Intro to the Command Line
Joe Fass
jnfass@ucdavis.edu
Basic Intro Material
Getting There
Secure SHell … ssh. Log in to the server using the following:
ssh [username]@servername
… where ‘username’ and the brackets are replaced by your username (class10, for example), and servername is replaced by the full address of the server your logging into.
Achieving Clarity, and Bugging Out
To start off: there will be many commands that will fill your screen with text. There are multiple ways to clear the clutter, and have an empty screen:
<enter> <enter> <enter>
<ctrl-l>
clear
And once you’re really done working on the command line:
exit # kills the current shell and exits!
# Note that any text *following* a "#" symbol is ignored.
But don’t exit yet … or if you did, just ssh back in. We’ve got work to do!
Command Line Basics: Listing Files and Killing Commands
First some basics - how to look at your surroundings.
pwd # present working directory ... where am I?
ls # list files here ... you should see nothing since your 'class##' homes are empty
ls /tmp/ # list files somewhere else
Let’s run our first time-consuming command … because one of the first things that’s good to know is how to escape once you’ve started something you don’t want.
sleep 1000 # wait for 1000 seconds!
<ctrl-c> # shows as '^C'; exits command entry without executing
# In some cases, a different key sequence is required
python # enter Python interactive session
<ctrl-d> # escape from some repl's ... "Read, Execute, & Print Loop"
R # enter R interactive session
<ctrl-d>
# Try this fun command:
yes # that's a lot of yes
<ctrl-c>
yes | more # "pipe," or direct, the yes output to the 'more' command
<spacebar>
<spacebar> # next page
<q> # quits from more, less, 'man' pages, etc.
So, ^C, ^D, ‘q’, and (from above) ‘exit’. Generally can’t hurt to try until one of them works! Now that we’ve seen the ‘less’ “paginator” (program that display output from other programs in pages, instead of all at once), here’s how to move around while in it.
yes | less # pipe to 'less' paginator, instead of 'more'
<spacebar>
<arrow keys, pgup, pgdn> # forward or back through file
g *or* G # beginning or end of file
/yes # '/' enters search mode, "yes" is pattern looked for (could be any string)
/y
/no
# After successfully finding some text matches, you can use:
n # next pattern match
N # previous pattern match
<spacebar> # all the navigation keys, like 'g' and <pgup>, still work
q # to quit
You’ve Got Options
One reason you’ll appreciate ‘less’ is that it’s the default paginator for the ‘man’ command. ‘Man’ stands for “manual,” and it’s the main way to get more detail on any of the commands we’ll introduce today. Each command can act as a basic tool, or you can add “options” or “flags” that modify the default behavior of the tool. These flags come in the form of ‘-v’ … or, when it’s a more descriptive word, two dashes: ‘–verbose’ … that’s a common (but not universal) one that tells a tool that you want it to give you output with more detail. Sometimes, options require specifying amounts or strings, like ‘-o results.txt’ or ‘–output results.txt’ … or ‘-n 4’ or ‘–numCPUs 4’. Let’s try some, and see what the man page for the “list files” command ‘ls’ is like.
ls -R /software
# ack! too much going to the screen!
<ctrl-c>
ls -R /software/scythe # lists directories and files *recursively*
# how do I know which options do what?
man ls
# navigate like in "less" (up/down,pgup/dn,g,G,/pattern,n,N,q)
# look up and try the following:
ls -l
ls -a
ls -l -a
ls -la # option "smushing" ... when no values need specifying
ls -ltrha
ls -ltrha --color # single letter (smushed) vs word options
# what if I want to use same options repeatedly? and be lazy?
alias # lists current *command aliases* or *shortcuts*
alias ll='ls -ltrhaF --color'
ll
alias l='ls --color'
Getting Around
The filesystem you’re working on is like the branching root system of a tree (image borrowed from web.sonoma.edu):
The top level, right at the root of the tree, is called the “root” directory, specified by ‘/’ … which is also the divider for directory addresses, or “paths.” (Note that there’s also often a directory named “root” just under the filesystem root … this is for the “root user” or “superuser” … but here we’re talking about the true filesystem root, ‘/’). We move around using the “change directory” command, ‘cd’:
cd # no effect? that's because by itself it sends you home (to ~, or /home/class10/ if you're class10)
cd / # go to root of tree's root system
cd home # go to where everyone's homes are
pwd
cd class10 # use your actual home, not class10
pwd
cd / # back to the root
pwd
cd ~ # a shortcut to home, from anywhere
pwd
cd . # "." always means *this* directory
pwd
cd .. # ".." always means *one directory up*
pwd
Don’t get confused between the “.” directory name and filenames that start with the “.” character. The latter are just valid filenames or directory names, but are usually hidden (use ls’s “-a” option to see hidden files).
