Skip to content

Notes on the Unix Programming Environment buy B. Kernighan and R. Pike

Notifications You must be signed in to change notification settings

ahmaazouzi/unix_programming_environment

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

76 Commits
 
 
 
 

Repository files navigation

THE UNIX PROGRAMMING ENVIRONMENT

This is an old book but it's held in high esteem by many. Seeing how many linux commands resemble those of unix and having enjoyed the C book, co-written by Kernighan, I thought I'd skim over this one and see what it still has to offer. These are my notes on what I understand of the book and what I manage to read of it. I will use macOS 10.x to test the scripts provided in the book and probably report if they are still relevant. I might skip the few commands and things I already know about how to use the terminal.

PREFACE AND TRIVIA:

  • Unix was created in 1969 in Bell Labs and rewritten in C in 1973. In 1974, it was licensed to universities (the so called BSD).
  • Unix is successful because of its portability, the fact that it's written in C, a high level language.
  • The Unix philosophy states that the relationship among different programs is what makes a great system. Even trivial programs can be combined to create great software, if they work well together.
  • An important theme of the book, is how programs work together and how they are used to create other programs, and how they fit in the environment where they run.
  • Use existing unix tools instead of writing programs that the same task clumsily.

CHAPTER 1: UNIX FOR BEGINNERS

  • Unix is a time sharing operating system kernel. A kernel is program that let users run their programs, controls resources and provides a file system. Time sharing means that the the use of the CPU is rapidly switched among different users and programs that seem to be running simultaneously.
  • Unix can also refer to the kernel along with other system tools such as editors, compilers, command languages.. etc.

1.1 Getting started:

  • Unix is full duplex, meaning when you type on the keyboard, characters are sent to the system which echoes them (send them back) to the screen. This echoing functionality is turned off when you type a password, for example.
  • Most keyboard characters are printing characters, but there are also control characters such as the RETURN key which signifies the end of input line. Control characters are invisible characters that control some aspect on the terminal.
  • Return has a key of its own, but most other control characters are typed by holding the control (CTL/CTRL/ctl) character and some other character. RETURN, for example, is is composed of ctl-m, backspace is ctl-h..etc (I don't know if these). Other important control keys include, BREAK and DELETE, which stop running programs immediately. This is the familiar ctl-c.
  • A few useful commands in a typical session in 1984 would include the following commands:
    • date: shows date and time.
    • mail: shows mail.
    • mail ahmed: sends mail to specified recipient.
    • d: delete mail inside mail
    • ctl-d: exit mail inside mail.
    • login: prompt you to login using the credentials you provide.
    • 'logout': can also be done using ctl-d.
    • who: shows who is logged into the system. tty... stands for the teletype (old word for terminal). It's the system name for a certain connection (user).
    • who am i: tells you which user you are.
    • ctl-u: kills a line (very useful, INDEED).
    • WHat the fuck are you talking about.
  • Type ahead: you can type while the terminal is outputting text. Your input gets mixed up with the output, but the system still interprets it correctly. COOOL!!!
  • To stop a program, I use mainly ctl-c and sometimes ctl-z. I still don't know the difference as of right now!
  • Chatting with other users of the same system can be done with e.g. write Ahmed. To exit chat, type ctl-d.
  • The manual: The book doesn't say much more than what I already know man. I need to work on this on my own.

1.2 Files and common commands:

  • Information is stored in files which have names, content, locations and metadata such as who owns them.
  • ed 101: ed is an old text editor, much older than both vi and emacs. Basic commands of ed:
    • a: start adding text.
    • .: By itself on a line. Stop adding text.
    • w <filename>: save to named file
    • q: quit ed.
    • 1,$p: prints to screen contents of file from line 1 to last line
    • g/<regular-expression>/p: same as grep.
  • Weird thing is that when I open a file with ed, I can't see any of the text previously written into it.
  • ls, which is used for listing files in a directory has many options. An interesting one is -t which lists files by time they were last modified and not alphabetically. -l list long info lines which include who owns the files and where they were created. -r reverses the order files are listed in. Options can be combined so ls -lrt lists long descriptions of files according when they were last modified in reverse order.
  • Options have to precede file names.
  • Options can be combined or left separate but this is not always the case.
  • Options in unix are not uniform between different commands, so -l means something in one command and a totally different thing in another commands.
  • cat and pr print the contents of a file or several files to screen. pr cut content into pages and put dates on top of pages
  • lp and lpr are used for printing. I don't know if this is still relevant!
  • File path manipulation:
    • mv: moves a file from one location to another or simply renames a file, or replace target with the moved file.
    • cp: copy the contents of a file into another.
    • rm: removes an existing file a warning a files doesn't exist.
  • file names should only be composed of underscores, letters and numbers...etc. Periods are used in extensions...
  • More file processing commands::
    • wc: spits out the count of lines, words and characters in a file or a group of files.
    • Grep: grep which is used to grab lines containing patterns originates from the ed command g/<regular-expression>/p. Used with the option -v, it grabs the lines not containing the provided regular-expression. ch re on grep is on chapter 4.
    • sort sorts a file line by line alphabetically. There are many options to make sort more useful.
    • cmp file1 file2: Used to check if two files are the same. Works on any type of file. It only shows in which line the first difference occurs.
    • diff file1 file2: used ti tell the differences between text files and where they occur.

1.3 Directories:

  • When you login, you are in your home directory. The working directory or current directory is the one where you are working now.
  • When a file is created, it by default resides in the current directory.
  • pwd tells you where you are located in the file system tree.
  • A pathname is the full name of a file starting at the root, something like /usr/you/junk.
  • When you try to run a command, the system looks inside the current directory and if it doesn't find it, it searches /bin, and if not then /usr/bin. If you run ls /bin /usr/bin, you get a list of commands like mv, cp, ed .. etc.
  • mkdir is used to create a new directory and rmdir removes an empty directory.
  • cd is used to changree directories. By itself it takes you to home directory. cd .. takes you one level back up towards the root. cd /usr takes you down to the user directory.

