Go to file
Czar 65b41786a0
fix setup
2023-02-19 23:10:58 -03:00
PythonSed fix setup 2023-02-19 23:10:58 -03:00
tests fix setup 2023-02-19 23:10:58 -03:00
.travis.yml Fix version tests. 2020-03-22 19:02:00 +01:00
LICENSE Initial commit 2016-12-01 22:47:38 +01:00
README.md Update README.md 2021-04-24 22:49:07 +02:00
sed.html Update repository URL 2019-06-04 00:11:34 +02:00
setup.py fix setup 2023-02-19 23:10:58 -03:00



A full and working Python implementation of sed

Build Status Coverage Status


General Information

Usage as a command line utility

Usage as a Python module

Sed dialect



To do list

General Information

Important note

pythonsed has been rewrited and completed by Frank Schäckermann. This new version has its own repository and is very close to be fully compatible with GNU sed 4.8. Any comment or pull request should be addressed to Frank.


pythonsed is a full and working Python implementation of sed. Its reference is GNU sed 4.2 of which it implements almost all commands and features. It may be used as a command line utility or it can be used as a module to bring sed functionality to Python scripts.

A complete set of tests is available as well as a testing utility. These tests include scripts from various origins and cover all aspects of sed functionalities.


Version Status
Python 3.7 Fully compatible except s///g with zero length matches. See this question at stackoverflow
Python 3 Fully compatible
Python 2.7.4 and above Fully compatible
Python 2.7 to Python 2.7.3 Fully compatible except regexps of the form ((.*)*). This causes one of the script from Chang suite to fail.
Python 2.6 Fully compatible except regexps of the form ((.*)*). argparse module must be installed.
Python 2.5 and below Not tested

Compatibility status applies also to the testing utility test-suite.py.


PythonSed is released under the MIT license.


To install, just clone or download the repository zip file and run the setup in download directory:

pip install .

This installs a command line utility named pythonsed and a package named PythonSed.

Usage as a command line utility

pythonsed is as console program receiving information from the command line. The format of the command line is:

pythonsed [options] -e<script expression> <input text file>

pythonsed [options] -f<script file> <input text file>

Note that pythonsed accepts only one script file or expression, and only one input file. options may be one or both of:

-n disable automatic printing

-ruse extended regular expressions

pythonsed may also use redirection to receive its input or send its output with the usual syntax:

cat myfile | pythonsed -f myscript1.sed | pythonsed -f myscript2.sed > myresultfile

It is also possible for pythonsed to receive its input from the keyboard by omitting any input file:

pythonsed -f myscript.sed

Usage as a Python module

An example covering all necessary symbols:

from PythonSed import Sed, SedException

sed = Sed()
    sed.no_autoprint = True
    sed.regexp_extended = False
except SedException as e:

sed.apply() input parameter may be a string (which is interpreted as a filename) or file-like object (including streams). Note that sed.apply() returns the list of lines printed by the script. As a default, these lines are printed to stdout. sed.apply() has an output parameter which enables to inhibit printing the lines (output=None) or enables to redirect the output to some text file (output='somefile.txt') or to a file-like object (including streams). Note also that if myinput or myoutput are file-like objects, they must be closed by the caller.

The script may also be read from a string by using sed.load_string(my_script_string).

sed dialect

PythonSed implements all standard commands and regular expression features of sed. Its reference is GNU sed 4.2. It implements almost all its features except the most specific ones.

GNU sed manual page can serve as a reference for PythonSed given the differences described in the following.


numberstandard behavior
$standard behavior
/regexp/standard behavior
\%regexp%standard behavior
address,addressstandard behavior
address!standard behavior
0,/regexp/not implemented
first~stepnot implemented
addr1,+Nnot implemented
addr1,~Nnot implemented

Regular expressions

charstandard behavior
*standard behavior
\+standard behavior
\?standard behavior
\{i\} \{i,j\} \{i,\}standard behavior
\(regexp\)standard behavior
.standard behavior
^standard behavior. When not at start of regexp, matches as itself
$standard behavior. When not at end of regexp, matches as itself
[list] [^list]standard behavior. [.ch.], [=a=], [:space:] are not implemented
regexp1\|regexp2standard behavior
regexp1regexp2standard behavior
\digitstandard behavior (back reference)
\n \tstandard behavior (extensions \s\S etc. are not handled)
\charstandard behavior (disable special regexp characters)

Note that for any combination of quantifiers (*, +, ?, {}), consecutive quantifiers or a quantifier starting a regexp will launch an error. This is true in basic or extended regular expression modes.

