sedparse

GNU sed's parser translated from C to Python


Keywords
parser, sed-scripts
License
GPL-3.0
Install
pip install sedparse==0.1.2

Documentation

sedparse

  • Author: Aurelio Jargas
  • License: GPLv3
  • Tested with Python 2.7, 3.4, 3.5, 3.6, 3.7 and 3.8 (see .travis.yml)

A translation from C to Python of GNU sed's parser for sed scripts.

After running sedparse in your sed script, the resulting "AST" is available in different formats:

  • List of objects (translated C structs)
  • List of dictionaries
  • JSON

For a complete reference on how the different sed commands are mapped by the parser, see:

About the translation

I copied the original code in C and translated everything to Python, line by line.

To make it feasible to keep this code updated with future GNU sed code, this is a literal translation, trying to mimic as much as possible of the original code. That includes using the same API, same logic, same variable and method names and same data structures. Pythonic code? Sorry, not here.

The accuracy of the parser is checked by extensive unit tests in tests/.

Sedparse was translated from this GNU sed version:

http://git.savannah.gnu.org/cgit/sed.git/commit/?id=a9cb52bcf39f0ee307301ac73c11acb24372b9d8

commit a9cb52bcf39f0ee307301ac73c11acb24372b9d8
Author: Assaf Gordon <assafgordon@gmail.com>
Date:   Sun Jun 2 01:14:00 2019 -0600

Note that this is not a full GNU sed implementation. Only the parser for sed scripts was translated. Check https://github.com/GillesArcas/PythonSed for a working sed in Python.

Sedparse extensions to the original parser

  • Preserves comments
  • Preserves blank lines between commands
  • Preserves original flags for the s command
  • Preserves original flags for regex addresses

Installation

pip install --user sedparse
sedparse --help

Alternatively, you can just download and run the sedparse.py file, since it is self-contained with no external dependencies.

Usage from the command line

The informed sed script will be parsed and checked for syntax errors. If everything is fine, a JSON representation of the script is sent to STDOUT.

Just like in sed, you can inform the sed script using one or more -e options:

$ sedparse -e "s/foo/bar/g" -e "5d"
[
    {
        "cmd": "s",
        "line": 1,
        "x": {
            "cmd_subst": {
                "regx": {
                    "flags": "g",
                    "pattern": "foo",
                    "slash": "/"
                },
                "replacement": {
                    "text": "bar"
                }
            }
        }
    },
    {
        "a1": {
            "addr_number": 5,
            "addr_type": 3
        },
        "cmd": "d",
        "line": 1
    }
]
$

Or you can inform the sed script as a file argument using -f:

$ echo '1,10!d' > head.sed
$ sedparse -f head.sed
[
    {
        "a1": {
            "addr_number": 1,
            "addr_type": 3
        },
        "a2": {
            "addr_number": 10,
            "addr_type": 3
        },
        "addr_bang": true,
        "cmd": "d",
        "line": 1
    }
]
$ rm head.sed
$

Or even as text coming from STDIN when using the special - file:

$ echo '\EXTREMITIES' | sedparse -f -
[
    {
        "a1": {
            "addr_regex": {
                "flags": "MI",
                "pattern": "XTR",
                "slash": "E"
            },
            "addr_type": 2
        },
        "cmd": "T",
        "line": 1,
        "x": {
            "label_name": "IES"
        }
    }
]
$

Usage as a Python module

Use sedparse.compile_string() to parse a string as a sed script. You must inform a list that will be appended in-place with the parsed commands.

>>> import sedparse
>>> sedscript = """\
... 11,/foo/ {
...     $!N
...     s/\\n/-/gi
... }
... """
>>> parsed = []
>>> sedparse.compile_string(parsed, sedscript)
>>>

Each sed command is represented by a struct_sed_cmd instance.

>>> import pprint
>>> pprint.pprint(parsed)  # doctest:+ELLIPSIS
[struct_sed_cmd(line=1, cmd='{', ...),
 struct_sed_cmd(line=2, cmd='N', ...),
 struct_sed_cmd(line=3, cmd='s', ...),
 struct_sed_cmd(line=4, cmd='}', ...)]
>>>

You can str() each command, or any of its inner structs, to get their "human readable" representation.

>>> [str(x) for x in parsed]
['11,/foo/ {', '$ !N', 's/\\n/-/gi', '}']
>>> str(parsed[0])
'11,/foo/ {'
>>> str(parsed[0].a1)
'11'
>>> str(parsed[0].a2)
'/foo/'
>>>

Use .to_dict() to convert a command into a Python dictionary.

>>> cmd_n = parsed[1]
>>> str(cmd_n)
'$ !N'
>>> pprint.pprint(cmd_n.to_dict())
{'a1': {'addr_type': 7}, 'addr_bang': True, 'cmd': 'N', 'line': 2}
>>>
>>> pprint.pprint(cmd_n.to_dict(remove_empty=False))
{'a1': {'addr_number': 0, 'addr_regex': None, 'addr_step': 0, 'addr_type': 7},
 'a2': None,
 'addr_bang': True,
 'cmd': 'N',
 'line': 2,
 'x': {'cmd_subst': {'outf': {'name': ''},
                     'regx': {'flags': '', 'pattern': '', 'slash': ''},
                     'replacement': {'text': ''}},
       'cmd_txt': {'text': []},
       'comment': '',
       'fname': '',
       'int_arg': -1,
       'label_name': ''}}
>>>

Use .to_json() to convert a command into JSON.

>>> print(cmd_n.to_json())
{
    "a1": {
        "addr_type": 7
    },
    "addr_bang": true,
    "cmd": "N",
    "line": 2
}
>>>

Have fun!

>>> [x.cmd for x in parsed]  # list of commands
['{', 'N', 's', '}']
>>> [str(x) for x in parsed if x.a1 is None]  # commands with no address
['s/\\n/-/gi', '}']
>>> [str(x) for x in parsed if x.addr_bang]  # commands with bang !
['$ !N']
>>> [x.x.comment for x in parsed if x.x.comment]  # extract all comments
[]
>>> [x.x.fname for x in parsed if x.cmd in "rRwW"]  # list of read/write filenames
[]
>>>