1.4 The shell:

  • Between you and the kernel sits a program called the shell or command interpreter. The existance of such a program separating you from the kernel has some advantages including:
    • Filename shorthands: batches of files can be processed all at once using certain patterns.
    • Input-output redirection: You can send your output to a file or gran input from a file or directly make the output an input to another program
    • Customizing the environment: you can create your own tools.
  • Filename shorthands: These are the basic shorthands found in SQL, regular expressions .. etc. * means any string of characters can be used for example to grab all files that start with 'ch' as in ch*, or all the files that end with '.pdf' as in *pdf. It's the shell that interprets * as a string of any length in the example ls *.pdf and not ls.
  • Other important filename shorthands include:
    • [...]: matches a single one of any of the characters inside the bracket and this can be done in one of the following notations:
      1. [01234546789]
      2. [1-46-9]
      3. [a-z]
    • ? is similar to * but it matches any single character instead of a sequence of characters.
  • IMPORTANT: Patterns only match existing filenames, so the following expression doesn't work: mv ch.* chapter.*
  • * is also very useful in path names. Think of the possibilities and what you can do with something like usr/*/calendar or usr/* etc.
  • To neutralize the special functionality of ? and * precede them with backslashes as in \* and \?. Now * and ? are just regular characters.
  • Input-output redirection: The terminal doesn't have to always be the input, output or both of a command. The results of a command such as cat can be output into a file. Here are some examples of conventions for redirecting input-output that are part of the shell:
    • cat f1 f2 f3 > temp.txt: Instead of catting the contents to the terminal, they are catted to the file temp.txt. > means put the output of the command into temp.txt. If the file exists, it overwrites it and if not it creates a new file with the name provided.
    • cat f1 f2 f3 >> temp.txt: Instead of overwriting the content of temp.txt, it adds the content to the end of it.
    • mail a b c < letter.txt: < means input the contents of letter.txt into a, b and c.
  • The use of temp files allows you to do many nifty things.
  • Pipes: or pipelines connect the output of a program to the input of another program without the need for temporary files. This is a fundamental contribution of the unix system. who | sort, for example, prints a sorted list of logged in users and ls | sort -r lists a revers-sorted list of files, and ls | grep .pdf lists file whose names include the pattern '.pdf'.
  • Pipes are versatile and anything that can be input into, or output that can be output from the terminal can be done to a pipe. You can chain as many programs in a pipe as you wish. Programs run simultaneously in a pipe, so they allow for interactivity.
  • I feel like a pipe including commands with optional filenames and arguments might cause confusion. Anyways, creating efficient pipes might need some training and getting used to.
  • Processes: You can string two programs together using a semicolon. date; who prints the date and list logged users. These programs run in sequence. Using &, the user can go ahead and start another program while the first program is still running. So by using wc upe.md &, you can start another program even thiugh wc is still running. When you use &, a number is printed to the screen. That number is process-id of the running process. A process is different from a program; it's an instance of a program. Two instances of the same program are two different processes. With an & at the end of a pipe, only the id of the last process is printed out. The command wait waits until all the process started with &.
  • The process-id can be used ot kill the process with the command kill <process-id>.
  • ps is used to list all the running processes. PID is the process-id. TTY is the terminal associated with the process and TIME is how much time the process used.
  • Processes might have parents and children (I won't go into this headache here).
  • nohup: nohup <command> & allows you to keep running a program after you log out of the terminal.
  • nice <command> & keeps running without taking too much resources.
  • A job can also be scheduled through the use of at. For example at 4pm echo ahmed. I am having a hard time running this command, because of time formatting. I keep getting the message "garbled time". I'll come back to it when I have an Internet connection.
  • Environment Tailoring: Unix can be customized to meet one's needs. The command ssty changes the kill line functionality from ctl-u to @ through the following command stty kill @. Once you logout or turn off the terminal, this change disappears, but to make it persist, you can use the .profile or bash_profile file. This can be discussed somewhere else. Some commands can be added to the profile, so that every time you login, some information is printed first like how busy the system is ..etc.
  • Shell variables: are shell properties that can be controlled and set by the user. They include
    • PS1: or prompt string and it can be set to anything. It usually has the user name and $. This can be icluded in .profile or .bash_profile as in PS1="Yeas dear?".
    • PATH: This is the most useful shell variable. It sets where the shell should look for commands. By default, the sell looks for commands in the current directory, then in bin, then in /usr/bin. This sequence of directories is called the search path which is stored in the shell variable PATH. PATH has a strange syntax where directories are separated by colons. This an example of a path that includes the current directory, /bin, /usr/bin and /usr/bin/games: PATH = .:/bin:/usr/bin:/usr/bin/games
    • There is also the strange {$PATH}, which I don't understand. More search paths can be added to the first one in this manner PATH=$PATH:/usr/news. This new search path is added to the value of PATH which is denoted here by $PATH. The value of any shell variable is obtained by preceding it with $.
    • . means the current directory, and a null value of PATH also means current directory.
    • You can create your own commands, collect them in a certain directory and add their search path to your PATH.
    • HOME: denotes the home directory.
    • TERM: is used by editors to manage screen and tells the editor what type of terminal you have.
  • You can have your own variables to store different information such as a certain path, so g=/usr/bin/games allows you to do something like cd g. g here stores the path to said directory.
  • export tells the shell that other programs can use sell or personl variables. This can be done in two ways:
    1. PATH = .:/bin:/usr/bin:/usr/bin/games: where the variable PATH is also defined and then exported.
    2. export PATH: where the previously defined variable is exported.
  • You can create your own commands by packaging existing commands. This can create some very powerful tools.

1.5 The rest of the unix system:

  • There is much more functionality in the unix system that's explained exhaustively in the manual. I need to learn more about how to use the manual.

CHAPTER 2: THE FILE SYSTEM

  • Everything in the Unix system is a file. To succeed using unix commands, one need to get familiar with its file system and how files are manipulated, structured..etc.

2.1 The basics of files:

  • Files are just byte sequences. The meaning of those bytes depends solely on the programs that interpret them. The system doesn't care what they mean or what structure they have.
  • The od -cb <filenames> commands print the bytes of a file as characters and their octal values. od stands for octal dump. This commands prints the byte content of a file, but doesn't print the end of the file. The kernel keeps track of the length of the file, so the end of the file is signaled by the fact that there is no more bytes to print out.
  • Programs retrieve the data in a file through a system call (OS voodoo) called read. Each time read is called it returns the next part of the file and also returns the number of bytes it reads. When, there are no more bytes, it returns zero, signaling the end of the file. Returning zero is interpretation-independent and a good way of telling the end of file without returning a special character.
  • More voodoo. I don't see the point of cat and cat -u

2.2 What is in a file?

  • The format or type of a file is determined by the program that interprets it. The file system and kernel have no idea what the type is, but it can make an educated guess through the command file <filename>.
  • To determine the type of a file file reads the first few hundred bytes of the file to tell its type. Reading the extension of the file only is not reliable. Some files have system-dependent magic numbrs that determine their types. In binary files, it is octal 410. In C files, file command looks for something like "#include" to determine it's a C file. Reducing and eliminating differences between file types is an advantage for the unix system. This makes the commands more versatile and everything in unix is a file. A binary, a c file, a text file..etc., they can all be treated in the same way. It's up to specific programs to deal with distinct types of files.

2.3 Filenames and directories:

  • Each running program, process, has a working directory. The process doesn't need to specify any pathname when it's interacting with the files residing in its working directory.
  • mkdir: creates a new directory.
  • du: prints a list of directories and subdirectories. With the option -a, this command can print all directories and files as in du -a, and the output of this program can also be grepped in a pipe to look for a certain pattern as in du -a | grep batata. If you don't specify a certain directory, this command assumes you mean the current directory.
  • In pure unix directories are files. An octal dump of a directory reveals a mix of binary and textual data. The textual data have the names of files inside the directory and the directory's name. I tried this on mac OS but couldn't od a directory.