Absolute and Relative Paths
The sequence above was probably confusing, if you’re not used to navigating filesystems this way. You can think of paths like addresses. You can tell your friend how to go to a particular store from where they are currently (a “relative” path), or from the main Interstate Highway that everyone uses (in this case, the root of the filesystem, ‘/’ … this is an “absolute” path). Both are valid. But absolute paths can’t be confused, because they tell you where to start off, and all the steps along the way. Relative paths, on the other hand, could be totally wrong for your friend if you assume they’re somewhere they’re not. With this in mind, let’s try a few more:
cd ~ # let's start at home
cd ../../home/class10/ # *relative* (start here, take two steps up, then down through home and class10)
pwd
cd /home/class10/ # *absolute* (start at root, take steps)
pwd
Linux also tolerates “empty” steps and loops, even if they look ugly. So:
cd # starting at your home again
cd /home//class10/ # used two slashes; it's treated as an empty step
cd ../class10/../class10/../class10/ # back and forth a few times.
cd /software/bwa/../bowtie/../../home/class10/ # are you lost or something?
There’s no real point to such weird paths, but it helps illustrate how paths work. In the last example, an absolute path, we start at the root of the filesystem (the initial ‘/’), move down into the ‘software’ directory, then down from there into the ‘bwa’ directory, then back up (into the software directory), down into ‘bowtie’, then up twice (which gets you back to the root directory), then down through the familiar ‘home’ and your own ‘class##’ directories. Now, wasn’t that a pain to type out all those directory names? So, let’s discuss …
Tab Completion - A Real Tendon-Saver
Using tab-completion will literally save your life. Hours of it. A single
touch one seven september # create three empty files using 'touch' command
cat o<tab><no enter> # will complete to "one"
<enter>
cat s<tab><no enter> # completes up to 'se' since that's in common between seven and september
<tab><no enter> # this second tab should cause listing of seven and september
v<tab><no enter> # now it's unique to, and should complete to seven
<enter> # runs "cat seven" command
# we often literally autocomplete commands letter by letter
# comes in handy if we don't exactly remember what name we want
ls /hom<tab>j<tab><tab>f<tab> # completes to my home directory, /home/jfass/
I can’t overstate how useful tab completion is. You should get used to using it constantly. Watch experienced users type and they maniacally hit tab once or twice in between almost every character. You don’t have to go that far, of course, but get used to constantly getting feedback from hitting tab and you will save yourself a huge amount of typing and trying to remember weird directory and filenames.
Create and Destroy
OK, so let’s get down to actually making some changes to the filesystem.
cd # home again
mkdir temp # make a directory called 'temp'
cd temp/
echo "Hello, world!" > first.txt # push text into a file using the '>' character
file first.txt # tells us what kind of file it is
If a file isn’t text, you probably don’t want to look at it. Binary (‘data’) files get pushed to the screen by the ‘cat’ command, and the screen takes character sized bites and displays whatever character that bite corresponds to … which can be weird non-printing characters like bells and interrupts that can really muck up your shell!
cat first.txt # 'cat' means "concatenate"
# 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 temp # shouldn't work!
rm temp/first.txt
rm temp/second.txt # clear directory first
rmdir temp # should succeed now
Piping and Redirection
| Pipes (‘ | ’) allow commands to hand output to other commands, and redirection characters (‘>’ and ‘»’) allow you to put output into files. |
mkdir CLB
cd CLB/
echo "first" > test.txt
cat test.txt
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. ‘>’ replaces, ‘»’ appends.
cut -c 1-3 test.txt # cuts character one to three, from every line, from file 'test.txt'
cat test.txt | cut -c 1-3 # same thing, piping output of one command into input of another
cat test.txt | cut -c 1-3 | sort -r
cat test.txt | cut -c 1-3 | sort -r | grep s
# pipes cat to cut to sort (-r means reverse order sort, grep searches for pattern matches)
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.
