History Repeats Itself
Linux remembers everything you’ve done (at least in the current shell session), which allows you to pull steps from your history, potentially modify them, and redo them. This can obviously save a lot of time and typing.
The ‘head’ command views the first 10 (by default) lines of a file. The ‘tail’ commands views the last 10 (by default) lines of a file. Type ‘man head’ or ‘man tail’ to consult their manuals.
<up arrow> # last command <up> # next-to-last command <down> # last command, again <down> # current command, empty or otherwise history # usually too much for one screen, so ... history | head # we discuss pipes (the vertical bar) below history | tail history | less # use 'q' to exit less ls -l pwd history | tail !560 # re-executes 560th command (yours will have different numbers; choose the one that recreates your really important result!)
Here are some more ways to make editing previous commands, or novel commands that you’re building up, easier:
<up><up> # go to some previous command, just to have something to work on <ctrl-a> # go to the beginning of the line <ctrl-e> # go to the end of the line # now use left and right to move to a single word (surrounded by whitespace: spaces or tabs) <ctrl-k> # delete from here to end of line <ctrl-w> # delete from here to beginning of preceeding word blah blah blah<ctrl-w><ctrl-w> # leaves you with only one 'blah'
You can also search your history from the command line:
<ctrl-r>fir # should find most recent command containing 'fir' string: echo 'first' > test.txt <enter> # to run command <ctrl-c> # get out of recursive search <ctr-r> # repeat <ctrl-r> to find successively older string matches
Create and Destroy
We already learned one command that will create a file, touch. Lets create a folder in /share/workshop for you to work in and then another directory cli. We will use the environment variable $USER, that contains your username.
cd # home again echo $USER # echo to screen the contents of the variable $USER mkdir ~/tmp2 cd ~/tmp2 echo 'Hello, world!' > first.txt
echo text then redirect (‘>’) to a file.
cat first.txt # 'cat' means 'concatenate', or just spit the contents of the file to the screen
why ‘concatenate’? try this:
cat first.txt first.txt first.txt > second.txt cat second.txt
OK, let’s destroy what we just created:
cd ../ rmdir tmp2 # 'rmdir' meands 'remove directory', but this shouldn't work! rm tmp2/first.txt rm tmp2/second.txt # clear directory first rmdir tmp2 # should succeed now
So, ‘mkdir’ and ‘rmdir’ are used to create and destroy (empty) directories. ‘rm’ to remove files. To create a file can be as simple as using ‘echo’ and the ‘>’ (redirection) character to put text into a file. Even simpler is the ‘touch’ command.
mkdir ~/cli cd ~/cli touch newFile ls -ltra # look at the time listed for the file you just created cat newFile # it's empty! sleep 60 # go grab some coffee touch newFile ls -ltra # same time?
So ‘touch’ creates empty files, or updates the ‘last modified’ time. Note that the options on the ‘ls’ command you used here give you a Long listing, of All files, in Reverse Time order (l, a, r, t).
When you’re on the command line, there’s no ‘Recycle Bin’. Since we’ve expanded a whole directory tree, we need to be able to quickly remove a directory without clearing each subdirectory and using ‘rmdir’.
cd mkdir -p rmtest/dir1/dir2 # the -p option creates all the directories at once rmdir rmtest # gives an error since rmdir can only remove directories that are empty rm -rf rmtest # will remove the directory and EVERYTHING in it
Here -r = recursively remove sub-directories, -f means force. Obviously, be careful with ‘rm -rf’, there is no going back, if you delete something with rm, rmdir its gone! There is no Recycle Bin on the Command-Line!
Piping and Redirection
Pipes (‘|’) allow commands to hand output to other commands, and redirection characters (‘>’ and ‘»’) allow you to put output into files.
echo 'first' > test.txt cat test.txt # outputs the contents of the file to the terminal echo 'second' > test.txt cat test.txt echo 'third' >> test.txt cat test.txt
The ‘>’ character redirects output of a command that would normally go to the screen instead into a specified file. ‘>’ overwrites the file, ‘»’ appends to the file.