2.4 Permissions:

  • Different users have different permissions. The super-user has more permissions. The system recognize users by u-id's or user ids. Two different login ids can have the same u-id. Users are also classed into groups. The system determines what you are allowed to do based on your uid or group-id.
  • The etc/passwd is the password file. It contains all the login information about each user of the system. (Tried this on mac but it didn't work exactly as described here but there is an etc/passwd and it has login information, but I couldn't see user id there). Each line in the password file follows the following format: login-id:encrypted-password:uid:group-id:miscellany:login-directory:shell The shell field is often empty because you're using the default shell /bin/sh. Miscellany contains anything, usually name, email or phone. The second field has a an encrypted form of the password. Login encrypts the password you provide and compares it to this encrypted form and if they agree, you are allowed in (Encryption and security are vast topics beyond the scope of this book and notes).
  • There are three kinds of permissions: read, write and execute. Each file has diffrent sets of permissions. As the owner of a file, you have a set of execute, write, read permissions. The group you belong to has its own set and everybody else has a different set. Listing files with the long option ls -l let you know what permissions a file has which are show in the first column of the list in the following form -rw-r--r-- . The first letter in this string stands for whether the file is a directory or not. If it were, the file would look this way drw-r--r--. The next 3 characters rw-encode the permissions associated with the owner of the file. It can both write and read the file, but the file is not an executable, hence the last character is slash; it would be an x if it were an executable and the owner (has the permission to execute it!) The next 3 characters are the set of permissions of the group you belong too and the last 3 are for everybody else.
  • The file etc/group encode group names, group ids and which users belong to which groups.
  • There is talk of set-uid, which I really don't understand. I will leave that to a course or a book on linux.
  • x in a directory doesn't mean it's executable, but it means it can be searched. It's possible to give a user the permission set --x. This means the file can be accessed if the user knows its name, but can't run ls on it to see what other files reside in it. Conversely, r-- means that one can run ls and see what files are in it, but can't use them.
  • The chmod command changes permissions on a file. This can be done in two ways:
    • Octal numbers: This is done by adding a 4 for read, 2 for write and 1 for execute. E.g to change permission of a file to -rw-rw-rw-, we use chmod 666 <filename>.
    • Symbolic description: This is complicated, but chmod +w this would remove writing permissions from all users and chmod -w remove writing permissions from all. How to do it individually for the 3 sets of users, who knows??!!!
  • Only super-users and the owner of a file can change its permissions. Even if somebody else allows you to write a file, the system won't allow you to change its permissions.
  • If a directory is writable, users can remove files from it regardless of what these permissions these files have. To prevent certain users from deleting files from a directory, make the directory non-writable.
  • Permissions and dates are not stored in files, but in a system structure called i-nodes.

2.5 Inodes:

  • "A file has several components such as name, contents, and administrative information such as permissions and modification times. Inodes contain administrative information about the files as well as system information such as how long the file is and where it resides in disc.. etc.
  • Inode has 3 times:
    • When contents were last modified.
    • When the file was last used (read or executed).
    • When the inode itself was last changed, such as when permission were set.
  • These different times can be examined through different options of the ls command such as -c (use time when file status was last changed) and -u (use time of last access, instead of last modification of the file) along with the long format option -l. (Check the manual for ls options).
  • Inodes are extremely important for the file system, because they are actually the files. Directory hierarchies are just a naming convenience. The system recognize files by their i-numbers, the numbers of the inodes that hold the files information. ls -i reports the i-number of a file.
  • od doesn't seem to work on macOS, but according to the book, 'od' on a directory prints a mix of binary and textual data where the text refers to file names and the binary refer to the i-numbers which are the only links between file names and their contents. A filename in a directory is called a link because it links a name a to an inode, and hence the data. The same i-number can appear in different directories. rm only removes the link to an inode. Only when the last link to a file is removed that the system removes the inode, and hence the file itself.
  • When the i-number is zero, that means the contents of the file is still there and there might be another link to it.
  • The ln command makes a link to an existing file with the syntax ln <old-file> <new-file>. This gives the same file two different names, allowing the same file to appear in two different directories. This seems like it creates a new file but all it does is creating a new name that points to the same file. Both names have the same i-number, which is COOOL!!!
  • The number that ls -l prints between the permissions and the owner of the file is the number of links to the file.
  • All links to an inodes are equally important. If a change is made to a file through a link, the same change will appear in all other links since all links just point to the same inode or file. The file again is not removed until the last link to it is removed.
  • It's easy to lose files and unless the system is properly backed up, be careful when messing with links.
  • While ln only creates a new link or pointer to the same old content, cp creates a new copy of the old file content.
  • mv is a little similar to cp and ln. While cp creates new centent and while ln creates a new link to the same inode, mv keeps the content and the i-number, but only change the name of the file.
  • ln, cp, mv can shuffle across directories, not just in the same directory. as in mv /usr/junk.txt tuts/l_unix/unix_programming_env/.

2.6 The directory hierarchy:

  • Being familiar with the directory hierarchy of a system makes using it easier and more efficient. This structure might have changed. it's fairly different on the mac OS system, but I'll stick here to the original hierarchy.
  • The top directory or the root is / and under it are the following directories:
    • bin: binaries such as who, ed, ls.
    • boot: bootstrap. Program that starts before the unix system is loaded.
    • dev: devices.
    • etc: et cetera or miscellany such as administrative files like passwd and groups.
    • lib: library. Parts of the C compiler.
    • tmp: temporary files. When programs are running, they create their own files or copy files they work on inside tmp. tmp is automatically cleaned when the system starts.
    • unix: The unix program, kernel..etc.
    • usr: user file system.
  • This is very similar to linux and it's better to invest time getting familiar with that directory hierarchy.

2.7 Devices:

  • Unix is or was special in that peripherals were run using files. There are references to these files inside the kernel that are converted to hardware commands that access these peripherals. Here is an example of what ls -l /dev would produce: crw--w---- 1 ahmed tty 16, 0 Aug 26 00:30 ttys000 This information is derived from the inode of the device file. It contains its internal name which consists of its type, either c (character) or b (block) and a pair of characters, its major and device numbers. The type is in the beginning of the permissions block produced by ls crw--w----. Discs and tapes were block devices, while terminals and others were character devices. The major device encodes the device type, while the minor number registers its instance. In our example 16 is the major number while 0 is the minor one
  • tty commands tells you which terminal you are using.
  • Many other incomprehensible details about booting, mounting, dismounting.. etc.

CHAPTER 3: USING THE SHELL

  • The shell is the program that interprets the requests you make to the system. A lot can be accomplished with the shell alone.

3.1 Command line structure:

  • The simplest command is made of one word such as ls, who or date. A command is usually terminated with new line, but it can also be terminated with a semicolon, so multiple commands can be present in the same line as in date; ls; who.
  • Because of operator precedence rules, a semicolon may introduce some issues especially in pipes. In date; who | wc you might want to get the word count of the output of both who and date, but you only get those of date. | has a higher precedence than ;. To solve this problem, group who and date with parentheses as in (date; who) | wc. This gives the word count of both date and who. Now, the outputs of the two commands "are concatenated into a single stream that can be sent down a pipe."
  • tee saves the a pipe's data into a file and its output as in (date; ls) | tee save.txt | wc. In this example, tee dumps data in save.txt and wc receives that same data. (I don't know what's the point of this).
  • & is another important command terminator. long-running-command & tells the shell to not wait for the command to end. It can keep running in the background while the user can keep using other commands of the shell. This terminator can be used with such commands as sleep ot do interesting things. sleep is a command that waits a specified time in seconds before it exists. It can be used to create reminders (asynchronous stuff??!!!). The following example should be self-explanatory: (sleep 5; echo tea is ready) & echo tea will be ready in 5 seconds. The precedence of & is higher than that of a semicolon, so the parentheses are essential in this example.
  • Since pipes are commands, & also applies to pipes.
  • Command arguments: commands can take arguments which are words separated by spaces or tabs. These arguments are usually files. Commands also take options which are arguments starting with minus signs such as -c.
  • Special characters such as >, <, | ,&, ; are not arguments but are used to control how the shell run commands.

