             # nest sub commands
MATCHES=$(grep -l "text match" *.txt)                       # assign output to var

Process substitution

A command is replaced by its output, but as if its stored in a file:

transform output into a file so you can run commands that only read from disk to read from stdin
helpful to compare files, bc diff can only compare files

Process substitution runs a command and associates its output with a file descriptor:

echo <(ls -l)
/dev/fd/63          # file descriptor

set _o posix        # (if necessary) enable process substition

<(command)          # syntax

diff <(ls *.jpg | sort -n) <(seq 1 1000 | sed 's/$/.jpg/') \    # compare contents of one dir with test vals generated on the fly and output as a file
> | grep '>' \                                                  # only output file lines
> | cut -c3-                                                    # display the filenames only

Command-as-string techniques

Construct a string piece-by-piece and then run the string as a command:

Argument to `bash`

When you run a command with bash, it creates a child process with its own environment, including current working directory and variables.

Changes to the shell in a bash command do not affect your present shell
Helpful with sudo commands that use redirection, because the shell runs the redirection command after evaluating the entire command

bash -c "<command>"                 # syntax

sudo bash -c 'echo "new log file" > /var/log/custom.log'        # sudo redirection, otherwise would only run echo as sudo
[sudo] password for user: 

cat /var/log/custom.log 
new log file

Piping to `bash`

bash reads anything from stdin, so you can pipe commands to bash:

good when you need to run multiple similar commands in a row

These steps are going to move every file in a directory into another directory whose first letter corresponds to the first letter in the file name. For example, files apple and aardvark move to the a/ directory. The general process is as follows:

Print the commands by manipulating strings
View the resuls with less
Pipe the results to bash

Here are the steps:

Create the subdirectories:
```
mkdir {a..z}
```
Create the regex for the sed command you will use in mv. Begin by capturing the first character of the filename as expression #1 (\1) for sed:
```
^$.$
```
Get the rest of the filename as expression #2 (\2):
```
$.*$$
```
Connect the two:
```
^$.$$.*$$
```
Create the mv command, passing the sed expressions as the filename:
```
mv \1\2 \1           # mv apple a
```

Check the full command with less:

ls -1 ??* | sed 's/^\(.\)\(.*\)$/mv \1\2 \1/' | less

When satisfied, pipe it to bash to execute:

ls -1 ??* | sed 's/^\(.\)\(.*\)$/mv \1\2 \1/' | bash          # ??* makes sure the file is 2 chars long so it doesn't collide with the filenames

Remote commands with SSH

You can execute a single command by passing it after the ssh command:

quicker than logging in, executing command, logging out
quote or escape special chars and redirection that need to be evaluated by remote shell
you can pipe commands to ssh like you would bash
- suppress pseudo-terminal output with -T option
- suppress welcome text by processing with bash

ssh user@host <command>                     # syntax

ssh user@host ls > output                   # output file on local machine
ssh user@host "ls > output"                 # output file on remote machine
echo "ls > output" | ssh user@host          # pipe to SSH
echo "ls > output" | ssh -T user@host       # pipe to SSH, suppress pseudo-terminal info
echo "ls > output" | ssh user@host bash     # pipe to SSH, suppress welcome text

xargs

Helps you construct and run multiple similar commands. It accepts two inputs:

List of whitespace-separated strings in stdin
Incomplete command template that is missing some arguments

xargs merges the list of strings and the command template create and run complete commands.

with find, xargs is usually a cleaner option than -exec
-n: controls now many arguments are appended by xargs onto each generated command
I: controls where the input strings appear in the generated command. Follow -I by any arbitrary string to make that string the placeholder. Seems to be like the {} in find’s -exec. Using -I limits xargs arguments to 1.
-0: changes the input separator from whitespace to the null character (ASCII zero, \0). With find, protects against special characters in input. Pass -print0 to find to separate input strings with nulls. ls cannot cleanly change space to nulls, so you can create an alias: ls0="find . -maxdepth=1 -print0"

find . -type f -name \*.py -print0 | xargs -0 wc -l     # print num of lines in all pythong files in pwd

# -n option
ls
apple  banana  cantaloupe  date  watermelon

ls | xargs -n1 echo                             # one option at a time
apple
banana
cantaloupe
date
watermelon

ls | xargs -n2 echo                             # two options at a time
apple banana
cantaloupe date
watermelon

Process-control techniques

Background commands

Immediately return the prompt and execute the command out of sight (in the background). Good for commands that take a long time to complete. Commands that occupy the shell are called foreground commands. A shell can run one foreground command, and any number of background commands.

Background commands are part of the shell’s job control:

A job is a shell’s unit of work, its a single instance of a command running in the shell
Each job has an ID that is a positive integer, or job ID
A good use of background jobs is to suspend your text editor with CTRL+Zwhile you go back to the shell to run commands.

To run a command in the background, append an ampersand to the end of the command:

<command> &                                     # syntax
<command1> & <command2> & <command3> &          # multiple bg commands

Job control commands

Command	Description
`jobs`	View shell’s jobs
`CTRL + Z`	Suspend a running job
`bg`	Run current suspended job in the background
`bg %n`	Move suspended job n into the background
`fg %n`	Move background job n into the foreground
`kill %n`	Kill background job n

Explicit subshells

Launch these in the middle of a combined command. Enclose the command in parentheses (()) to run in a child shell:

useful if you need to change dirs in a pipeline command

(cd /usr/local)                     # doesn't change parent shell pwd

Process replacement

Supersedes the parent shell. Use with exec, but I couldn’t find any practical examples for what I do!