The ‘cut’ command pieces of lines from a file line by line. This command cuts characters 1 to 3, from every line, from file ‘test.txt’
cut -c 1-3 test.txt
same thing, piping output of one command into input of another
cat test.txt | cut -c 1-3
This pipes (i.e., sends the output of) cat to cut to sort (-r means reverse order sort), and then grep searches for pattern (‘s’) matches (i.e. for any line where an ‘s’ appears anywhere on the line.)
cat test.txt | cut -c 1-3 | sort -r cat test.txt | cut -c 1-3 | sort -r | grep s
This is a great way to build up a set of operations while inspecting the output of each step in turn. We’ll do more of this in a bit.
Compression and Archives
As file sizes get large, you’ll often see compressed files, or whole compressed folders. Note that any good bioinformatics software should be able to work with compressed file formats.
gzip test.txt cat test.txt.gz
To uncompress a file
gunzip -c test.txt.gz
The ‘-c’ leaves the original file alone, but dumps expanded output to screen
gunzip test.txt.gz # now the file should change back to uncompressed test.txt
Tape archives, or .tar files, are one way to compress entire folders and all contained folders into one file. When they’re further compressed they’re called ‘tarballs’. We can use wget (web get).
The .tar.gz and .tgz are commonly used extensions for compressed tar files, when gzip compression is used. The application tar is used to uncompress .tar files
tar -xzvf PhiX_Illumina_RTA.tar.gz
Here -x = extract, -z = use gzip/gunzip, -v = verbose (show each file in archive), -f filename
Note that, unlike Windows, linux does not depend on file extensions to determine file behavior. So you could name a tarball ‘fish.puppy’ and the extract command above should work just fine. The only thing that should be different is that tab-completion doesn’t work within the ‘tar’ command if it doesn’t see the ‘correct’ file extension.
BASH Wildcard Characters
We can use ‘wildcard characters’ when we want to specify or operate on sets of files all at once.
list files in Illumina sub-directory of any directory ending in ‘hiX’
list all files ending in ‘.fa’ a few directories down. So, ‘?’ fills in for zero or one character, ‘*’ fills in for zero or more characters. The ‘find’ command can be used to locate files using a similar form.
find . -name "*.f*" find . -name "*.f?"
how is this different from the previous ls commands?
Quick Note About the Quote(s)
The quote characters “ and ‘ are different. In general, single quotes preserve the literal meaning of all characters between them. On the other hand, double quotes allow the shell to see what’s between them and make substitutions when appropriate. For example:
VRBL=someText echo '$VRBL' echo "$VRBL"
However, some commands try to be ‘smarter’ about this behavior, so it’s a little hard to predict what will happen in all cases. It’s safest to experiment first when planning a command that depends on quoting … list filenames first, instead of changing them, etc. Finally, the ‘backtick’ characters ` (same key - unSHIFTED - as the tilde ~) causes the shell to interpret what’s between them as a command, and return the result.
# counts the number of lines in file and stores result in the LINES variable LINES=`cat PhiX/Illumina/RTA/Sequence/Bowtie2Index/genome.1.bt2 | wc -l` echo $LINES
Since copying or even moving large files (like sequence data) around your filesystem may be impractical, we can use links to reference ‘distant’ files without duplicating the data in the files. Symbolic links are disposable pointers that refer to other files, but behave like the referenced files in commands. I.e., they are essentially ‘Shortcuts’ (to use a Windows term) to a file or directory.
The ‘ln’ command creates a link. You should, by default, always create a symbolic link using the -s option.
ln -s PhiX/Illumina/RTA/Sequence/WholeGenomeFasta/genome.fa . ls -ltrhaF # notice the symbolic link pointing at its target grep -c ">" genome.fa
STDOUT & STDERR
Programs can write to two separate output streams, ‘standard out’ (STDOUT), and ‘standard error’ (STDERR). The former is generally for direct output of a program, while the latter is supposed to be used for reporting problems. I’ve seen some bioinformatics tools use STDERR to report summary statistics about the output, but this is probably bad practice. Default behavior in a lot of cases is to dump both STDOUT and STDERR to the screen, unless you specify otherwise. In order to nail down what goes where, and record it for posterity:
wc -c genome.fa 1> chars.txt 2> any.err
the 1st output, STDOUT, goes to ‘chars.txt’
the 2nd output, STDERR, goes to ‘any.err’
Contains the character count of the file genome.fa
Empty since no errors occured.
Saving STDOUT is pretty routine (you want your results, yes?), but remember that explicitly saving STDERR is important on a remote server, since you may not directly see the ‘screen’ when you’re running jobs.