3.2 Metacharacters:

  • These are special characters that are interpreted differently by the shell. The asterix * is the most commonly used metcharacter. It is used to search a directory and stands for any string of characters. Filename-matching characters, though, ignore files starting with dots, to avoid problems with . (current directory) and .. (parent directory).
  • There are many metacharacters. To prevent the shell from interpreting these characters as metacharacters, enclose them in single quotes ($ echo '****' instead of $ echo ****) or precede them with backslashes ($ echo \*\*\*\* instead of $ echo ****). Single quotes are less problematic than double quotes and the user needs the keep the distinction between the two. Backslashes are still the easiest way to avoid all confusion.
  • Another vital use of single quotes is that they allow for including newline in the printed text.
$ echo 'hello
> world'

will include a new line after hello.

  • Secondary prompt appears when you type newline inside quotes or double quotes. It is printed by the shell in a new line when it expects you to type more to complete a command. It is stored in the shell variable PS2 and can be changed and stored in .profile through this command:
PS2=">>>"
  • A backslash at the end of a line allows you to continue typing on the next line. The backslash is discarded in general but is retained inside single quotes:
$ echo abc\
> ddd\
> zz
abcdddzz

while:

$ echo 'abc\
> ddd\
> zz'
abc\
ddd\
zz
  • # is used for comments.
  • Some super nerdy talk of why echo and why not the newline supplemented by echo is a good idea!!
  • Shell metacharacters:
Character Description
> prog > file direct standard output to file
>> prog >> file append standard output to file
< prog < file take standard input from file file
` `
<< str here document: standard input follows, up to next str on a line by itself
* match any string of zero or more characters in filenames
? match any single character in filenames.
[ccc] match any single character from ccc in filenames.
; command terminator: p1;p2 does p1, then p2
& like ; but doesn't wait for p1 to finish
... run command(s) in ...; output replaces `...`
(...) run command(s) in ... in a subshell
{...} run commands(s) in ... in current shell(rarely used)
&1, $2, etc. $0...$9 replaced by arguments to shell file
$var value of shell variable var
${var} value of var; avoids confusion when concatonated with text;
\ \c take character c literally, newline discarded
'...' take ... literally
"..." take ... literally after $,`...` and \ interpreted
# if # starts word, rest of line is comment
var=value assign to variable var
p1 && p2 run p1, if unsuccessful, run p2
`p1

3.3 Creating new commands:

  • When a sequence of commands are run frequently, it would be a good idea to package them in a new command.
  • The book provides an example of a program that counts how many users are logged in the system. The following pipe does that job:
who | wc -l

This pipe is written into a file nuusing an editor or simply echoed into that file:

echo 'who | wc -l' > nu

The shell, named sh in unix, is then used to to run the program. It takes its input from the file nu as in:

sh <nu

It's ridiculous to type sh every time you want to run this type of file. This file needs to be turned into a shell file. Generally speaking, if a file is executable and contains text, the shell assumes it is a shell file. Change permissions of the file to make it executable:

chmod +x nu

Now nu is a shell command that can be run just by typing nu.

  • The way the shell runs nu is to create a new shell process as if sh nu were typed. This shell process is called a sub shell (a shell process invoked by the current shell).
  • The new command is only available in your current directory. To make it available to your system or your repertoire, put it in a specific directory and add it to your PATHin your .profile or .bash_profile. There are different conventions for this. The book suggests a certain usr/you/bin.
  • To create a temporary sub shell, you can use the following command:
sh # This creates a sub shell that is different from the parent shell

You can do work in this shell and when you are done, just run the command exit. (The book mentions ctl-d, but that doesn't work in my machine.)

3.4 Command arguments and parameters:

  • How do we supply our new command with arguments such as files to work on? The book provides the example of a command cx that make files executable. Arguments are denoted by a sequence of symbols $1 to $9 where $1 stands for the first argument, $2 stands for the second argument and so on until $9. If our shell file contains the following command chmod +x $1, then the command will work on the first argument, and the sub-shell replaces $1 with nu in the command cx nu.
  • How to handle more than one argument? You can do the naive chmod +x $1 $2 $3 $4 $5 $6 $7 $8 $9. This will work if you have less than 9 files. The unused arguments will be empty strings. If you have more than 9 arguments, the program fails. This can be solved by the the following nifty convention:
chmod +x $*

This way, your command will work no matter how many arguments you have.

  • Files are not the only type of arguments you can add to your commands. Consider the example provided by the book where grep is used to grap text from files:
echo 'grep $* usr/lib/phone_book' > search_phones

This command returns all lines containing the grepped pattern. There is only one little problem. When the grepped pattern is made of more than one word, it becomes more than one argument. Our command gets two arguments instead of one resulting in the failure of the program. To fix this problem, $* needs to be placed inside double quotes. While single quotes doesn't allow any character to be interpreted by the shell, double quotes allow $, \, and ... to be interpreted instead of just read as regular characters. The correct command then becomes:

echo 'grep "$*" usr/lib/phone_book' > search_phones

Now, the strings that replaces "$*" will be a single argument even if it contains a blank.

  • Argument $0 is the name of the program itself. (I saw this in the C book).

3.5 Program output as arguments:

  • Arguments can be generated explicitly, with metacharacters like * or they can be generated by running programs. The output of a pgromar itself can be the argument to a command. You can get the output of a program by enclosing it in back quotes `...`'. The following command:
echo the date is `date`

prints the date, not the string "date". The command date runs and its output becomes one of echo's arguments.

  • Imagine a command that would grab a title of a to be created file from another file:
touch `cat lala`

3.6 Shell variables:

  • Shell variables are also called parameters. They include variables such PATH and HOME that we saw earlier. They are different from positional arguments such as $1 and $2.
  • Shell variables can be created, accessed and modified. Here is an example of modifiying a shell variable, the PATH:
PATH=:/bin:/usr/bin

There must be no spaces around the equal sign and value must be a single word with no blanks. If it contains shell metacharacters, it should be enclosed in single quotes.

  • The value of a parameter can be accessed by preceding it with a dollar sign.
  • Conventionally, special shell variables are in uppercase such as PATH and HOME. You can create your own shell variables to store paths or strings and give them lowercase names.
  • set prints out all shell variables.. There are way too many. To print specific variable values use echo as in echo $PATH.
  • The value of a variable is associated with the shell that created it. It is not passed to its child shells.
  • A shell file (dicussed at 3.3) can't change the variables of their parent shell. But there is a workaround where commands in files are run using the dot command . Let's say we have a file games_dir containing a command to add the games directory to the PATH variable. Running this file can be done as follows:
. games_dir

games_dir need not be made executable, because the dot comand doesn't actually execute the commands in the file. It merely read them and the current shell executes them. The disadvantage is that the file can't have arguments $1, $2.. etc.