Extended regular expressions

Using the -r switch enables to simplify regular expressions by removing the antislah character before the special characters +, ?, (, ), |, { and }. If these characters must appear as regular characters in a regexp, they must be slashed.


Compliant(including one liner syntax and double address extensions)
b labelCompliant 
: labelCompliant 
Compliant(including single line and double address extensions)
=Compliant(including double address extension)
Compliant(including single line and double address extensions)
lCompliant(length parameter not implemented)
qCompliant(except exit code extension)
r filenameCompliant(including double address extension but not reading from stdin)
sCompliant(except escape sequences in replacement (\L, \l, \U, \u, \E), modifiers e and M/m, and combination of modifier g and number)
t labelCompliant 
w filenameCompliant(including double address extension but not writing to stdout or stderr)
#Compliant(comments start anywhere in the line.)

Compliant means compliant with GNU sed description.

The other commands specific to GNU sed are not implemented.



The working of PythonSed is tested and compared to the behavior of GNU sed with a set of tests and a testing utility.

The tests are either coded in text files with .suite extension or may be stored in test directories as standard sed scripts.

The test suites are:

unit.suitea text filecontaining unitary tests
chang.suitea text file containing scripts from Roger Chang web site
test-suite1a set of scripts from GNU sed test suite
test-suite2a set of scripts from the seder's grab-bag, Rosetta code web site and GitHub (lisp!)
test-suite3additional unitary tests better stored in a folder with some extra data text files
test-suite4a set of scripts from the sed $HOME
Note that the goal of these tests is not to check the correctness of the scripts but to verify that `sed.py` and GNU sed have the same behavior.

Testing utility

Tests are launched and checked with the test-suite.py Python script. This script uses either PythonSed package to run the sed scripts, or any sed executable. This enables to compare the working of PythonSed with the one of GNU sed.

The calling syntax is:

test-suite.py <testsuite> [number] [-b executable] [-x list of script references]
testsuite either a text file with .suite extension or a test directory
number an optional reference number of a test, when present only this tests is run
executable an optional name or path of a sed executable to use for testing
list of script references an optional list of tests to exclude for instance when a feature is not implemented. A script reference is either the title of the test for tests stored in modules, or the the name of the script file.

Text file test suites

When tests are stored in a text file (with .suite extension), they are made of four elements:

  • the title of the test
  • the script itself
  • the input list of lines
  • the expected result

The four elements of a test are separated with lines made of three identical characters, for instance:

Test substitution with global flag
In Xanadu did Kubhla Khan
In XANadu did Kubhla KhAN

Note also that:

  • the script section may be empty, enabling to test a script on various data without repeating the script.
  • The input and output sections may be empty, enabling to test various scripts on the same data, without repeating the data.
  • Flags are set with a comment on the first line. As usual, #n stops autoprint mode and extended regexp mode is set with #r or #nr.
  • The expected result may be ??? when the test has no result and ends with an error.
  • All text outside the test, i.e. before first delimiter or after last delimiter, is ignored and acts like a comment.

Directory test suites

When tests are stored in a directory, they are represented by three or four files with same name but different extensions:

  • the script itself, with '.sed' extension
  • the input of the script, with '.inp' extension
  • the expected result of the script, with '.good' extension
  • possibly a file, with '.flags' extension, containing the sed switches -n and/or -r.

Some other files may be used when using reading or writing commands in scripts. In that case, the expected written files must be named with extension '.wgoodN' where N is the number of the expected written file.


A python implementation of sed has to face legitimate questions about timing. Fortunately, results are not bad. Unfortunately, they seem correlated with version number. Timings are given in seconds.

Platform GNU sed 4.2.1 sed.py python 2.6 sed.py python 2.7 sed.py python 3.4
Windows7, Intel Xeon 3.2 GHz, 6 Gb RAM 19.4 19.1 22.6 26.9
Windows XP, Intel Pentium4 3.2 GHz, 4 Gb RAM 47.5 50.7 56.5 71.2
Linux, Intel Pentium4 3.2 GHz, 4 Gb RAM - - 51.0 -

Test conditions:

  • Only script files are used (scripts from folders testsuiteN). This is to avoid measuring the time to extract scripts, inputs and results from .suite files.
  • The given values are averaged from three consecutive test runs.

To do list

At one moment, one has to decide what will be in the release to come, and what can be delayed. Here are some features which would be nice to have but can be delayed to a future version.

  • Better POSIX compliance:

  • multiple scripts on the command line (-e, -f)

  • multiple input file

  • character classes

  • Better error handling (display of the number of the line in error)

  • Better error handling when testing (the error message could be tested)

  • Use PythonSed as a basis for a sed debugger.

  • ...