Bob@localhost [/home/Bob]
$ cat phone.list
Smith, Terry 7-7989
Adams, Fran 2-3876
StClair, Pat 4-6122
Brown, Robin 1-3745
Stair, Chris 5-5972
Benson, Sam 4-5587
Bob@localhost [/home/Bob]
$ cp phone.list phone.list2

Bob@localhost [/home/Bob]
$ sed '/Brown/s//Browne/' phone.list2 >> phone.list2

Bob@localhost [/home/Bob]
$ cat phone.list2
Smith, Terry 7-7989
Adams, Fran 2-3876
StClair, Pat 4-6122
Brown, Robin 1-3745
Stair, Chris 5-5972
Benson, Sam 4-5587

Smith, Terry 7-7989
Adams, Fran 2-3876
StClair, Pat 4-6122
Browne, Robin 1-3745
Stair, Chris 5-5972
Benson, Sam 4-5587
Bob@localhost [/home/Bob]
$ cp phone.list phone.list3

Bob@localhost [/home/Bob]
$ sed '/Brown/s//Browne/' phone.list3 > phone.list3

Bob@localhost [/home/Bob]
$ cat phone.list3

Bob@localhost [/home/Bob]



Can someone please help me understand how this works.

When a double redirection is given in the command it displays <content> along with the updated content appended to it. It is mentioned in the book(Unix Applications Programming Mastering the Shell by Ray Swartz) that the shell always scans a command line for shell meta characters before creating a process to execute the command. As a result (when single redirection is used), the shell sets up the redirection, which erases the contents of phone.list, before sed can get a chance to look at it.

With this mentioned in the book, when we give double redirection ">>" wouldn't the shell first scan the file phone.list2 and then perform the sed operation. If it does so then shouldn't we be getting a different output.

Please explain. Thanks in advance.

All the ">>" redirection does is open the file at the end of the file, any output will then be appended to the end.

It is asking for trouble (as in doing this is an error) to try to process that file AND append to the end in the same command sequence. If you have a large enough file (due to buffering issues), the new data being appended will be read to be processed - thus forming an infinite loop, and a constantly growing file.

It isn't consistent - because if the file is small enough to fit in one buffer then the sed command could fully process the file (generate the output), and start exiting (which flushes the output file) before the new data shows up.

1 members found this post helpful.