  • There is a different way to change a shell variable by sub shells. (I couldn't get it to work and thought the syntax was bad. I will come back to it when need for it rises or see if there is a more modern way of doing that).
  • To make shell variable values available ot sub shells, the export command is used. There is no way of and no reason for making sub shell variables visible to parent shells.
$ x="2000"
$ sh 			# start subshell
$ echo $x
				# blank; $x not avaiblable
$ exit 			# leave subshell
$ export x 		# make x available to subshell
$ sh 			# start subshell again
$ echo $x 		
2000			# VoilĂ ! x is here! It was exported, don't you know?!

export is Weird, so only export variables you need to use in all shells and subshells, especially special ones such PATH and HOME.

3.7 More on I/O redirection:

  • When an error occurs, an error message is printed on the terminal. Each program has 3 default files established when it starts. These are file descriptors. They are numbered by small integers and are 0, standard input and 1, standard output and 2, standard error.
  • To redirect standard output to some file, you can use the following construct:
diff file1 file2 file3 2>errorfile

There should be no space between 2 and >.

  • Redirecting both standard error and standard output to standard output is done with the construct 2>1&. Redirecting both standard output and standard error to standard error can be done with the construct 1>2&
  • Commands can be packaged together with their arguments in the same file so that a shell file can be self contained. Such a shell command is called a here document. It means the input is right here and not somewhere else. The following is an example of such shell file:
grep "$*" << End
baba
caca
dada
wawa
wawa
End

The construct << End means everything until the world "End".

  • The next table lists the various shell I/O redirections: | Character | Description | --- | --- | | >file | direct standard output to file | >>file | append standard output to file | <file | take standard input from file | p1 | p2 | connect standard output of p1 to standard input of p2 | ^ | obsolete synonym for | | n>file | direct output from file descriptor n to file | n>>file | append output from file descriptor n to file | n>&m | merge output from file descriptor n with file descriptor m | n<&m | merge input from file descriptor n with file descriptor m | <<s | here document: take standard input until next s at beginning of a line; subtitute for $, '...' and ''' | <<\s | here document iwht no substitution | <<'s' | here document iwht no substitution

3.8 Looping in shell programs:

  • Being a programming language, the shell has variables, loopings, conditionals.. etc. Looping over filenames is very common and is done with the shell for statment; (might be only one to type in terminal rather than in a a shell file). A for statment follows this general syntax:
for var in list of words
do
	commands
done

The following program prints filenames one per line:

for i in *
do
	echo $i # notice we are accessing the value here. This is a little different from regular programming languages.
done
  • It might be desirable to compress this expression into a one liner. do and done are only recognized when they appear after a newline or a semicolon.
for i in *; do echo $i; done
  • Loop can be done over not only filenames, but from anything, a list that you type for i in 1 2 3 4; ... or for i in `cat ...` ; ....

3.9 Bundle: Putting it all together:

  • (Didn't quite get it! Might come back later!)

3.10 Why a programmable shell:

  • Some praise of the UNIX shell and enumeration or why it is or superior. It's not just a program for running commands, but a complete language that is both easy to use, conventient and can be extended to do much much more.

CHAPTER 4: FILTERS

  • Filters are a "family of UNIX programs that read some inoput, perform a simple transformation on it, and write some ounput." They include programs we've seen before such as wc, grep, sort. In addition to these filters, this chapter discusses programmable filters such sed and awk which are more generalized and more sophisticated.

4. The grep family

  • grep follows this general syntax:
grep pattern filename

It searches standard output or named files for pattern and print each line containing that pattern. Here are examples of grep with some of its common options:

grep -n potato *.java # Output lines with their numbers
grep From $MAIL  # input from a shell variable
grep potato s | grep -v mashed # Get line where non-mashed potato appears
grep -i ahmed phonebook # ignore case
who | grep ahmed # pipe as input
ls | grep -v *.c
  • grep works on a restricted form of regular expressions. It was invented in an evening through a surgery on the ed editor. Unfortunately, regular expression metacharacters are not the same as shell metacharacters. They are generally similar but there are many differences. To avoid potential problems, enclose grep patterns with single quotes.
  • Here is a brief review of these regular expressions along with a detailed table:
    • ^ anchors the pattern to the beginning of a line and $ anchors it to the end.
grep '^lala' baba # -> lala is sick
grep '$lala' baba # -> There is no one but lala

(End anchor doesn't work on mac. Read somwhere it has to do with how newline is constructed)

  • grep supports character classes just like shell as in [a-z] which matches any lower case letter. If the character class starts with a circumflex ^, the pattern matches any character except those in the class. [^A-Z] means any character but uppercase letters. There is mo backslash in character classes, to match square brackets and minus signs you can use [][-]
  • A period . matches any single character.
  • The closure applies to the previous character, metacharacter or character class, and collectively matches any number (zero or more) matches of the character, metacharacter, or character class:
    • x* matches a sequence of x's as long as possiblexxxxxxxx.
    • [1-2]* matches a numerical string 12345567.
    • .* matches everything until a new line das#(*@ _)*&?221HHbB.
    • .*x matches everything up to and including the last x in the line. Two important observations about the closure operator:
    1. Closures apply to only one character and not a sequence. xy* is matched by xyyyyy and not xyxyxyxy.
    2. Any number of matches includes zero. If you want at least one character matched, duplicate the character or metacharacter as in [0-9][0-9]*
  • No grep regular expression matches a new line. grep re are applied to each line individually.
  • grep is the oldest of a family that also includes egrep and fgrep. fgrep can search many literal strings simultaneously, egrep evaluates true regular expressions with an "or" operator and parantheses to group expressions.
  • egrep and fgrep have the option -f that allows them to take their input from a file that contains the lined separated patterns to be searched.
  • egrep has a few more features than grep:
    • parantheses can group expressions, so (xy)* can match an empty string, xy, xyxy or xyxyxyxyxy.
    • Vertical bar | is an "or" operator. (today | tomorrow) and to(day|morrow) match either today or tomorrow.
    • egrep also has two closure operator + and ?. x+ matches one more x's, while x? matches 0 or 1 x's but no more.
  • This one is cool. It searches a dictionary for words that contain all vowels in alphabetical order
egrep '^[^aeiou]*a[^aeiou]*e[^aeiou]*i[^aeiou]*o[^aeiou]*u[^aeiou]*$' dict 
  • fgrep can't interpret metacharacters, but can search for thousands of words in parallel.
  • grep is older has more options than egrep and has tagged regular expressions (WHAT"S THAT???????) egrep is much faster, but the two can combined in the same program.
  • Table. grep and egrep Regular Expressions (Decreasing order of precedence):
Pattern Description
c any non-special character c matches itself
\c turn off any special meaning of character c
^ beginning of line
$ end of line
. any single character
[...] any one of characters in ...; ranges like a-z are legal
[^...] any single character not in ...; ranges are legal
\n what the nth (...) matched (grep only) (WHAT?)
r* zero or more occurrances of r
r+ one or more occurrances of r (egrep only)
r? zero or one occurrances of r (egrep only)
r1r2 r1 followed by r2
`r1 r2`
\(r\) tagged regular expression (grep only); can be nested
(r) regular expression r (egrep only); can be nested
no regular expression matches a newline.