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.
<up arrow> # last command
<up> # next-to-last command
<down> # last command, again
<down> # current command, empty or otherwise
# try however many ups and downs you want, to step around in your history
# for a more global view:
history # usually too much for one screen, so ...
history | head
history | tail
history | tail -n 30
history | less
cat test.txt | cut -c 1-3 | sort -r | grep -s > reallyImportantResult.txt
history | tail
!560 # re-executes 560th command (yours will have different numbers; choose your last one)
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
Editing Yourself
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
<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)
<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"
Installing (Simple) Software
Let’s install a straightforward tool … another read aligner by the author of BWA that’s intended for longer, lower accuracy reads (such as from the PacBio or Oxford Nanopore sequencers): minimap2.
cd # returns you to your home directory, since we've been working in the 'CLB' directory
mkdir tools
cd tools/
git clone https://github.com/lh3/minimap2.git
cd minimap2/
make
./minimap2
cd
Now you can run this tool (an executable file) by fully specifying the path to your ‘tools’ directory from wherever you’re running the tool. In addition, you could copy, move, or put symbolic links to the tool in a common place, like /usr/bin/, which should be in everybody’s path.
Not-so-basic Intro Material (for Homework!)
Compression
With BIG DATA(TM), you’ll often see compressed files, or whole compressed folders.
gzip test.txt
file test.txt
gunzip -c test.txt | more # '-c' leaves the original file alone, but dumps expanded output to screen
ls -ltrha
gunzip test.txt
bzip2 test.txt; bunzip2 test.txt.bz2 # note the ';' is a substitute for <enter>
Note that there are analogs of several file manipulation tools that allow you to view compressed files as if they weren’t compressed, like ‘zless’ and ‘zcat’ for gzip-compressed files, and ‘bzless’ and ‘bzcat’ for bzip2-compressed files.
Archives
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.” Let’s grab one.
wget ftp://igenome:G3nom3s4u@ussd-ftp.illumina.com/PhiX/Illumina/RTA/PhiX_Illumina_RTA.tar.gz
# ... or ...
curl ftp://igenome:G3nom3s4u@ussd-ftp.illumina.com/PhiX/Illumina/RTA/PhiX_Illumina_RTA.tar.gz > PhiX.tgz
# .tar.gz and .tgz are *commonly used* extensions for compressed tar files, when gzip compression is used
tar -xzvf PhiX_Illumina_RTA.tar.gz # or whatever you called it, if you used curl
# -x = extract, -z = use gzip/gunzip, -v = verbose (show each file in archive), -f = use a file, not a tape drive(!)
Forced Removal
This gets a heading all its own. Because 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’.
rm -rf PhiX # be sure ... there's no going back!
# -r = recursively remove sub-directories, -f means *force* (auto-"yes")
# We actually want to use those directories, so un-archive them again!
Obviously, be careful with ‘rm -rf’.
BASH Wildcard Characters and Find
When we want to specify or operate on sets of files all at once.
ls ?hiX/Illumina # list files in Illumina sub-directory of any directory ending in "hiX"
ls PhiX/Illumina/RTA/Sequence/*/*.fa # list all .fa files a few directories down
# So, "?" fills in for zero or one character, "*" fills in for zero or more characters
find . -name "*.fa"
find . -name "*.f?" # how is this different from the previous command?
Manipulation of a FASTA File
We just found the phiX-174 genome, so let’s copy it to our current directory so we can play with it:
cp ./PhiX/Illumina/RTA/Sequence/WholeGenomeFasta/genome.fa phix.fa
# Note how we copied the 'genome.fa' file to a different name: 'phix.fa'
wc -l phix.fa
We can use the ‘grep’ command to search for matches to patterns (more flexibly than by using less’s ‘/’ key). ‘grep’ comes from “globally search for a regular expression and print.”
grep -c ">" phix.fa # only one FASTA sequence entry, since only one header line (">gi|somethingsomething...")
cat phix.fa # this may not be useful for anything larger than a virus!
