sed Code Blocks in Babel

The Unix utility sed (stream editor) is a non-interactive command-line text editor that can be used to filter text in order to extract part of a file or to substitute multiple occurrences of a string in a file.

Alternatives to sed include awk and Perl.

The sed utility was created more than 50 years ago and is currently available on most operating systems.

To enable Babel evaluation of sed code blocks, add a sexp to org-babel-load-languages as follows:

(org-babel-do-load-languages
 'org-babel-load-languages
 '((sed . t))) ;; this line configures Babel =sed= support

Examples that use the in-file header argument read this input file:

one two three, one two three
four three two one
one hundred

By default, sed code blocks return a table.

#+name: sed_default
#+begin_src sed :in-file ~/temp/sed-test.txt
s/one/ONE/g
#+end_src

#+RESULTS: sed_default
| ONE  | two     | three, | ONE | two | three |
| four | three   | two    | ONE |     |       |
| ONE  | hundred |        |     |     |       |

The header argument :results verbatim returns the contents of the filtered input.

#+name: sed_return
#+begin_src sed :in-file ~/temp/sed-test.txt :results verbatim
s/one/ONE/g
#+end_src

#+RESULTS: sed_return
: ONE two three, ONE two three
: four three two ONE
: ONE hundred

Command line arguments to sed can be passed with the header argument :cmd-line.

#+name: sed_silent
#+begin_src sed :in-file ~/temp/sed-test.txt :results verbatim :cmd-line -n
s/one/ONE/g
#+end_src

#+RESULTS: sed_silent

Here is the obligatory 'Hello, world!' example.

#+name: sed_hello
#+begin_src sed  :in-file ~/temp/sed-test.txt :results verbatim
# convert input text stream to "Hello, world!"
s/.*/Hello, world!/
q
#+end_src

#+RESULTS: sed_hello
: Hello, world!

The output from sed can be directed to a file with the header arguments :file and, optionally :output-dir. The header argument :results file inserts a link to the output file in the Org mode buffer.

Note the comment on the first line of the code block. Note also the last line, which instructs sed to quit.

#+name: sed_file
#+begin_src sed  :in-file ~/temp/sed-test.txt :results file :file sed-results.txt :output-dir ~/temp/
# convert input text stream to "Hello, world!"
s/.*/Hello, world!/
q
#+end_src

#+RESULTS: sed_file
[[file:~/temp/sed-results.txt]]

The file sed-results.txt looks like this:

Hello, world!

Nevertheless, sed is a stream editor designed to be used in a Unix pipeline. The following examples show how to use sed to edit a stream of text from the Org buffer.

This example uses a relatively new Org feature to name a piece of text, then sends it through the stdin header argument to sed.

Another way to achieve this refers to the results of source code block evaluation.

Preparation of this documentation was aided by Bruce Barnett's introduction and tutorial.