4. Other filters

  • Unix has a bunch of small filters with a ton of functionality. Some of the more common filtes include:
  • sort as seen before sort input by lines according to ASCII order. A rich set of options makes sort very useful:
    • -f: folds characters, hence ignoring case.
    • -d (dictionary): ignores characters other than letters, numbers and blanks.
    • -n: sorts by numeric value.
    • -r: reverses sorting order. For example, ls -s | sort -nr reverse sorts files by size.
    • -o: specifies an output file (which can be one of the input files).
    • -u: discards all but one of identical line groups (WHAT??) sort acts mainly on lines, but it can also be directed to certain fields. +m means the comparison skips the first m fields, while +0 starts at the start of the field:
ls -l | sort +3n # sorts files by byte count numeric value

multiple keys can be used for sorting as in

sort +0f +0 -u filenames

This line seems to first sort line alphabetically ignoring case distinctions with +0f, then do secondary sorts on alphabetically equal lines based on case with +0 and then -u discards adjacent duplicates.

  • uniq functions in a similar way to the -u option of sort. It discard adjacent duplicates regardless of weather lines are sorted or not. It works great for removing multiple blanks. Options include:
    • -d: prints only duplicated lines.
    • -u: prints inly unique lines.
    • -c: counts the number or occurrances of a line.
  • comm compares files. Given twp sorted files, comm prints 3 columns, first column as only lines from f1, column 2 only files from f2, and column 3 has files from both. Options include the possibility of suppressing any of the output columns. -2 suppresses the second column, and -23 suppresses columns 2 and 3. These options can be useful is comparing directories.
comm -12 f1 f2 # prints lines in both files
comm -23 f1 f2 # prints lines only in first file
  • tr transliterate characters. This is mainly used to convert between character cases.
tr a-zA-Z # this needs some experimentation

4. The stream editor sed

  • sed is based on the ed editor. Its commands follow this general format:
sed 'list of ed commands' filenames ...

This command reads lines from the input files one line at a time, apply each of the commands in the list in order, and writes its edited form to the standard output. As an example, the following the command will change each instance of unix to linux:

sed 's/unix/linux/g' filenames >output
sed 's/unix/linux/g' file >file # This is bad bad, a big NO NO
  • s stands for substitute. It matches does the regex matching.
  • g makes the change globally for every word in the line, not just the first occurrance.
  • sed does not change its input, but write the altered form to standard output.
  • Don't redirect the output of sed to its input file. You need to use a temporary file for that or the overwrite program from chapter 5.
  • It is necessary to always enclose sed arguments in single quotes because of similarities with shell metacharacters.
  • Here are more examples of how sed works:
sed 's/.*\t//' # deletes all characters up to and including a rightmost tab
sed 's/ .* //' # replaces blank and everything that follow it up to another blank by a single blank
who | sed 's/.* //' # grabs user's name

-(typing a tab in osx terminal in sed is weird. It's done through: ctrl-V followed by TAB key)

  • A sequence for indenting lines in a file can be done through these different versions of the same program:
sed 's/^/\t/' file

This script puts indents even in empty lines. A better version that doesn't indent empty lines is:

sed '/./s/^/\t/' file

Prefixing a pattern in the form /pattern/ allows sed to select which lines to act upon. The /./ matches lines that have least one character other than a newline. This will still output all lines, either changed or unchanged. It is also possible to do commands on lines that don't match the selection pattern with the use of !.

sed '/^$/!s/^/\t/' file

/^$/ means the beginning is adjacent to the end, meaning it's an empty line. (A little bug I encountered is that the first empty line for some reason is not recognized as empty)

  • sed allows to print just the x leading lines with the command xq (3q, 10q ... etc.).
sed 3q file

This will print the first 3 lines.

  • A chain of sed commands must be separated by a newline:
sed 's/^/\t/
	3q'
  • With the option -f, sed can get its commands from a file as in:
sed -f commandfile sourcefile
  • q prints its input up to and including the first line matching pattern:
sed '/pattern/q' file
  • d deletes every line containing the matching pattern:
sed '/pattern/d' file
  • With a combination of the command p which prints lines containing the given patterns and the option -n for turning off the usual sed printing sed can act like grep:
sed -n '/pattern/p' file  
sed -n '/pattern/!p' file # this acts in the same way as grep -v
  • Philosophy time: why do we still have both grep and sed? grep is older and simpler and sometimes does things sed can't do!
  • A new line can be inserted after each line with:
sed 's/$/\
/' file
  • The next example is of a command that replaces each string of blanks or tabs with a newline, splitting input into single words
sed 's/[ \t][ \t]*/\
/g' file

[ \t] means a blank or tab; [ \t][ \t]*, as seen before, means 1 or more. ([ \t]* is for zero or more, but we have 2 [ \t]*'s).

  • Pairs of regular expressions or line numbers can be used to select ranges of lines to be acted upon:
sed  -n '3,9p' file
sed '1,10p' file
sed '$d' file # delete last line
  • Line numbers don't reset at the beginning of file, so numbering flows from one file to the next.
  • sed can't read backward. Once a line is read, it's gone forever.
  • sed also allows outputting to multiple files, possibly different patterns to different files:
sed -n '/pat/w file1
	/pat/!w file2' filenames...
  • sed has its limitations but it's fast and efficient.
  • (I skipped a few things in this section. sed needs its own notes! I might work on that in the future)

4. The awk pattern scanning and processing language

  • Some of sed limitations are remedied by awk. While sed is based the ed text editor, awk is based more on the C language and its usage is similar to that of sed:
awk 'program' filenames
  • The structure of 'program' itself is different:
pattern {action}
pattern {action}
...

awk reads each line, and for any pattern it finds, it performs the corresponding action. It doesn't change its input.

  • Patterns can similar regex like in egrep or contain C-type conditionals. The following does the exact same thing as egrep, it prints lines matching the regex pattern:
awk '/regex/ { print }' filenames
  • Both patterns and actions are optionals, so using the pattern only has an identical output to egrep. If you you omit the pattern, you get the action performed on every line of the input.
  • awk's program can be read from a file with the -f option just like sed:
awk -f programfile filenames

Fields:

  • awk splits each line into fields of non-blank strings separated by tabs or blanks
  • These fields are called $1, $2, $3, ... $NF where the variable NF's value is the number of fields. Only fields in awk have are preceded with $. Other variables are not. To only print name of logged in users and the time they logged in, we use the following:
who | awk '{ print $1, $5 }'

To sort the output of the previous command, one can pipe it into sort and rearrange fields as in:

who | awk '{print $5, $1}' | sort
  • These operations are easier to use in awk than in sed, but awk is much slower to start and perform these operations.
  • Field separator can be changed with the option -F as in:
sed -n '6,13p' README.md | awk -F: '{print $2}'

This command changes the separator to a colon :.

  • Tabs and blanks are special as separators. They are bey default separators and leading blanks and tabs are discarded. If anything other than blanks is set as a sperator, then leading separators are counted to determine the fields. this applies to tabs as well

Fields:

  • You can print line numbers using NR:
awk '{print NR, $0}' filenames

$0 stands for the unchanged line in its entirety.