# let's look at start codon and 2 following:
grep "ATG......" phix.fa # '.' characters are the single-character wildcards for grep
# use the '-o' option to **o*nly print the pattern matches, one per line
grep -o "ATG......" phix.fa
# use the 'cut' command with '-c' to select characters 4-6, the second codon
grep -o "ATG......" phix.fa | cut -c4-6
# 'sort' the second codon sequences (default order is same as ASCII table; see 'man ascii')
grep -o "ATG......" phix.fa | cut -c4-6 | sort
# combine successive identical sequences, but count them ('-c' option)
grep -o "ATG......" phix.fa | cut -c4-6 | sort | uniq -c
# and finally sort using only the 1st "field" as a key ('-k1,1'), in reverse numeric order ('-rn')
grep -o "ATG......" phix.fa | cut -c4-6 | sort | uniq -c | sort -rn -k1,1
# ... which gives us the most common codons first
This may not be a particularly useful thing to do with a genomic FASTA file, but it illustrates the process by which one can build up a string of operations, using pipes, in order to ask quantitative questions about sequence content. More generally than that, this process allows one to ask questions about files and file contents and the operating system, and verify at each step that the process so far is working as expected. The command line is, in this sense, really a modular workflow management system.
Symbolic Links
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.
ln -s PhiX/Illumina/RTA/Sequence/WholeGenomeFasta/genome.fa .
ls -ltrhaF # notice the symbolic link pointing at its target
grep -c ">" genome.fa
Bioinformatics, At Last! … ?
OK, let’s try to do some sequence alignment (similar to a BLAST search).
bwa mem genome.fa phix.fa > aln.sam
# We get the following:
# [E::bwa_idx_load] fail to locate the index files
What went wrong? We redirected output to the ‘aln.sam’ file using the ‘>’ character. But there’s nothing in the output file:
cat aln.sam
# <nothing>
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:
bwa mem genome.fa phix.fa 1> aln.sam 2> aln.err
# the 1st output, STDOUT, goes to 'aln.sam'
# the 2nd output, STDERR, goes to 'aln.err'
cat aln.sam
# <nothing>
cat aln.err
# [E::bwa_idx_load] fail to locate the index files
That didn’t really help us with our problem, but at least you can direct output where you want it now. Saving STDOUT is pretty routine (you want your results, yes?).
For the curious, our problem was that we didn’t index the “target” or “reference” of our sequence alignment. Try this:
bwa index genome.fa
bwa mem genome.fa phix.fa 1> aln.sam 2> aln.err
cat aln.err
# see? summary of the alignment process / results
cat aln.sam
The resulting SAM file shows a perfect match between the query sequence (‘phix.fa’) and the reference (‘genome.fa’) … which is good, because they’re the same file (see the symbolic links section above)! Beyond that, don’t worry about the SAM format; we’ll get into that tomorrow.
Loops
Loops are useful for quickly telling the shell to perform one operation after another, in series. For example:
for i in {1..21}; do echo $i >> a; done
cat a
# <1 through 21 on separate lines>
The general form is:
for name in {list}; do
commands
done
The list can be a sequence of numbers or letters, or a group of files specified with wildcard characters:
# imagine you have 20 sequence files, in a 'fastqs' directory:
bwa index reference.fa
for sample in fastqs/*.fastq; do
bwa mem reference.fa $sample 1> $sample.sam 2> $sample.err
done
# this would produce, for example, ./fastqs/sample1.fastq.sam and ./fastqs/sample1.fastq.err, etc.
Sometimes a “while” loop is more convenient than a “for” loop … if you don’t readily know how many iterations of the loop you want:
while {condition}; do
commands
done
Or, imagining a file that contains the filenames (one per line) of samples’ sequence data:
cat file-of-filenames.txt | while read sample; do
bwa mem reference.fa $sample 1> $sample.sam 2> $sample.err
done
Paste Command
The paste command is useful in creating tables.