  • To have more control over how printing is done, use printf which is similar if not identical to C's printf:
awk '{ printf "%4d %s\n", NR, $0}' filenames

with printf you need to add newline and you can for example specify the length of a number field with something like %4d.

Patterns:

  • There are several ways of checking if a field is empty:
awk -F: '$2 == ""' Filenames 		# empty field
awk -F: '$2 ~ /^$/' Filenames 		# matches empty line
awk -F: '$2 !~ /./' Filenames 		# does't match any character
awk -F: 'length($2) == 0' Filenames # length is zero
  • Patterns are generally used for data validation. Regular expressions are enclosed in slashes and matching is based on the operator ~. ! is for negation.
  • substr(s,m,n) can be used to select substrings as in:
date | awk '{ print substr($4, 1, 5)}' # produces something like 02:00

The BEGIN and END patterns:

  • BEGIN actions are executed before the first line is read. Initialize variables, print headings.. etc.
  • END actions are executed after the last line is read. Can e.g. for printing the number of lines with NR.
awk 'BEGIN {FS = ":"}
{print $2}' filenames
awk 'END {print NR}' filenames

Arihtmetic and variables:

  • awk does computations on its input making powerful. The following sums the values of the first column:
awk '{s += $1}
END {print s}' filenames

In this example, there was no need for Initializing the s variables. awk variables are automatically Initialized to 0 by default.

  • Strings can also be stored in variables as in {s = "abc"}. They are also initialized to empty strings. The context determines whether a variable is treated as a number or string.

  • Next table is of awk built-in variables. | Variable | Value | --- | --- | | FILENAME | name of current file | FS | field separator character (default blank & tab) | NF | number of fields in input record | NR | number of input record | OFMT | output format for numbers (default %g) | OFS | output field separator string (default tab) | ORS | output record separator string (default newline) | RS | input record separator character (default newline)

  • As for operators, they are similar to those of C and most other C-like languages. The one exception is tilde ~ and !~ used for matching regex.

Control flow:

  • The following is fizzbuzz in awk. It shows how control flow works:
awk '
BEGIN {
	for (i = 1; i <= 100; i++){
		if (i % 15 == 0) print"fizzbuzz"
		else if (i % 5 == 0) print "buzz"
		else if (i % 3 == 0) print "fizz"
		else print i
	}
}' $*
  • This is very similar to how C-like languages do thing. The few things that are different include the keywords BEGIN and END. When not preceded by BEGIN, a comparison like NR > 0 is a equivalent to if (NR > 0). This seems like a python with curly brackets or groovy
  • There is also a while statment that is syntactically identical to those found in C languages.
  • You can break out of a loop using break. continue resumes iteration. exit transfers execution immediately to the END pattern. next jumps to the next input line.
  • Statements can be separated by new lines or semicolons.

Arrays:

  • It's easy to work with arrays in awk. They are similar to python's lists. They resize and are declared and initialized automatically as in the following example:
awk '{ line[NR] = $1}
END { for (i = 0; i < 10; i++ )
	print line[i]
}' $*

Built in variables:

Variable Usage
cos(x) cosine
exp(x) exponential
getline() reads next input file line, returns 0 if end of file, 1 otherwise
index(s1,s2) position of string s2 in string s1; returns 0 if not present
int(x) integer part of x; truncates toward 0
length(x) get length of string s
log(x) natural logarithm
sin(x) sine
split(s, a, c) split string s using separator c into array a; returns length of a
sprintf(fmt, ...) print according to specification fmt
substr(s, m, n) n-character substring of s beginning at position m

Associative arrays

  • Values other than integers can be used as subscripts in associative arrays. Associative arrays are hashed key value pairs. The follwing program sums values whose first field is identical:
	# list of pairs with duplicate keys
	# a 1
	# b 2
	# b 2
	# d 4
	# d 3
	# a 0

awk '{sum[$1] += $2}
END for (i in sum) {print sum[i]}
'
  • The for ... in: loop is not the same as the regular array for loop. It doesn't iterate over the elements of the array but it iterates over the subscripts but the result is similar, although it might require sorting after the loop is done.
  • The following program outputs the word frequency of its input:
awk '{ for ( i = 1; i <= NF; i++ ) freq[$i]++} # iterates over line fields
END {for (w in freq) print w, freq[w]}
' $* | sort +1 -nr | sed 20q # 20q only print first 20 line

interacting with the shell

  • Interacting with the shell can get a little tricky. Interactively choosing what field number to print out can be done in different ways:
awk '{ print $'$1'}' # $1 is outside quotes so it's interpreted as a shell variable, an argument to this program
awk "{ print \$ $1 }" # Remember from shell var table that \ and $ are interpreted inside double quotes
  • These scripts can't can only take standard input and can't get their input from files. That requires shell programming from the next chapter.
  • (Skipping the following calender subsection. I prefer to use small atomic filters piped together as the next section recommend).

4. Good files and good filters

  • It's more common to use awk and other filters in small scripts to do a single thing, following the unix philosophy. Small scripts are also piped together to solve bigger problems.
  • The output of unix programs follow a general easy to understand format that can also be easily input into pipes and other programs. That output is divided into lines, divided into fields by tabs or blanks. Following the general way unix programs are written can make your scripts easy to use and effective.

CHAPTER 5: SHELL PROGRAMMING

  • (I will try to create my own little programlets that imitate the book programs if they are not time consuming).
  • The shell is also a programming language, that's why it has too many details.
  • The authors Admit that the manual can be cryptic, so they focused on examples rather than listing features of the shell.
  • Shell programs should follow better standards and be robust; handle bad input and gives meaningful error messages.

5.1 Modifying a shell program:

  • The book suggests modifying the cal program, allowing it to take month names as arguments while preserving its original behavior and functionality.
  • The case statement allows you do take different numbers and types of arguments; it similar switch statments in C and it follows the following general fomrat:
case word in
	pattern ) command ;;
	pattern ) command ;;
	...
esac

This statements allows you to execute a command if word matches the first one of the listed patterns.

  • The following is a little program based on wc with the added functionality of selecting the first n lines for which words are to be counted:
 #wcp: wc on steroids; also counts words on first n lines

case $# in
	[0-1]) /usr/bin/wc $1; exit;
esac

case $1 in
	[1-9]|[1-9][0-9]|[1-9][0-9][0-9]
	) head -n $1 $2 | /usr/bin/wc;;
	* ) echo "That ain't gonna work, mate!";;
esac

Notice how the original system wc is called with it's full path /usr/bin/wc to prevent the system from confusing it with the wc we just created. Relying on the old name while keeping the old functionality allows the users to avoid learning new tricks and new command names.

  • The shell variable $# denotes the number of arguments a file was called with. Other special shell variables are included in the following table: | shell variable| usage | --- | --- | | $# | the number of arguments | $* | all arguments tp shell | $@ | similar to $*; see section 5.7 | $- | options applied to the shell | $? | return value of the last | $$ | process-id of the shell | $! | process-id of the last command started with & | $HOME | default argument for cd command | $IFS | list of characters that separate words in arguments | $MAIL | file that, when changed, triggers "you have mail" | $PATH | list of directories to search for commanda | $PS1 | prompt string, default '$ ' | $PS2 | prompt string for continued command line, default '> '

  • One nifty shell command is set. When called without arguments, it shows the values of shell variables. When called with an argument, it resets the values of $1, $2 .. etc. with the values of parts of that argument. The following example clarifies such usage:

$ set `date`;
$ echo $1
Mon
$ echo $2
Sep 

Some of set's more prominent options include -v and -x They echo commands as they are being processed by the shell.

  • This patterns might be confusing [1-9][0-9][0-9]. This is how number cases are dealt with in the shell; a little similar to regex. Another similar pattern is [Jj]an* which should be self-explanatory and is the same as [jan*|Jan*]. This table lists the shell pattern matching rules:
pattern rules
* match any string, including
? match any single character
[ccc] match any of the characters in [ccc]
"..." same as '...' ; match ... exactly, protects soecial characters
\c match c literally
`a b`
/ In filenames, matched only by an explicit / in the expression. In case matched like any other character.
. as the first character of a filename, is matched by an explicit . in the expression.
  • It's unfortunate that shell pattern matching is different from regex. Some of the differences are mere design mistakes, while others make a lot of sense. The book offers the following examples where the regex is too clunky and would be really clumsy to type on the terminal. The terse shell expression *.php would've been replaced by the hideous ^?*.php$.

5.2 Which command is which:

  • It is easy to get a version of a command that is not what you expect when you make your own private versions of general commands like ls or cal. This has to do mainly with the components of the shell variable PATH. Remember that the system searches the path sequentially; the executable found in the first directory of the path is what gets run.
  • The following is the "pseudo-code" of a shell program that loop over the PATH directories and search for the named executable file.
for i in each component of PATH
do
	if given name is in directory i
		print its full pathname
done
  • The PATH should first be separated by spaces. colons need be replaced by blanks. A null string is the same as a dot in the shell which means the current directory. The following sed script can do this:
echo $PATH | sed 's/^:/.:/
				  s/::/:.:/g
				  s/:$/:./
				  s/:/ /g'

The null can be either in the middle or either end of the PATH string. Those sed commands can be run separately through pipes, but as you can see, they can also be done this way.

  • The test command can tell you if a file exists in a directory. It has the following options: -f: file exists and is not a directory -w: It's writable. -r: It's readable. testt doesn't do anything besides returning an exit value. If the exit value is 0, it's true. Otherwise it's false (we know this already). Every time you run a program, its exit value is stored in the shell variable $? as seen in one of the table.
  • If statement in the shell follows the following genral format:
if command
then
	commands if condition true
else
	commands if condition false
fi

The newline or a semicolon are necessary before then, else and fi.

  • case statements can be faster than if. They do pattern matching directly in the shell, while if run commands and can do extra stuff case can't. Basically if you can use both, prefer case.
  • The following is the which program that fetches a program in the PATH list of directories:
case $# in
	0) echo 'Usage: which command' 1>&2; exit 2
		;;
esac
for i in `echo $PATH | sed 's/^:/.:/
				  s/::/:.:/g
				  s/:$/:./
				  s/:/ /g'`
do
	if test -x $i/$1
	then 
		echo $i/$1
		exit 0
	fi
done
exit 1
  • This is a beautiful program and the way it was used to teach features of the shell is ingenious. I bet Kernighan wrote this chapter.
  • 1>&2 redirects an error message to standard error so it doesn't get lost in the pipeline.
  • Two interesting shell operators that can save a lot of keystrokes in conditionals are || and &. Here are two examples that illustrate their use:
test -f somefile || echo somefile "does not exist"
test -f somefile && echo somefile "exists, indeed"

If the first command's return status is 0, the command to the right of || is ignored; if it were 1, the command to the right is executed. The command to the right of && will only run if the test evaluates to true.

5.3 while and until loops:

  • The while loops follows this general formula:
while command
do
	loop as long as as command returns true
done

while the until command can be represented by:

until command
do
	loop as long as as command returns false
done
  • The following examples illustrate these constructs use:
while sleep 60
do
	who | grep johndoe
done

This program will keep checking if johndoe every minute and print his info when he logs in. If he is already logged in, or just has just logged in after this program starts, you will have to wait 60 seconds before this program tells you that. It will aslo keep printing johndoe information as long as he is logged in. The next program is more interesting; it keeps looping and checking if johndoe is logged in. Once he is logged in, it prints his info. Grep returns an exti status 0 (true) which signals that the loop stops:

until who | grep johndoe
do
	sleep 60
done
  • The book goes on in describing a program that keeps track of all system users. A few thing to note:
  1. the shell variable $$ is used to distinguish temporary files (that reside in the /tmp directory) based on which processes run them as in newfile=/tmp/johndoe.$$. When you print this file in the shell, you'd get something like johndoe.40301 where 40301 refers to the id of the process running this file.
  2. A forever loop: can be done with the following expression:
while true
do
	echo something
	sleep 3
done

An even shorter and more efficient construct that doesn't use a command from the file system is : which can be used this way:

while :
do
	echo something
	sleep 3
done
  1. The PATH is usually fixed in the shell file so that it looks only within that path and nowhere else. This prevents running unwanted programs with the same name as the current one that reside somewhere else in the system. So the path that you specify in your shell file overrides the actual system-wide PATH.
  • The use of braces with variables has special meanings. the following construct:
t=${1-60}

This is similar to default variable values in languages like python. It means set t to the given argument $1, or to 60, otherwise. Another rather self-explanatory example is:

echo $varx
echo ${var}x

these two variables are different. The first example prints whatever is contained by the variable varx. The second one prints the contents of the variable var and append 'x' to its end.

echo ${something?I am sick}

In then above example the shell will try to print the variable something, but if it's not defined, it prints out the message following ? sign or a default message if nothing follows the ? sign. Similar constructs are illustrated by the following examples:

echo ${junk-thing}  # evaluates to junk if defined, otherwise thing.
echo ${junk-$thing} # same but to evaluates to the value of thing.
echo ${junk=thing}  # sets junk to thing.
echo ${junk=$thing} # sets junk to the value of thing.
echo ${junk+thing}  # thing if junk is defined, otherwise nothing.

5.4 Traps: catching interrupts

  • DEL sends an interrupt signal hungup sends a hungup signal. Book doesn't expain what a signal is. Most common signals include:
shell number Meaning
0 shell exit (for any reason, including end of file)
1 hangup
2 interrupt
3 quit (ctl-: causes program to produce core dump)
9 kill (cannot be caught or ignored)
15 terminate, default signal generated by kill(1)
  • (I have no idea what the rest of this section is saying, trap??!! Trap what??)

5.5 Replacing a file:

  • The -o option is used to overwrite files. It allows you, for example, to sort a file in place:
sort file1 > file1 # This is extremely wrong
sort file1 -o file1 # This works!!

(BORED!!!! My original intent was to get a better knowledge of the shell and grep, sed and awk!! Things are getting a little too tough. I might come back to the rest of the book. These 4.~ chapters covered a lot of material that needs to be reviewed).

Releases

No releases published

Packages

No packages published