echo "WT1" >> b
echo "WT2" >> b
echo "control1" >> b
echo "control2" >> b
# now we can number our four samples to conveniently refer to them in order
for i in {1..4}; do echo $i >> a; done # semicolons terminate separate lines of multi-line commands like loops
paste a b > c
cat c
Running in the Background
Sometimes it’s useful to continue working on the command line, even after you’ve executed a command that’s going to take a while to finish. Normally this command would occupy the shell, and prevent you from typing in commands and receiving results. But we can “put jobs in the background” so that they don’t occupy your shell directly:
sleep 1000000
<control-z> # shows as '^Z'
# [1]+ Stopped sleep 1000000
bg
# [1]+ sleep 1000000 &
‘^Z’ first suspends the sleep command (or whatever command is running when you hit
jobs
# [1]+ Running sleep 1000000 &
We can also start a job in the background in one step, without having to suspend then resume it, using the ‘&’ character at the end of the command:
sleep 5000000 &
If we want to delete these jobs for any reason, we can kill them using the numbering that ‘jobs’ reveals:
jobs
# [1]- Running sleep 1000000 &
# [2]+ Running sleep 5000000 &
kill %1
jobs
# [1]- Terminated sleep 1000000
# [2]+ Running sleep 5000000 &
kill %2
jobs
# [2]+ Terminated sleep 5000000
Finally, the ‘nohup’ command (from “no hangup”!) makes jobs extra resistant to lost connections or terminal problems. In other words, even jobs running in the background can be terminated if one’s shell dies. ‘nohup’ separates the running job from the shell, so it’ll keep running until it dies or is actively killed by you.
nohup sleep 1000000 &
# [1] 34993
# class##@c4-0:~/CLB$ nohup: ignoring input and appending output to ‘nohup.out’
jobs
# [1]+ Running nohup sleep 1000000 &
# output is dumped into the 'nohup.out' file unless specifically redirected in your command
kill %1
Table of Processes
The ‘top’ command prints a self-updating table of running processes and system stats. Use ‘q’ to exit top, ‘z’ to toggle better color contrast, ‘M’ to sort by memory use, ‘P’ to sort by processor use, and ‘c’ to toggle display of the full commands. Hit ‘1’ to toggle display of all processors, and hit ‘u’ followed by typing in a username in order to only show processes (jobs) owned by that user.
Shell Scripts, File Permissions
Often it’s useful to define a whole string of commands to run on some input, so that (1) you can be sure you’re running the same commands on all data, and (2) you don’t have to type the same commands in over and over! Let’s use the ‘nano’ text editor program that’s pretty reliably installed on most linux systems (MacOS??).
nano test.sh
# nano now occupies the whole screen; see commands at the bottom
# type/paste in the following ...
# (note that '#!' is an interpreted command to the shell, not a comment)
# (also note that you may have to add and delete spaces and <enter>s to get the spacing right, though that's not critical)
#!/bin/bash
grep -o . $1 | \
sort | \
uniq -c | \
sort -rn -k1,1
<control-o><control-x> # to write **o**ut to test.sh, and then e**x**it nano
Note that ‘$1’ means ‘the value of the 1st argument to the shell script’ … in other words, the text that follows the shell script name when we run it (see below).
Though there are ways to run the commands in test.sh right now, it’s generally useful to give yourself (and others) “execute” permissions for test.sh, really making it a shell script. Note the characters in the first (left-most) field of the file listing:
ll test.sh
# -rw-r--r-- 1 jfass biocore 79 Aug 19 15:05 test.sh
The first ‘-‘ becomes a ‘d’ if the “file” is actually a directory. The next three characters represent read, write, and execute permissions for the file owner (you), followed by three characters for users in the owner’s group, followed by three characters for all other users. Run the ‘chmod’ command to change permissions for the ‘test.sh’ file, adding execute permissions (‘+x’) for the user (you) and your group (‘ug’):
chmod ug+x test.sh
ll test.sh
# -rwxr-xr-- 1 jfass biocore 79 Aug 19 15:05 test.sh*
OK! So let’s run this script, feeding it the phiX genome. When we put the genome file 1st after the name of the script, this filename becomes variable ‘1’, which the script can access by specifying ‘$1’.
./test.sh genome.fa
# 1686 T
# 1292 A
# 1253 G
# 1155 C
# 1 x
# 1 p
# 1 i
# 1 h
# 1 >
The script’s grep command splits out every character in the file on a separate line, then sorts them so it can count the occurrences of every unique character and show the most frequent characters first … a quick and dirty way to get at GC content.